diff options
Diffstat (limited to 'src')
| -rw-r--r-- | src/ChangeLog | 38 | ||||
| -rw-r--r-- | src/coding.c | 5 | ||||
| -rw-r--r-- | src/config.in | 9 | ||||
| -rw-r--r-- | src/fns.c | 20 | ||||
| -rw-r--r-- | src/gtkutil.c | 2 | ||||
| -rw-r--r-- | src/minibuf.c | 150 | ||||
| -rw-r--r-- | src/regex.c | 6335 | ||||
| -rw-r--r-- | src/regex.h | 576 | ||||
| -rw-r--r-- | src/tparam.c | 14 | ||||
| -rw-r--r-- | src/xfaces.c | 22 |
10 files changed, 173 insertions, 6998 deletions
diff --git a/src/ChangeLog b/src/ChangeLog index f6959b45efa..4adadd84db1 100644 --- a/src/ChangeLog +++ b/src/ChangeLog | |||
| @@ -1,3 +1,41 @@ | |||
| 1 | 2003-12-24 Jan Dj,Ad(Brv <jan.h.d@swipnet.se> | ||
| 2 | |||
| 3 | * gtkutil.c (xg_frame_set_char_size): Call x_wm_set_size_hint. | ||
| 4 | |||
| 5 | * xfaces.c (lface_fully_specified_p): Take into account that | ||
| 6 | MAC OS always have unspecified stipple. | ||
| 7 | |||
| 8 | 2003-12-24 Thien-Thi Nguyen <ttn@gnu.org> | ||
| 9 | |||
| 10 | * tparam.c (tparam1): Add handling for `%pN', which | ||
| 11 | means use param N for the next substitution. | ||
| 12 | |||
| 13 | 2003-12-24 Thien-Thi Nguyen <ttn@gnu.org> | ||
| 14 | |||
| 15 | * xfaces.c (Fcolor_gray_p): Fix omission bug: | ||
| 16 | In case `frame' is nil, consult the selected frame. | ||
| 17 | (Fcolor_supported_p): Likewise. | ||
| 18 | |||
| 19 | 2003-12-23 Luc Teirlinck <teirllm@auburn.edu> | ||
| 20 | |||
| 21 | * fns.c (Frandom, Fstring_make_multibyte, Fset_char_table_range): | ||
| 22 | Doc fixes. | ||
| 23 | |||
| 24 | * minibuf.c (read_minibuf): Allow INITIAL to be a cons of a string | ||
| 25 | and an integer. Adapt the introductory comment accordingly. | ||
| 26 | (Fread_from_minibuffer): Delete code moved into read_minibuf. | ||
| 27 | Doc fix. | ||
| 28 | (Fread_minibuffer, Fread_no_blanks_input): Adapt to changes in | ||
| 29 | read_minibuf. | ||
| 30 | (Fcompleting_read): Delete code moved into read_minibuf. | ||
| 31 | (Ftest_completion): Make it handle obarrays and hash tables | ||
| 32 | correctly. | ||
| 33 | |||
| 34 | 2003-12-03 Kenichi Handa <handa@m17n.org> | ||
| 35 | |||
| 36 | * coding.c (decode_coding_iso2022): Fix for preserving UTF-8 | ||
| 37 | encoding sequence. | ||
| 38 | |||
| 1 | 2003-12-01 Kenichi Handa <handa@m17n.org> | 39 | 2003-12-01 Kenichi Handa <handa@m17n.org> |
| 2 | 40 | ||
| 3 | * composite.c (syms_of_composite): Don't make the compostion hash | 41 | * composite.c (syms_of_composite): Don't make the compostion hash |
diff --git a/src/coding.c b/src/coding.c index 910fc8faa68..3f11c6f41fb 100644 --- a/src/coding.c +++ b/src/coding.c | |||
| @@ -2125,7 +2125,10 @@ decode_coding_iso2022 (coding, source, destination, src_bytes, dst_bytes) | |||
| 2125 | && src + 1 < src_end | 2125 | && src + 1 < src_end |
| 2126 | && src[0] == '%' | 2126 | && src[0] == '%' |
| 2127 | && src[1] == '@') | 2127 | && src[1] == '@') |
| 2128 | break; | 2128 | { |
| 2129 | src += 2; | ||
| 2130 | break; | ||
| 2131 | } | ||
| 2129 | d += CHAR_STRING (c1, d), produced_chars++; | 2132 | d += CHAR_STRING (c1, d), produced_chars++; |
| 2130 | } | 2133 | } |
| 2131 | if (d + 3 > (dst_bytes ? dst_end : src)) | 2134 | if (d + 3 > (dst_bytes ? dst_end : src)) |
diff --git a/src/config.in b/src/config.in index 5024712ac43..0fb4b3f3444 100644 --- a/src/config.in +++ b/src/config.in | |||
| @@ -549,6 +549,9 @@ Boston, MA 02111-1307, USA. */ | |||
| 549 | /* Define to 1 if you have the <sys/select.h> header file. */ | 549 | /* Define to 1 if you have the <sys/select.h> header file. */ |
| 550 | #undef HAVE_SYS_SELECT_H | 550 | #undef HAVE_SYS_SELECT_H |
| 551 | 551 | ||
| 552 | /* Define to 1 if you have the <sys/socket.h> header file. */ | ||
| 553 | #undef HAVE_SYS_SOCKET_H | ||
| 554 | |||
| 552 | /* Define to 1 if you have the <sys/soundcard.h> header file. */ | 555 | /* Define to 1 if you have the <sys/soundcard.h> header file. */ |
| 553 | #undef HAVE_SYS_SOUNDCARD_H | 556 | #undef HAVE_SYS_SOUNDCARD_H |
| 554 | 557 | ||
| @@ -748,9 +751,9 @@ Boston, MA 02111-1307, USA. */ | |||
| 748 | /* If using the C implementation of alloca, define if you know the | 751 | /* If using the C implementation of alloca, define if you know the |
| 749 | direction of stack growth for your system; otherwise it will be | 752 | direction of stack growth for your system; otherwise it will be |
| 750 | automatically deduced at run-time. | 753 | automatically deduced at run-time. |
| 751 | STACK_DIRECTION > 0 => grows toward higher addresses | 754 | STACK_DIRECTION > 0 => grows toward higher addresses |
| 752 | STACK_DIRECTION < 0 => grows toward lower addresses | 755 | STACK_DIRECTION < 0 => grows toward lower addresses |
| 753 | STACK_DIRECTION = 0 => direction of growth unknown */ | 756 | STACK_DIRECTION = 0 => direction of growth unknown */ |
| 754 | #undef STACK_DIRECTION | 757 | #undef STACK_DIRECTION |
| 755 | 758 | ||
| 756 | /* Define to 1 if you have the ANSI C header files. */ | 759 | /* Define to 1 if you have the ANSI C header files. */ |
| @@ -95,7 +95,7 @@ DEFUN ("identity", Fidentity, Sidentity, 1, 1, 0, | |||
| 95 | DEFUN ("random", Frandom, Srandom, 0, 1, 0, | 95 | DEFUN ("random", Frandom, Srandom, 0, 1, 0, |
| 96 | doc: /* Return a pseudo-random number. | 96 | doc: /* Return a pseudo-random number. |
| 97 | All integers representable in Lisp are equally likely. | 97 | All integers representable in Lisp are equally likely. |
| 98 | On most systems, this is 28 bits' worth. | 98 | On most systems, this is 29 bits' worth. |
| 99 | With positive integer argument N, return random number in interval [0,N). | 99 | With positive integer argument N, return random number in interval [0,N). |
| 100 | With argument t, set the random number seed from the current time and pid. */) | 100 | With argument t, set the random number seed from the current time and pid. */) |
| 101 | (n) | 101 | (n) |
| @@ -1099,8 +1099,14 @@ string_make_unibyte (string) | |||
| 1099 | DEFUN ("string-make-multibyte", Fstring_make_multibyte, Sstring_make_multibyte, | 1099 | DEFUN ("string-make-multibyte", Fstring_make_multibyte, Sstring_make_multibyte, |
| 1100 | 1, 1, 0, | 1100 | 1, 1, 0, |
| 1101 | doc: /* Return the multibyte equivalent of STRING. | 1101 | doc: /* Return the multibyte equivalent of STRING. |
| 1102 | The function `unibyte-char-to-multibyte' is used to convert | 1102 | If STRING is unibyte and contains non-ASCII characters, the function |
| 1103 | each unibyte character to a multibyte character. */) | 1103 | `unibyte-char-to-multibyte' is used to convert each unibyte character |
| 1104 | to a multibyte character. In this case, the returned string is a | ||
| 1105 | newly created string with no text properties. If STRING is multibyte | ||
| 1106 | or entirely ASCII, it is returned unchanged. In particular, when | ||
| 1107 | STRING is unibyte and entirely ASCII, the returned string is unibyte. | ||
| 1108 | \(When the characters are all ASCII, Emacs primitives will treat the | ||
| 1109 | string the same way whether it is unibyte or multibyte.) */) | ||
| 1104 | (string) | 1110 | (string) |
| 1105 | Lisp_Object string; | 1111 | Lisp_Object string; |
| 1106 | { | 1112 | { |
| @@ -2482,9 +2488,9 @@ a character set name, or a character code. */) | |||
| 2482 | DEFUN ("set-char-table-range", Fset_char_table_range, Sset_char_table_range, | 2488 | DEFUN ("set-char-table-range", Fset_char_table_range, Sset_char_table_range, |
| 2483 | 3, 3, 0, | 2489 | 3, 3, 0, |
| 2484 | doc: /* Set the value in CHAR-TABLE for a range of characters RANGE to VALUE. | 2490 | doc: /* Set the value in CHAR-TABLE for a range of characters RANGE to VALUE. |
| 2485 | RANGE should be t (for all characters), nil (for the default value) | 2491 | RANGE should be t (for all characters), nil (for the default value), |
| 2486 | a vector which identifies a character set or a row of a character set, | 2492 | a character set, a vector which identifies a character set, a row of a |
| 2487 | a coding system, or a character code. */) | 2493 | character set, or a character code. Return VALUE. */) |
| 2488 | (char_table, range, value) | 2494 | (char_table, range, value) |
| 2489 | Lisp_Object char_table, range, value; | 2495 | Lisp_Object char_table, range, value; |
| 2490 | { | 2496 | { |
| @@ -5648,7 +5654,7 @@ This applies to commands from menus and tool bar buttons. The value of | |||
| 5648 | `use-dialog-box' takes precedence over this variable, so a file dialog is only | 5654 | `use-dialog-box' takes precedence over this variable, so a file dialog is only |
| 5649 | used if both `use-dialog-box' and this variable are non-nil. */); | 5655 | used if both `use-dialog-box' and this variable are non-nil. */); |
| 5650 | use_file_dialog = 1; | 5656 | use_file_dialog = 1; |
| 5651 | 5657 | ||
| 5652 | defsubr (&Sidentity); | 5658 | defsubr (&Sidentity); |
| 5653 | defsubr (&Srandom); | 5659 | defsubr (&Srandom); |
| 5654 | defsubr (&Slength); | 5660 | defsubr (&Slength); |
diff --git a/src/gtkutil.c b/src/gtkutil.c index 4fc3351ab7e..4d0b50573f7 100644 --- a/src/gtkutil.c +++ b/src/gtkutil.c | |||
| @@ -515,7 +515,7 @@ xg_frame_set_char_size (f, cols, rows) | |||
| 515 | gtk_window_resize (GTK_WINDOW (FRAME_GTK_OUTER_WIDGET (f)), | 515 | gtk_window_resize (GTK_WINDOW (FRAME_GTK_OUTER_WIDGET (f)), |
| 516 | pixelwidth, pixelheight); | 516 | pixelwidth, pixelheight); |
| 517 | xg_resize_widgets (f, pixelwidth, pixelheight); | 517 | xg_resize_widgets (f, pixelwidth, pixelheight); |
| 518 | 518 | x_wm_set_size_hint (f, 0, 0); | |
| 519 | SET_FRAME_GARBAGED (f); | 519 | SET_FRAME_GARBAGED (f); |
| 520 | cancel_mouse_face (f); | 520 | cancel_mouse_face (f); |
| 521 | } | 521 | } |
diff --git a/src/minibuf.c b/src/minibuf.c index 542fdd54f1d..4c98f7ba708 100644 --- a/src/minibuf.c +++ b/src/minibuf.c | |||
| @@ -404,10 +404,15 @@ minibuffer_completion_contents () | |||
| 404 | return make_buffer_string (prompt_end, PT, 1); | 404 | return make_buffer_string (prompt_end, PT, 1); |
| 405 | } | 405 | } |
| 406 | 406 | ||
| 407 | /* Read from the minibuffer using keymap MAP, initial contents INITIAL | 407 | /* Read from the minibuffer using keymap MAP and initial contents INITIAL, |
| 408 | (a string), putting point minus BACKUP_N bytes from the end of INITIAL, | 408 | putting point minus BACKUP_N bytes from the end of INITIAL, |
| 409 | prompting with PROMPT (a string), using history list HISTVAR | 409 | prompting with PROMPT (a string), using history list HISTVAR |
| 410 | with initial position HISTPOS. (BACKUP_N should be <= 0.) | 410 | with initial position HISTPOS. INITIAL should be a string or a |
| 411 | cons of a string and an integer. BACKUP_N should be <= 0, or | ||
| 412 | Qnil, which is equivalent to 0. If INITIAL is a cons, BACKUP_N is | ||
| 413 | ignored and replaced with an integer that puts point N characters | ||
| 414 | from the beginning of INITIAL, where N is the CDR of INITIAL, or at | ||
| 415 | the beginning of INITIAL if N <= 0. | ||
| 411 | 416 | ||
| 412 | Normally return the result as a string (the text that was read), | 417 | Normally return the result as a string (the text that was read), |
| 413 | but if EXPFLAG is nonzero, read it and return the object read. | 418 | but if EXPFLAG is nonzero, read it and return the object read. |
| @@ -419,7 +424,7 @@ minibuffer_completion_contents () | |||
| 419 | 424 | ||
| 420 | If ALLOW_PROPS is nonzero, we do not throw away text properties. | 425 | If ALLOW_PROPS is nonzero, we do not throw away text properties. |
| 421 | 426 | ||
| 422 | if INHERIT_INPUT_METHOD is nonzeor, the minibuffer inherit the | 427 | if INHERIT_INPUT_METHOD is nonzero, the minibuffer inherits the |
| 423 | current input method. */ | 428 | current input method. */ |
| 424 | 429 | ||
| 425 | static Lisp_Object | 430 | static Lisp_Object |
| @@ -441,6 +446,7 @@ read_minibuf (map, initial, prompt, backup_n, expflag, | |||
| 441 | Lisp_Object mini_frame, ambient_dir, minibuffer, input_method; | 446 | Lisp_Object mini_frame, ambient_dir, minibuffer, input_method; |
| 442 | struct gcpro gcpro1, gcpro2, gcpro3, gcpro4, gcpro5; | 447 | struct gcpro gcpro1, gcpro2, gcpro3, gcpro4, gcpro5; |
| 443 | Lisp_Object enable_multibyte; | 448 | Lisp_Object enable_multibyte; |
| 449 | int pos = INTEGERP (backup_n) ? XINT (backup_n) : 0; | ||
| 444 | 450 | ||
| 445 | /* String to add to the history. */ | 451 | /* String to add to the history. */ |
| 446 | Lisp_Object histstring; | 452 | Lisp_Object histstring; |
| @@ -456,6 +462,27 @@ read_minibuf (map, initial, prompt, backup_n, expflag, | |||
| 456 | cancel_hourglass (); | 462 | cancel_hourglass (); |
| 457 | #endif | 463 | #endif |
| 458 | 464 | ||
| 465 | if (!NILP (initial)) | ||
| 466 | { | ||
| 467 | if (CONSP (initial)) | ||
| 468 | { | ||
| 469 | backup_n = Fcdr (initial); | ||
| 470 | initial = Fcar (initial); | ||
| 471 | CHECK_STRING (initial); | ||
| 472 | if (!NILP (backup_n)) | ||
| 473 | { | ||
| 474 | CHECK_NUMBER (backup_n); | ||
| 475 | /* Convert to distance from end of input. */ | ||
| 476 | if (XINT (backup_n) < 1) | ||
| 477 | /* A number too small means the beginning of the string. */ | ||
| 478 | pos = - SCHARS (initial); | ||
| 479 | else | ||
| 480 | pos = XINT (backup_n) - 1 - SCHARS (initial); | ||
| 481 | } | ||
| 482 | } | ||
| 483 | else | ||
| 484 | CHECK_STRING (initial); | ||
| 485 | } | ||
| 459 | val = Qnil; | 486 | val = Qnil; |
| 460 | ambient_dir = current_buffer->directory; | 487 | ambient_dir = current_buffer->directory; |
| 461 | input_method = Qnil; | 488 | input_method = Qnil; |
| @@ -482,7 +509,8 @@ read_minibuf (map, initial, prompt, backup_n, expflag, | |||
| 482 | 509 | ||
| 483 | if (noninteractive) | 510 | if (noninteractive) |
| 484 | { | 511 | { |
| 485 | val = read_minibuf_noninteractive (map, initial, prompt, backup_n, | 512 | val = read_minibuf_noninteractive (map, initial, prompt, |
| 513 | make_number (pos), | ||
| 486 | expflag, histvar, histpos, defalt, | 514 | expflag, histvar, histpos, defalt, |
| 487 | allow_props, inherit_input_method); | 515 | allow_props, inherit_input_method); |
| 488 | UNGCPRO; | 516 | UNGCPRO; |
| @@ -633,8 +661,7 @@ read_minibuf (map, initial, prompt, backup_n, expflag, | |||
| 633 | if (!NILP (initial)) | 661 | if (!NILP (initial)) |
| 634 | { | 662 | { |
| 635 | Finsert (1, &initial); | 663 | Finsert (1, &initial); |
| 636 | if (INTEGERP (backup_n)) | 664 | Fforward_char (make_number (pos)); |
| 637 | Fforward_char (backup_n); | ||
| 638 | } | 665 | } |
| 639 | 666 | ||
| 640 | clear_message (1, 1); | 667 | clear_message (1, 1); |
| @@ -884,8 +911,9 @@ Fifth arg HIST, if non-nil, specifies a history list | |||
| 884 | which INITIAL-CONTENTS corresponds to). | 911 | which INITIAL-CONTENTS corresponds to). |
| 885 | Positions are counted starting from 1 at the beginning of the list. | 912 | Positions are counted starting from 1 at the beginning of the list. |
| 886 | Sixth arg DEFAULT-VALUE is the default value. If non-nil, it is available | 913 | Sixth arg DEFAULT-VALUE is the default value. If non-nil, it is available |
| 887 | for history commands; but `read-from-minibuffer' does NOT return DEFAULT-VALUE | 914 | for history commands; but, unless READ is non-nil, `read-from-minibuffer' |
| 888 | if the user enters empty input! It returns the empty string. | 915 | does NOT return DEFAULT-VALUE if the user enters empty input! It returns |
| 916 | the empty string. | ||
| 889 | Seventh arg INHERIT-INPUT-METHOD, if non-nil, means the minibuffer inherits | 917 | Seventh arg INHERIT-INPUT-METHOD, if non-nil, means the minibuffer inherits |
| 890 | the current input method and the setting of `enable-multibyte-characters'. | 918 | the current input method and the setting of `enable-multibyte-characters'. |
| 891 | If the variable `minibuffer-allow-text-properties' is non-nil, | 919 | If the variable `minibuffer-allow-text-properties' is non-nil, |
| @@ -895,33 +923,10 @@ If the variable `minibuffer-allow-text-properties' is non-nil, | |||
| 895 | Lisp_Object prompt, initial_contents, keymap, read, hist, default_value; | 923 | Lisp_Object prompt, initial_contents, keymap, read, hist, default_value; |
| 896 | Lisp_Object inherit_input_method; | 924 | Lisp_Object inherit_input_method; |
| 897 | { | 925 | { |
| 898 | int pos = 0; | 926 | Lisp_Object histvar, histpos, val; |
| 899 | Lisp_Object histvar, histpos, position, val; | ||
| 900 | struct gcpro gcpro1; | 927 | struct gcpro gcpro1; |
| 901 | 928 | ||
| 902 | position = Qnil; | ||
| 903 | |||
| 904 | CHECK_STRING (prompt); | 929 | CHECK_STRING (prompt); |
| 905 | if (!NILP (initial_contents)) | ||
| 906 | { | ||
| 907 | if (CONSP (initial_contents)) | ||
| 908 | { | ||
| 909 | position = Fcdr (initial_contents); | ||
| 910 | initial_contents = Fcar (initial_contents); | ||
| 911 | } | ||
| 912 | CHECK_STRING (initial_contents); | ||
| 913 | if (!NILP (position)) | ||
| 914 | { | ||
| 915 | CHECK_NUMBER (position); | ||
| 916 | /* Convert to distance from end of input. */ | ||
| 917 | if (XINT (position) < 1) | ||
| 918 | /* A number too small means the beginning of the string. */ | ||
| 919 | pos = - SCHARS (initial_contents); | ||
| 920 | else | ||
| 921 | pos = XINT (position) - 1 - SCHARS (initial_contents); | ||
| 922 | } | ||
| 923 | } | ||
| 924 | |||
| 925 | if (NILP (keymap)) | 930 | if (NILP (keymap)) |
| 926 | keymap = Vminibuffer_local_map; | 931 | keymap = Vminibuffer_local_map; |
| 927 | else | 932 | else |
| @@ -944,7 +949,7 @@ If the variable `minibuffer-allow-text-properties' is non-nil, | |||
| 944 | 949 | ||
| 945 | GCPRO1 (default_value); | 950 | GCPRO1 (default_value); |
| 946 | val = read_minibuf (keymap, initial_contents, prompt, | 951 | val = read_minibuf (keymap, initial_contents, prompt, |
| 947 | make_number (pos), !NILP (read), | 952 | Qnil, !NILP (read), |
| 948 | histvar, histpos, default_value, | 953 | histvar, histpos, default_value, |
| 949 | minibuffer_allow_text_properties, | 954 | minibuffer_allow_text_properties, |
| 950 | !NILP (inherit_input_method)); | 955 | !NILP (inherit_input_method)); |
| @@ -960,8 +965,6 @@ is a string to insert in the minibuffer before reading. */) | |||
| 960 | Lisp_Object prompt, initial_contents; | 965 | Lisp_Object prompt, initial_contents; |
| 961 | { | 966 | { |
| 962 | CHECK_STRING (prompt); | 967 | CHECK_STRING (prompt); |
| 963 | if (!NILP (initial_contents)) | ||
| 964 | CHECK_STRING (initial_contents); | ||
| 965 | return read_minibuf (Vminibuffer_local_map, initial_contents, | 968 | return read_minibuf (Vminibuffer_local_map, initial_contents, |
| 966 | prompt, Qnil, 1, Qminibuffer_history, | 969 | prompt, Qnil, 1, Qminibuffer_history, |
| 967 | make_number (0), Qnil, 0, 0); | 970 | make_number (0), Qnil, 0, 0); |
| @@ -1012,9 +1015,6 @@ the current input method and the setting of `enable-multibyte-characters'. */) | |||
| 1012 | Lisp_Object prompt, initial, inherit_input_method; | 1015 | Lisp_Object prompt, initial, inherit_input_method; |
| 1013 | { | 1016 | { |
| 1014 | CHECK_STRING (prompt); | 1017 | CHECK_STRING (prompt); |
| 1015 | if (! NILP (initial)) | ||
| 1016 | CHECK_STRING (initial); | ||
| 1017 | |||
| 1018 | return read_minibuf (Vminibuffer_local_ns_map, initial, prompt, Qnil, | 1018 | return read_minibuf (Vminibuffer_local_ns_map, initial, prompt, Qnil, |
| 1019 | 0, Qminibuffer_history, make_number (0), Qnil, 0, | 1019 | 0, Qminibuffer_history, make_number (0), Qnil, 0, |
| 1020 | !NILP (inherit_input_method)); | 1020 | !NILP (inherit_input_method)); |
| @@ -1578,13 +1578,10 @@ Completion ignores case if the ambient value of | |||
| 1578 | Lisp_Object prompt, table, predicate, require_match, initial_input; | 1578 | Lisp_Object prompt, table, predicate, require_match, initial_input; |
| 1579 | Lisp_Object hist, def, inherit_input_method; | 1579 | Lisp_Object hist, def, inherit_input_method; |
| 1580 | { | 1580 | { |
| 1581 | Lisp_Object val, histvar, histpos, position; | 1581 | Lisp_Object val, histvar, histpos; |
| 1582 | Lisp_Object init; | ||
| 1583 | int pos = 0; | ||
| 1584 | int count = SPECPDL_INDEX (); | 1582 | int count = SPECPDL_INDEX (); |
| 1585 | struct gcpro gcpro1; | 1583 | struct gcpro gcpro1; |
| 1586 | 1584 | ||
| 1587 | init = initial_input; | ||
| 1588 | GCPRO1 (def); | 1585 | GCPRO1 (def); |
| 1589 | 1586 | ||
| 1590 | specbind (Qminibuffer_completion_table, table); | 1587 | specbind (Qminibuffer_completion_table, table); |
| @@ -1593,23 +1590,6 @@ Completion ignores case if the ambient value of | |||
| 1593 | EQ (require_match, Qt) ? Qnil : require_match); | 1590 | EQ (require_match, Qt) ? Qnil : require_match); |
| 1594 | last_exact_completion = Qnil; | 1591 | last_exact_completion = Qnil; |
| 1595 | 1592 | ||
| 1596 | position = Qnil; | ||
| 1597 | if (!NILP (init)) | ||
| 1598 | { | ||
| 1599 | if (CONSP (init)) | ||
| 1600 | { | ||
| 1601 | position = Fcdr (init); | ||
| 1602 | init = Fcar (init); | ||
| 1603 | } | ||
| 1604 | CHECK_STRING (init); | ||
| 1605 | if (!NILP (position)) | ||
| 1606 | { | ||
| 1607 | CHECK_NUMBER (position); | ||
| 1608 | /* Convert to distance from end of input. */ | ||
| 1609 | pos = XINT (position) - SCHARS (init); | ||
| 1610 | } | ||
| 1611 | } | ||
| 1612 | |||
| 1613 | if (SYMBOLP (hist)) | 1593 | if (SYMBOLP (hist)) |
| 1614 | { | 1594 | { |
| 1615 | histvar = hist; | 1595 | histvar = hist; |
| @@ -1628,7 +1608,7 @@ Completion ignores case if the ambient value of | |||
| 1628 | val = read_minibuf (NILP (require_match) | 1608 | val = read_minibuf (NILP (require_match) |
| 1629 | ? Vminibuffer_local_completion_map | 1609 | ? Vminibuffer_local_completion_map |
| 1630 | : Vminibuffer_local_must_match_map, | 1610 | : Vminibuffer_local_must_match_map, |
| 1631 | init, prompt, make_number (pos), 0, | 1611 | initial_input, prompt, Qnil, 0, |
| 1632 | histvar, histpos, def, 0, | 1612 | histvar, histpos, def, 0, |
| 1633 | !NILP (inherit_input_method)); | 1613 | !NILP (inherit_input_method)); |
| 1634 | 1614 | ||
| @@ -1650,7 +1630,7 @@ the values STRING, PREDICATE and `lambda'. */) | |||
| 1650 | (string, alist, predicate) | 1630 | (string, alist, predicate) |
| 1651 | Lisp_Object string, alist, predicate; | 1631 | Lisp_Object string, alist, predicate; |
| 1652 | { | 1632 | { |
| 1653 | Lisp_Object regexps, tem = Qnil; | 1633 | Lisp_Object regexps, tail, tem = Qnil; |
| 1654 | int i = 0; | 1634 | int i = 0; |
| 1655 | 1635 | ||
| 1656 | CHECK_STRING (string); | 1636 | CHECK_STRING (string); |
| @@ -1676,20 +1656,56 @@ the values STRING, PREDICATE and `lambda'. */) | |||
| 1676 | else | 1656 | else |
| 1677 | string = Fstring_make_multibyte (string); | 1657 | string = Fstring_make_multibyte (string); |
| 1678 | 1658 | ||
| 1679 | tem = oblookup (Vminibuffer_completion_table, | 1659 | tem = oblookup (alist, |
| 1680 | SDATA (string), | 1660 | SDATA (string), |
| 1681 | SCHARS (string), | 1661 | SCHARS (string), |
| 1682 | SBYTES (string)); | 1662 | SBYTES (string)); |
| 1683 | if (!SYMBOLP (tem)) | ||
| 1684 | return Qnil; | ||
| 1685 | } | 1663 | } |
| 1664 | |||
| 1665 | if (completion_ignore_case && !SYMBOLP (tem)) | ||
| 1666 | { | ||
| 1667 | for (i = XVECTOR (alist)->size - 1; i >= 0; i--) | ||
| 1668 | { | ||
| 1669 | tail = XVECTOR (alist)->contents[i]; | ||
| 1670 | if (SYMBOLP (tail)) | ||
| 1671 | while (1) | ||
| 1672 | { | ||
| 1673 | if (EQ((Fcompare_strings (string, make_number (0), Qnil, | ||
| 1674 | Fsymbol_name (tail), | ||
| 1675 | make_number (0) , Qnil, Qt)), | ||
| 1676 | Qt)) | ||
| 1677 | { | ||
| 1678 | tem = tail; | ||
| 1679 | break; | ||
| 1680 | } | ||
| 1681 | if (XSYMBOL (tail)->next == 0) | ||
| 1682 | break; | ||
| 1683 | XSETSYMBOL (tail, XSYMBOL (tail)->next); | ||
| 1684 | } | ||
| 1685 | } | ||
| 1686 | } | ||
| 1687 | |||
| 1688 | if (!SYMBOLP (tem)) | ||
| 1689 | return Qnil; | ||
| 1686 | } | 1690 | } |
| 1687 | else if (HASH_TABLE_P (alist)) | 1691 | else if (HASH_TABLE_P (alist)) |
| 1688 | { | 1692 | { |
| 1689 | i = hash_lookup (XHASH_TABLE (alist), string, NULL); | 1693 | struct Lisp_Hash_Table *h = XHASH_TABLE (alist); |
| 1694 | i = hash_lookup (h, string, NULL); | ||
| 1690 | if (i >= 0) | 1695 | if (i >= 0) |
| 1691 | tem = HASH_KEY (XHASH_TABLE (alist), i); | 1696 | tem = HASH_KEY (h, i); |
| 1692 | else | 1697 | else |
| 1698 | for (i = 0; i < HASH_TABLE_SIZE (h); ++i) | ||
| 1699 | if (!NILP (HASH_HASH (h, i)) && | ||
| 1700 | EQ (Fcompare_strings (string, make_number (0), Qnil, | ||
| 1701 | HASH_KEY (h, i), make_number (0) , Qnil, | ||
| 1702 | completion_ignore_case ? Qt : Qnil), | ||
| 1703 | Qt)) | ||
| 1704 | { | ||
| 1705 | tem = HASH_KEY (h, i); | ||
| 1706 | break; | ||
| 1707 | } | ||
| 1708 | if (!STRINGP (tem)) | ||
| 1693 | return Qnil; | 1709 | return Qnil; |
| 1694 | } | 1710 | } |
| 1695 | else | 1711 | else |
diff --git a/src/regex.c b/src/regex.c deleted file mode 100644 index a145183510d..00000000000 --- a/src/regex.c +++ /dev/null | |||
| @@ -1,6335 +0,0 @@ | |||
| 1 | /* Extended regular expression matching and search library, version | ||
| 2 | 0.12. (Implements POSIX draft P1003.2/D11.2, except for some of the | ||
| 3 | internationalization features.) | ||
| 4 | |||
| 5 | Copyright (C) 1993,94,95,96,97,98,99,2000 Free Software Foundation, Inc. | ||
| 6 | |||
| 7 | This program is free software; you can redistribute it and/or modify | ||
| 8 | it under the terms of the GNU General Public License as published by | ||
| 9 | the Free Software Foundation; either version 2, or (at your option) | ||
| 10 | any later version. | ||
| 11 | |||
| 12 | This program is distributed in the hope that it will be useful, | ||
| 13 | but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 14 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | ||
| 15 | GNU General Public License for more details. | ||
| 16 | |||
| 17 | You should have received a copy of the GNU General Public License | ||
| 18 | along with this program; if not, write to the Free Software | ||
| 19 | Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, | ||
| 20 | USA. */ | ||
| 21 | |||
| 22 | /* TODO: | ||
| 23 | - structure the opcode space into opcode+flag. | ||
| 24 | - merge with glibc's regex.[ch]. | ||
| 25 | - replace (succeed_n + jump_n + set_number_at) with something that doesn't | ||
| 26 | need to modify the compiled regexp so that re_match can be reentrant. | ||
| 27 | - get rid of on_failure_jump_smart by doing the optimization in re_comp | ||
| 28 | rather than at run-time, so that re_match can be reentrant. | ||
| 29 | */ | ||
| 30 | |||
| 31 | /* AIX requires this to be the first thing in the file. */ | ||
| 32 | #if defined _AIX && !defined REGEX_MALLOC | ||
| 33 | #pragma alloca | ||
| 34 | #endif | ||
| 35 | |||
| 36 | #ifdef HAVE_CONFIG_H | ||
| 37 | # include <config.h> | ||
| 38 | #endif | ||
| 39 | |||
| 40 | #if defined STDC_HEADERS && !defined emacs | ||
| 41 | # include <stddef.h> | ||
| 42 | #else | ||
| 43 | /* We need this for `regex.h', and perhaps for the Emacs include files. */ | ||
| 44 | # include <sys/types.h> | ||
| 45 | #endif | ||
| 46 | |||
| 47 | /* Whether to use ISO C Amendment 1 wide char functions. | ||
| 48 | Those should not be used for Emacs since it uses its own. */ | ||
| 49 | #if defined _LIBC | ||
| 50 | #define WIDE_CHAR_SUPPORT 1 | ||
| 51 | #else | ||
| 52 | #define WIDE_CHAR_SUPPORT \ | ||
| 53 | (HAVE_WCTYPE_H && HAVE_WCHAR_H && HAVE_BTOWC && !emacs) | ||
| 54 | #endif | ||
| 55 | |||
| 56 | /* For platform which support the ISO C amendement 1 functionality we | ||
| 57 | support user defined character classes. */ | ||
| 58 | #if WIDE_CHAR_SUPPORT | ||
| 59 | /* Solaris 2.5 has a bug: <wchar.h> must be included before <wctype.h>. */ | ||
| 60 | # include <wchar.h> | ||
| 61 | # include <wctype.h> | ||
| 62 | #endif | ||
| 63 | |||
| 64 | #ifdef _LIBC | ||
| 65 | /* We have to keep the namespace clean. */ | ||
| 66 | # define regfree(preg) __regfree (preg) | ||
| 67 | # define regexec(pr, st, nm, pm, ef) __regexec (pr, st, nm, pm, ef) | ||
| 68 | # define regcomp(preg, pattern, cflags) __regcomp (preg, pattern, cflags) | ||
| 69 | # define regerror(errcode, preg, errbuf, errbuf_size) \ | ||
| 70 | __regerror(errcode, preg, errbuf, errbuf_size) | ||
| 71 | # define re_set_registers(bu, re, nu, st, en) \ | ||
| 72 | __re_set_registers (bu, re, nu, st, en) | ||
| 73 | # define re_match_2(bufp, string1, size1, string2, size2, pos, regs, stop) \ | ||
| 74 | __re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop) | ||
| 75 | # define re_match(bufp, string, size, pos, regs) \ | ||
| 76 | __re_match (bufp, string, size, pos, regs) | ||
| 77 | # define re_search(bufp, string, size, startpos, range, regs) \ | ||
| 78 | __re_search (bufp, string, size, startpos, range, regs) | ||
| 79 | # define re_compile_pattern(pattern, length, bufp) \ | ||
| 80 | __re_compile_pattern (pattern, length, bufp) | ||
| 81 | # define re_set_syntax(syntax) __re_set_syntax (syntax) | ||
| 82 | # define re_search_2(bufp, st1, s1, st2, s2, startpos, range, regs, stop) \ | ||
| 83 | __re_search_2 (bufp, st1, s1, st2, s2, startpos, range, regs, stop) | ||
| 84 | # define re_compile_fastmap(bufp) __re_compile_fastmap (bufp) | ||
| 85 | |||
| 86 | /* Make sure we call libc's function even if the user overrides them. */ | ||
| 87 | # define btowc __btowc | ||
| 88 | # define iswctype __iswctype | ||
| 89 | # define wctype __wctype | ||
| 90 | |||
| 91 | # define WEAK_ALIAS(a,b) weak_alias (a, b) | ||
| 92 | |||
| 93 | /* We are also using some library internals. */ | ||
| 94 | # include <locale/localeinfo.h> | ||
| 95 | # include <locale/elem-hash.h> | ||
| 96 | # include <langinfo.h> | ||
| 97 | #else | ||
| 98 | # define WEAK_ALIAS(a,b) | ||
| 99 | #endif | ||
| 100 | |||
| 101 | /* This is for other GNU distributions with internationalized messages. */ | ||
| 102 | #if HAVE_LIBINTL_H || defined _LIBC | ||
| 103 | # include <libintl.h> | ||
| 104 | #else | ||
| 105 | # define gettext(msgid) (msgid) | ||
| 106 | #endif | ||
| 107 | |||
| 108 | #ifndef gettext_noop | ||
| 109 | /* This define is so xgettext can find the internationalizable | ||
| 110 | strings. */ | ||
| 111 | # define gettext_noop(String) String | ||
| 112 | #endif | ||
| 113 | |||
| 114 | /* The `emacs' switch turns on certain matching commands | ||
| 115 | that make sense only in Emacs. */ | ||
| 116 | #ifdef emacs | ||
| 117 | |||
| 118 | # include "lisp.h" | ||
| 119 | # include "buffer.h" | ||
| 120 | |||
| 121 | /* Make syntax table lookup grant data in gl_state. */ | ||
| 122 | # define SYNTAX_ENTRY_VIA_PROPERTY | ||
| 123 | |||
| 124 | # include "syntax.h" | ||
| 125 | # include "charset.h" | ||
| 126 | # include "category.h" | ||
| 127 | |||
| 128 | # ifdef malloc | ||
| 129 | # undef malloc | ||
| 130 | # endif | ||
| 131 | # define malloc xmalloc | ||
| 132 | # ifdef realloc | ||
| 133 | # undef realloc | ||
| 134 | # endif | ||
| 135 | # define realloc xrealloc | ||
| 136 | # ifdef free | ||
| 137 | # undef free | ||
| 138 | # endif | ||
| 139 | # define free xfree | ||
| 140 | |||
| 141 | /* Converts the pointer to the char to BEG-based offset from the start. */ | ||
| 142 | # define PTR_TO_OFFSET(d) POS_AS_IN_BUFFER (POINTER_TO_OFFSET (d)) | ||
| 143 | # define POS_AS_IN_BUFFER(p) ((p) + (NILP (re_match_object) || BUFFERP (re_match_object))) | ||
| 144 | |||
| 145 | # define RE_MULTIBYTE_P(bufp) ((bufp)->multibyte) | ||
| 146 | # define RE_STRING_CHAR(p, s) \ | ||
| 147 | (multibyte ? (STRING_CHAR (p, s)) : (*(p))) | ||
| 148 | # define RE_STRING_CHAR_AND_LENGTH(p, s, len) \ | ||
| 149 | (multibyte ? (STRING_CHAR_AND_LENGTH (p, s, len)) : ((len) = 1, *(p))) | ||
| 150 | |||
| 151 | /* Set C a (possibly multibyte) character before P. P points into a | ||
| 152 | string which is the virtual concatenation of STR1 (which ends at | ||
| 153 | END1) or STR2 (which ends at END2). */ | ||
| 154 | # define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \ | ||
| 155 | do { \ | ||
| 156 | if (multibyte) \ | ||
| 157 | { \ | ||
| 158 | re_char *dtemp = (p) == (str2) ? (end1) : (p); \ | ||
| 159 | re_char *dlimit = ((p) > (str2) && (p) <= (end2)) ? (str2) : (str1); \ | ||
| 160 | re_char *d0 = dtemp; \ | ||
| 161 | PREV_CHAR_BOUNDARY (d0, dlimit); \ | ||
| 162 | c = STRING_CHAR (d0, dtemp - d0); \ | ||
| 163 | } \ | ||
| 164 | else \ | ||
| 165 | (c = ((p) == (str2) ? (end1) : (p))[-1]); \ | ||
| 166 | } while (0) | ||
| 167 | |||
| 168 | |||
| 169 | #else /* not emacs */ | ||
| 170 | |||
| 171 | /* If we are not linking with Emacs proper, | ||
| 172 | we can't use the relocating allocator | ||
| 173 | even if config.h says that we can. */ | ||
| 174 | # undef REL_ALLOC | ||
| 175 | |||
| 176 | # if defined STDC_HEADERS || defined _LIBC | ||
| 177 | # include <stdlib.h> | ||
| 178 | # else | ||
| 179 | char *malloc (); | ||
| 180 | char *realloc (); | ||
| 181 | # endif | ||
| 182 | |||
| 183 | /* When used in Emacs's lib-src, we need to get bzero and bcopy somehow. | ||
| 184 | If nothing else has been done, use the method below. */ | ||
| 185 | # ifdef INHIBIT_STRING_HEADER | ||
| 186 | # if !(defined HAVE_BZERO && defined HAVE_BCOPY) | ||
| 187 | # if !defined bzero && !defined bcopy | ||
| 188 | # undef INHIBIT_STRING_HEADER | ||
| 189 | # endif | ||
| 190 | # endif | ||
| 191 | # endif | ||
| 192 | |||
| 193 | /* This is the normal way of making sure we have memcpy, memcmp and bzero. | ||
| 194 | This is used in most programs--a few other programs avoid this | ||
| 195 | by defining INHIBIT_STRING_HEADER. */ | ||
| 196 | # ifndef INHIBIT_STRING_HEADER | ||
| 197 | # if defined HAVE_STRING_H || defined STDC_HEADERS || defined _LIBC | ||
| 198 | # include <string.h> | ||
| 199 | # ifndef bzero | ||
| 200 | # ifndef _LIBC | ||
| 201 | # define bzero(s, n) (memset (s, '\0', n), (s)) | ||
| 202 | # else | ||
| 203 | # define bzero(s, n) __bzero (s, n) | ||
| 204 | # endif | ||
| 205 | # endif | ||
| 206 | # else | ||
| 207 | # include <strings.h> | ||
| 208 | # ifndef memcmp | ||
| 209 | # define memcmp(s1, s2, n) bcmp (s1, s2, n) | ||
| 210 | # endif | ||
| 211 | # ifndef memcpy | ||
| 212 | # define memcpy(d, s, n) (bcopy (s, d, n), (d)) | ||
| 213 | # endif | ||
| 214 | # endif | ||
| 215 | # endif | ||
| 216 | |||
| 217 | /* Define the syntax stuff for \<, \>, etc. */ | ||
| 218 | |||
| 219 | /* Sword must be nonzero for the wordchar pattern commands in re_match_2. */ | ||
| 220 | enum syntaxcode { Swhitespace = 0, Sword = 1 }; | ||
| 221 | |||
| 222 | # ifdef SWITCH_ENUM_BUG | ||
| 223 | # define SWITCH_ENUM_CAST(x) ((int)(x)) | ||
| 224 | # else | ||
| 225 | # define SWITCH_ENUM_CAST(x) (x) | ||
| 226 | # endif | ||
| 227 | |||
| 228 | /* Dummy macros for non-Emacs environments. */ | ||
| 229 | # define BASE_LEADING_CODE_P(c) (0) | ||
| 230 | # define CHAR_CHARSET(c) 0 | ||
| 231 | # define CHARSET_LEADING_CODE_BASE(c) 0 | ||
| 232 | # define MAX_MULTIBYTE_LENGTH 1 | ||
| 233 | # define RE_MULTIBYTE_P(x) 0 | ||
| 234 | # define WORD_BOUNDARY_P(c1, c2) (0) | ||
| 235 | # define CHAR_HEAD_P(p) (1) | ||
| 236 | # define SINGLE_BYTE_CHAR_P(c) (1) | ||
| 237 | # define SAME_CHARSET_P(c1, c2) (1) | ||
| 238 | # define MULTIBYTE_FORM_LENGTH(p, s) (1) | ||
| 239 | # define PREV_CHAR_BOUNDARY(p, limit) ((p)--) | ||
| 240 | # define STRING_CHAR(p, s) (*(p)) | ||
| 241 | # define RE_STRING_CHAR STRING_CHAR | ||
| 242 | # define CHAR_STRING(c, s) (*(s) = (c), 1) | ||
| 243 | # define STRING_CHAR_AND_LENGTH(p, s, actual_len) ((actual_len) = 1, *(p)) | ||
| 244 | # define RE_STRING_CHAR_AND_LENGTH STRING_CHAR_AND_LENGTH | ||
| 245 | # define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \ | ||
| 246 | (c = ((p) == (str2) ? *((end1) - 1) : *((p) - 1))) | ||
| 247 | # define MAKE_CHAR(charset, c1, c2) (c1) | ||
| 248 | #endif /* not emacs */ | ||
| 249 | |||
| 250 | #ifndef RE_TRANSLATE | ||
| 251 | # define RE_TRANSLATE(TBL, C) ((unsigned char)(TBL)[C]) | ||
| 252 | # define RE_TRANSLATE_P(TBL) (TBL) | ||
| 253 | #endif | ||
| 254 | |||
| 255 | /* Get the interface, including the syntax bits. */ | ||
| 256 | #include "regex.h" | ||
| 257 | |||
| 258 | /* isalpha etc. are used for the character classes. */ | ||
| 259 | #include <ctype.h> | ||
| 260 | |||
| 261 | #ifdef emacs | ||
| 262 | |||
| 263 | /* 1 if C is an ASCII character. */ | ||
| 264 | # define IS_REAL_ASCII(c) ((c) < 0200) | ||
| 265 | |||
| 266 | /* 1 if C is a unibyte character. */ | ||
| 267 | # define ISUNIBYTE(c) (SINGLE_BYTE_CHAR_P ((c))) | ||
| 268 | |||
| 269 | /* The Emacs definitions should not be directly affected by locales. */ | ||
| 270 | |||
| 271 | /* In Emacs, these are only used for single-byte characters. */ | ||
| 272 | # define ISDIGIT(c) ((c) >= '0' && (c) <= '9') | ||
| 273 | # define ISCNTRL(c) ((c) < ' ') | ||
| 274 | # define ISXDIGIT(c) (((c) >= '0' && (c) <= '9') \ | ||
| 275 | || ((c) >= 'a' && (c) <= 'f') \ | ||
| 276 | || ((c) >= 'A' && (c) <= 'F')) | ||
| 277 | |||
| 278 | /* This is only used for single-byte characters. */ | ||
| 279 | # define ISBLANK(c) ((c) == ' ' || (c) == '\t') | ||
| 280 | |||
| 281 | /* The rest must handle multibyte characters. */ | ||
| 282 | |||
| 283 | # define ISGRAPH(c) (SINGLE_BYTE_CHAR_P (c) \ | ||
| 284 | ? (c) > ' ' && !((c) >= 0177 && (c) <= 0237) \ | ||
| 285 | : 1) | ||
| 286 | |||
| 287 | # define ISPRINT(c) (SINGLE_BYTE_CHAR_P (c) \ | ||
| 288 | ? (c) >= ' ' && !((c) >= 0177 && (c) <= 0237) \ | ||
| 289 | : 1) | ||
| 290 | |||
| 291 | # define ISALNUM(c) (IS_REAL_ASCII (c) \ | ||
| 292 | ? (((c) >= 'a' && (c) <= 'z') \ | ||
| 293 | || ((c) >= 'A' && (c) <= 'Z') \ | ||
| 294 | || ((c) >= '0' && (c) <= '9')) \ | ||
| 295 | : SYNTAX (c) == Sword) | ||
| 296 | |||
| 297 | # define ISALPHA(c) (IS_REAL_ASCII (c) \ | ||
| 298 | ? (((c) >= 'a' && (c) <= 'z') \ | ||
| 299 | || ((c) >= 'A' && (c) <= 'Z')) \ | ||
| 300 | : SYNTAX (c) == Sword) | ||
| 301 | |||
| 302 | # define ISLOWER(c) (LOWERCASEP (c)) | ||
| 303 | |||
| 304 | # define ISPUNCT(c) (IS_REAL_ASCII (c) \ | ||
| 305 | ? ((c) > ' ' && (c) < 0177 \ | ||
| 306 | && !(((c) >= 'a' && (c) <= 'z') \ | ||
| 307 | || ((c) >= 'A' && (c) <= 'Z') \ | ||
| 308 | || ((c) >= '0' && (c) <= '9'))) \ | ||
| 309 | : SYNTAX (c) != Sword) | ||
| 310 | |||
| 311 | # define ISSPACE(c) (SYNTAX (c) == Swhitespace) | ||
| 312 | |||
| 313 | # define ISUPPER(c) (UPPERCASEP (c)) | ||
| 314 | |||
| 315 | # define ISWORD(c) (SYNTAX (c) == Sword) | ||
| 316 | |||
| 317 | #else /* not emacs */ | ||
| 318 | |||
| 319 | /* Jim Meyering writes: | ||
| 320 | |||
| 321 | "... Some ctype macros are valid only for character codes that | ||
| 322 | isascii says are ASCII (SGI's IRIX-4.0.5 is one such system --when | ||
| 323 | using /bin/cc or gcc but without giving an ansi option). So, all | ||
| 324 | ctype uses should be through macros like ISPRINT... If | ||
| 325 | STDC_HEADERS is defined, then autoconf has verified that the ctype | ||
| 326 | macros don't need to be guarded with references to isascii. ... | ||
| 327 | Defining isascii to 1 should let any compiler worth its salt | ||
| 328 | eliminate the && through constant folding." | ||
| 329 | Solaris defines some of these symbols so we must undefine them first. */ | ||
| 330 | |||
| 331 | # undef ISASCII | ||
| 332 | # if defined STDC_HEADERS || (!defined isascii && !defined HAVE_ISASCII) | ||
| 333 | # define ISASCII(c) 1 | ||
| 334 | # else | ||
| 335 | # define ISASCII(c) isascii(c) | ||
| 336 | # endif | ||
| 337 | |||
| 338 | /* 1 if C is an ASCII character. */ | ||
| 339 | # define IS_REAL_ASCII(c) ((c) < 0200) | ||
| 340 | |||
| 341 | /* This distinction is not meaningful, except in Emacs. */ | ||
| 342 | # define ISUNIBYTE(c) 1 | ||
| 343 | |||
| 344 | # ifdef isblank | ||
| 345 | # define ISBLANK(c) (ISASCII (c) && isblank (c)) | ||
| 346 | # else | ||
| 347 | # define ISBLANK(c) ((c) == ' ' || (c) == '\t') | ||
| 348 | # endif | ||
| 349 | # ifdef isgraph | ||
| 350 | # define ISGRAPH(c) (ISASCII (c) && isgraph (c)) | ||
| 351 | # else | ||
| 352 | # define ISGRAPH(c) (ISASCII (c) && isprint (c) && !isspace (c)) | ||
| 353 | # endif | ||
| 354 | |||
| 355 | # undef ISPRINT | ||
| 356 | # define ISPRINT(c) (ISASCII (c) && isprint (c)) | ||
| 357 | # define ISDIGIT(c) (ISASCII (c) && isdigit (c)) | ||
| 358 | # define ISALNUM(c) (ISASCII (c) && isalnum (c)) | ||
| 359 | # define ISALPHA(c) (ISASCII (c) && isalpha (c)) | ||
| 360 | # define ISCNTRL(c) (ISASCII (c) && iscntrl (c)) | ||
| 361 | # define ISLOWER(c) (ISASCII (c) && islower (c)) | ||
| 362 | # define ISPUNCT(c) (ISASCII (c) && ispunct (c)) | ||
| 363 | # define ISSPACE(c) (ISASCII (c) && isspace (c)) | ||
| 364 | # define ISUPPER(c) (ISASCII (c) && isupper (c)) | ||
| 365 | # define ISXDIGIT(c) (ISASCII (c) && isxdigit (c)) | ||
| 366 | |||
| 367 | # define ISWORD(c) ISALPHA(c) | ||
| 368 | |||
| 369 | # ifdef _tolower | ||
| 370 | # define TOLOWER(c) _tolower(c) | ||
| 371 | # else | ||
| 372 | # define TOLOWER(c) tolower(c) | ||
| 373 | # endif | ||
| 374 | |||
| 375 | /* How many characters in the character set. */ | ||
| 376 | # define CHAR_SET_SIZE 256 | ||
| 377 | |||
| 378 | # ifdef SYNTAX_TABLE | ||
| 379 | |||
| 380 | extern char *re_syntax_table; | ||
| 381 | |||
| 382 | # else /* not SYNTAX_TABLE */ | ||
| 383 | |||
| 384 | static char re_syntax_table[CHAR_SET_SIZE]; | ||
| 385 | |||
| 386 | static void | ||
| 387 | init_syntax_once () | ||
| 388 | { | ||
| 389 | register int c; | ||
| 390 | static int done = 0; | ||
| 391 | |||
| 392 | if (done) | ||
| 393 | return; | ||
| 394 | |||
| 395 | bzero (re_syntax_table, sizeof re_syntax_table); | ||
| 396 | |||
| 397 | for (c = 0; c < CHAR_SET_SIZE; ++c) | ||
| 398 | if (ISALNUM (c)) | ||
| 399 | re_syntax_table[c] = Sword; | ||
| 400 | |||
| 401 | re_syntax_table['_'] = Sword; | ||
| 402 | |||
| 403 | done = 1; | ||
| 404 | } | ||
| 405 | |||
| 406 | # endif /* not SYNTAX_TABLE */ | ||
| 407 | |||
| 408 | # define SYNTAX(c) re_syntax_table[(c)] | ||
| 409 | |||
| 410 | #endif /* not emacs */ | ||
| 411 | |||
| 412 | #ifndef NULL | ||
| 413 | # define NULL (void *)0 | ||
| 414 | #endif | ||
| 415 | |||
| 416 | /* We remove any previous definition of `SIGN_EXTEND_CHAR', | ||
| 417 | since ours (we hope) works properly with all combinations of | ||
| 418 | machines, compilers, `char' and `unsigned char' argument types. | ||
| 419 | (Per Bothner suggested the basic approach.) */ | ||
| 420 | #undef SIGN_EXTEND_CHAR | ||
| 421 | #if __STDC__ | ||
| 422 | # define SIGN_EXTEND_CHAR(c) ((signed char) (c)) | ||
| 423 | #else /* not __STDC__ */ | ||
| 424 | /* As in Harbison and Steele. */ | ||
| 425 | # define SIGN_EXTEND_CHAR(c) ((((unsigned char) (c)) ^ 128) - 128) | ||
| 426 | #endif | ||
| 427 | |||
| 428 | /* Should we use malloc or alloca? If REGEX_MALLOC is not defined, we | ||
| 429 | use `alloca' instead of `malloc'. This is because using malloc in | ||
| 430 | re_search* or re_match* could cause memory leaks when C-g is used in | ||
| 431 | Emacs; also, malloc is slower and causes storage fragmentation. On | ||
| 432 | the other hand, malloc is more portable, and easier to debug. | ||
| 433 | |||
| 434 | Because we sometimes use alloca, some routines have to be macros, | ||
| 435 | not functions -- `alloca'-allocated space disappears at the end of the | ||
| 436 | function it is called in. */ | ||
| 437 | |||
| 438 | #ifdef REGEX_MALLOC | ||
| 439 | |||
| 440 | # define REGEX_ALLOCATE malloc | ||
| 441 | # define REGEX_REALLOCATE(source, osize, nsize) realloc (source, nsize) | ||
| 442 | # define REGEX_FREE free | ||
| 443 | |||
| 444 | #else /* not REGEX_MALLOC */ | ||
| 445 | |||
| 446 | /* Emacs already defines alloca, sometimes. */ | ||
| 447 | # ifndef alloca | ||
| 448 | |||
| 449 | /* Make alloca work the best possible way. */ | ||
| 450 | # ifdef __GNUC__ | ||
| 451 | # define alloca __builtin_alloca | ||
| 452 | # else /* not __GNUC__ */ | ||
| 453 | # if HAVE_ALLOCA_H | ||
| 454 | # include <alloca.h> | ||
| 455 | # endif /* HAVE_ALLOCA_H */ | ||
| 456 | # endif /* not __GNUC__ */ | ||
| 457 | |||
| 458 | # endif /* not alloca */ | ||
| 459 | |||
| 460 | # define REGEX_ALLOCATE alloca | ||
| 461 | |||
| 462 | /* Assumes a `char *destination' variable. */ | ||
| 463 | # define REGEX_REALLOCATE(source, osize, nsize) \ | ||
| 464 | (destination = (char *) alloca (nsize), \ | ||
| 465 | memcpy (destination, source, osize)) | ||
| 466 | |||
| 467 | /* No need to do anything to free, after alloca. */ | ||
| 468 | # define REGEX_FREE(arg) ((void)0) /* Do nothing! But inhibit gcc warning. */ | ||
| 469 | |||
| 470 | #endif /* not REGEX_MALLOC */ | ||
| 471 | |||
| 472 | /* Define how to allocate the failure stack. */ | ||
| 473 | |||
| 474 | #if defined REL_ALLOC && defined REGEX_MALLOC | ||
| 475 | |||
| 476 | # define REGEX_ALLOCATE_STACK(size) \ | ||
| 477 | r_alloc (&failure_stack_ptr, (size)) | ||
| 478 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) \ | ||
| 479 | r_re_alloc (&failure_stack_ptr, (nsize)) | ||
| 480 | # define REGEX_FREE_STACK(ptr) \ | ||
| 481 | r_alloc_free (&failure_stack_ptr) | ||
| 482 | |||
| 483 | #else /* not using relocating allocator */ | ||
| 484 | |||
| 485 | # ifdef REGEX_MALLOC | ||
| 486 | |||
| 487 | # define REGEX_ALLOCATE_STACK malloc | ||
| 488 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) realloc (source, nsize) | ||
| 489 | # define REGEX_FREE_STACK free | ||
| 490 | |||
| 491 | # else /* not REGEX_MALLOC */ | ||
| 492 | |||
| 493 | # define REGEX_ALLOCATE_STACK alloca | ||
| 494 | |||
| 495 | # define REGEX_REALLOCATE_STACK(source, osize, nsize) \ | ||
| 496 | REGEX_REALLOCATE (source, osize, nsize) | ||
| 497 | /* No need to explicitly free anything. */ | ||
| 498 | # define REGEX_FREE_STACK(arg) ((void)0) | ||
| 499 | |||
| 500 | # endif /* not REGEX_MALLOC */ | ||
| 501 | #endif /* not using relocating allocator */ | ||
| 502 | |||
| 503 | |||
| 504 | /* True if `size1' is non-NULL and PTR is pointing anywhere inside | ||
| 505 | `string1' or just past its end. This works if PTR is NULL, which is | ||
| 506 | a good thing. */ | ||
| 507 | #define FIRST_STRING_P(ptr) \ | ||
| 508 | (size1 && string1 <= (ptr) && (ptr) <= string1 + size1) | ||
| 509 | |||
| 510 | /* (Re)Allocate N items of type T using malloc, or fail. */ | ||
| 511 | #define TALLOC(n, t) ((t *) malloc ((n) * sizeof (t))) | ||
| 512 | #define RETALLOC(addr, n, t) ((addr) = (t *) realloc (addr, (n) * sizeof (t))) | ||
| 513 | #define RETALLOC_IF(addr, n, t) \ | ||
| 514 | if (addr) RETALLOC((addr), (n), t); else (addr) = TALLOC ((n), t) | ||
| 515 | #define REGEX_TALLOC(n, t) ((t *) REGEX_ALLOCATE ((n) * sizeof (t))) | ||
| 516 | |||
| 517 | #define BYTEWIDTH 8 /* In bits. */ | ||
| 518 | |||
| 519 | #define STREQ(s1, s2) ((strcmp (s1, s2) == 0)) | ||
| 520 | |||
| 521 | #undef MAX | ||
| 522 | #undef MIN | ||
| 523 | #define MAX(a, b) ((a) > (b) ? (a) : (b)) | ||
| 524 | #define MIN(a, b) ((a) < (b) ? (a) : (b)) | ||
| 525 | |||
| 526 | /* Type of source-pattern and string chars. */ | ||
| 527 | typedef const unsigned char re_char; | ||
| 528 | |||
| 529 | typedef char boolean; | ||
| 530 | #define false 0 | ||
| 531 | #define true 1 | ||
| 532 | |||
| 533 | static int re_match_2_internal _RE_ARGS ((struct re_pattern_buffer *bufp, | ||
| 534 | re_char *string1, int size1, | ||
| 535 | re_char *string2, int size2, | ||
| 536 | int pos, | ||
| 537 | struct re_registers *regs, | ||
| 538 | int stop)); | ||
| 539 | |||
| 540 | /* These are the command codes that appear in compiled regular | ||
| 541 | expressions. Some opcodes are followed by argument bytes. A | ||
| 542 | command code can specify any interpretation whatsoever for its | ||
| 543 | arguments. Zero bytes may appear in the compiled regular expression. */ | ||
| 544 | |||
| 545 | typedef enum | ||
| 546 | { | ||
| 547 | no_op = 0, | ||
| 548 | |||
| 549 | /* Succeed right away--no more backtracking. */ | ||
| 550 | succeed, | ||
| 551 | |||
| 552 | /* Followed by one byte giving n, then by n literal bytes. */ | ||
| 553 | exactn, | ||
| 554 | |||
| 555 | /* Matches any (more or less) character. */ | ||
| 556 | anychar, | ||
| 557 | |||
| 558 | /* Matches any one char belonging to specified set. First | ||
| 559 | following byte is number of bitmap bytes. Then come bytes | ||
| 560 | for a bitmap saying which chars are in. Bits in each byte | ||
| 561 | are ordered low-bit-first. A character is in the set if its | ||
| 562 | bit is 1. A character too large to have a bit in the map is | ||
| 563 | automatically not in the set. | ||
| 564 | |||
| 565 | If the length byte has the 0x80 bit set, then that stuff | ||
| 566 | is followed by a range table: | ||
| 567 | 2 bytes of flags for character sets (low 8 bits, high 8 bits) | ||
| 568 | See RANGE_TABLE_WORK_BITS below. | ||
| 569 | 2 bytes, the number of pairs that follow (upto 32767) | ||
| 570 | pairs, each 2 multibyte characters, | ||
| 571 | each multibyte character represented as 3 bytes. */ | ||
| 572 | charset, | ||
| 573 | |||
| 574 | /* Same parameters as charset, but match any character that is | ||
| 575 | not one of those specified. */ | ||
| 576 | charset_not, | ||
| 577 | |||
| 578 | /* Start remembering the text that is matched, for storing in a | ||
| 579 | register. Followed by one byte with the register number, in | ||
| 580 | the range 0 to one less than the pattern buffer's re_nsub | ||
| 581 | field. */ | ||
| 582 | start_memory, | ||
| 583 | |||
| 584 | /* Stop remembering the text that is matched and store it in a | ||
| 585 | memory register. Followed by one byte with the register | ||
| 586 | number, in the range 0 to one less than `re_nsub' in the | ||
| 587 | pattern buffer. */ | ||
| 588 | stop_memory, | ||
| 589 | |||
| 590 | /* Match a duplicate of something remembered. Followed by one | ||
| 591 | byte containing the register number. */ | ||
| 592 | duplicate, | ||
| 593 | |||
| 594 | /* Fail unless at beginning of line. */ | ||
| 595 | begline, | ||
| 596 | |||
| 597 | /* Fail unless at end of line. */ | ||
| 598 | endline, | ||
| 599 | |||
| 600 | /* Succeeds if at beginning of buffer (if emacs) or at beginning | ||
| 601 | of string to be matched (if not). */ | ||
| 602 | begbuf, | ||
| 603 | |||
| 604 | /* Analogously, for end of buffer/string. */ | ||
| 605 | endbuf, | ||
| 606 | |||
| 607 | /* Followed by two byte relative address to which to jump. */ | ||
| 608 | jump, | ||
| 609 | |||
| 610 | /* Followed by two-byte relative address of place to resume at | ||
| 611 | in case of failure. */ | ||
| 612 | on_failure_jump, | ||
| 613 | |||
| 614 | /* Like on_failure_jump, but pushes a placeholder instead of the | ||
| 615 | current string position when executed. */ | ||
| 616 | on_failure_keep_string_jump, | ||
| 617 | |||
| 618 | /* Just like `on_failure_jump', except that it checks that we | ||
| 619 | don't get stuck in an infinite loop (matching an empty string | ||
| 620 | indefinitely). */ | ||
| 621 | on_failure_jump_loop, | ||
| 622 | |||
| 623 | /* Just like `on_failure_jump_loop', except that it checks for | ||
| 624 | a different kind of loop (the kind that shows up with non-greedy | ||
| 625 | operators). This operation has to be immediately preceded | ||
| 626 | by a `no_op'. */ | ||
| 627 | on_failure_jump_nastyloop, | ||
| 628 | |||
| 629 | /* A smart `on_failure_jump' used for greedy * and + operators. | ||
| 630 | It analyses the loop before which it is put and if the | ||
| 631 | loop does not require backtracking, it changes itself to | ||
| 632 | `on_failure_keep_string_jump' and short-circuits the loop, | ||
| 633 | else it just defaults to changing itself into `on_failure_jump'. | ||
| 634 | It assumes that it is pointing to just past a `jump'. */ | ||
| 635 | on_failure_jump_smart, | ||
| 636 | |||
| 637 | /* Followed by two-byte relative address and two-byte number n. | ||
| 638 | After matching N times, jump to the address upon failure. | ||
| 639 | Does not work if N starts at 0: use on_failure_jump_loop | ||
| 640 | instead. */ | ||
| 641 | succeed_n, | ||
| 642 | |||
| 643 | /* Followed by two-byte relative address, and two-byte number n. | ||
| 644 | Jump to the address N times, then fail. */ | ||
| 645 | jump_n, | ||
| 646 | |||
| 647 | /* Set the following two-byte relative address to the | ||
| 648 | subsequent two-byte number. The address *includes* the two | ||
| 649 | bytes of number. */ | ||
| 650 | set_number_at, | ||
| 651 | |||
| 652 | wordbeg, /* Succeeds if at word beginning. */ | ||
| 653 | wordend, /* Succeeds if at word end. */ | ||
| 654 | |||
| 655 | wordbound, /* Succeeds if at a word boundary. */ | ||
| 656 | notwordbound, /* Succeeds if not at a word boundary. */ | ||
| 657 | |||
| 658 | /* Matches any character whose syntax is specified. Followed by | ||
| 659 | a byte which contains a syntax code, e.g., Sword. */ | ||
| 660 | syntaxspec, | ||
| 661 | |||
| 662 | /* Matches any character whose syntax is not that specified. */ | ||
| 663 | notsyntaxspec | ||
| 664 | |||
| 665 | #ifdef emacs | ||
| 666 | ,before_dot, /* Succeeds if before point. */ | ||
| 667 | at_dot, /* Succeeds if at point. */ | ||
| 668 | after_dot, /* Succeeds if after point. */ | ||
| 669 | |||
| 670 | /* Matches any character whose category-set contains the specified | ||
| 671 | category. The operator is followed by a byte which contains a | ||
| 672 | category code (mnemonic ASCII character). */ | ||
| 673 | categoryspec, | ||
| 674 | |||
| 675 | /* Matches any character whose category-set does not contain the | ||
| 676 | specified category. The operator is followed by a byte which | ||
| 677 | contains the category code (mnemonic ASCII character). */ | ||
| 678 | notcategoryspec | ||
| 679 | #endif /* emacs */ | ||
| 680 | } re_opcode_t; | ||
| 681 | |||
| 682 | /* Common operations on the compiled pattern. */ | ||
| 683 | |||
| 684 | /* Store NUMBER in two contiguous bytes starting at DESTINATION. */ | ||
| 685 | |||
| 686 | #define STORE_NUMBER(destination, number) \ | ||
| 687 | do { \ | ||
| 688 | (destination)[0] = (number) & 0377; \ | ||
| 689 | (destination)[1] = (number) >> 8; \ | ||
| 690 | } while (0) | ||
| 691 | |||
| 692 | /* Same as STORE_NUMBER, except increment DESTINATION to | ||
| 693 | the byte after where the number is stored. Therefore, DESTINATION | ||
| 694 | must be an lvalue. */ | ||
| 695 | |||
| 696 | #define STORE_NUMBER_AND_INCR(destination, number) \ | ||
| 697 | do { \ | ||
| 698 | STORE_NUMBER (destination, number); \ | ||
| 699 | (destination) += 2; \ | ||
| 700 | } while (0) | ||
| 701 | |||
| 702 | /* Put into DESTINATION a number stored in two contiguous bytes starting | ||
| 703 | at SOURCE. */ | ||
| 704 | |||
| 705 | #define EXTRACT_NUMBER(destination, source) \ | ||
| 706 | do { \ | ||
| 707 | (destination) = *(source) & 0377; \ | ||
| 708 | (destination) += SIGN_EXTEND_CHAR (*((source) + 1)) << 8; \ | ||
| 709 | } while (0) | ||
| 710 | |||
| 711 | #ifdef DEBUG | ||
| 712 | static void extract_number _RE_ARGS ((int *dest, re_char *source)); | ||
| 713 | static void | ||
| 714 | extract_number (dest, source) | ||
| 715 | int *dest; | ||
| 716 | re_char *source; | ||
| 717 | { | ||
| 718 | int temp = SIGN_EXTEND_CHAR (*(source + 1)); | ||
| 719 | *dest = *source & 0377; | ||
| 720 | *dest += temp << 8; | ||
| 721 | } | ||
| 722 | |||
| 723 | # ifndef EXTRACT_MACROS /* To debug the macros. */ | ||
| 724 | # undef EXTRACT_NUMBER | ||
| 725 | # define EXTRACT_NUMBER(dest, src) extract_number (&dest, src) | ||
| 726 | # endif /* not EXTRACT_MACROS */ | ||
| 727 | |||
| 728 | #endif /* DEBUG */ | ||
| 729 | |||
| 730 | /* Same as EXTRACT_NUMBER, except increment SOURCE to after the number. | ||
| 731 | SOURCE must be an lvalue. */ | ||
| 732 | |||
| 733 | #define EXTRACT_NUMBER_AND_INCR(destination, source) \ | ||
| 734 | do { \ | ||
| 735 | EXTRACT_NUMBER (destination, source); \ | ||
| 736 | (source) += 2; \ | ||
| 737 | } while (0) | ||
| 738 | |||
| 739 | #ifdef DEBUG | ||
| 740 | static void extract_number_and_incr _RE_ARGS ((int *destination, | ||
| 741 | re_char **source)); | ||
| 742 | static void | ||
| 743 | extract_number_and_incr (destination, source) | ||
| 744 | int *destination; | ||
| 745 | re_char **source; | ||
| 746 | { | ||
| 747 | extract_number (destination, *source); | ||
| 748 | *source += 2; | ||
| 749 | } | ||
| 750 | |||
| 751 | # ifndef EXTRACT_MACROS | ||
| 752 | # undef EXTRACT_NUMBER_AND_INCR | ||
| 753 | # define EXTRACT_NUMBER_AND_INCR(dest, src) \ | ||
| 754 | extract_number_and_incr (&dest, &src) | ||
| 755 | # endif /* not EXTRACT_MACROS */ | ||
| 756 | |||
| 757 | #endif /* DEBUG */ | ||
| 758 | |||
| 759 | /* Store a multibyte character in three contiguous bytes starting | ||
| 760 | DESTINATION, and increment DESTINATION to the byte after where the | ||
| 761 | character is stored. Therefore, DESTINATION must be an lvalue. */ | ||
| 762 | |||
| 763 | #define STORE_CHARACTER_AND_INCR(destination, character) \ | ||
| 764 | do { \ | ||
| 765 | (destination)[0] = (character) & 0377; \ | ||
| 766 | (destination)[1] = ((character) >> 8) & 0377; \ | ||
| 767 | (destination)[2] = (character) >> 16; \ | ||
| 768 | (destination) += 3; \ | ||
| 769 | } while (0) | ||
| 770 | |||
| 771 | /* Put into DESTINATION a character stored in three contiguous bytes | ||
| 772 | starting at SOURCE. */ | ||
| 773 | |||
| 774 | #define EXTRACT_CHARACTER(destination, source) \ | ||
| 775 | do { \ | ||
| 776 | (destination) = ((source)[0] \ | ||
| 777 | | ((source)[1] << 8) \ | ||
| 778 | | ((source)[2] << 16)); \ | ||
| 779 | } while (0) | ||
| 780 | |||
| 781 | |||
| 782 | /* Macros for charset. */ | ||
| 783 | |||
| 784 | /* Size of bitmap of charset P in bytes. P is a start of charset, | ||
| 785 | i.e. *P is (re_opcode_t) charset or (re_opcode_t) charset_not. */ | ||
| 786 | #define CHARSET_BITMAP_SIZE(p) ((p)[1] & 0x7F) | ||
| 787 | |||
| 788 | /* Nonzero if charset P has range table. */ | ||
| 789 | #define CHARSET_RANGE_TABLE_EXISTS_P(p) ((p)[1] & 0x80) | ||
| 790 | |||
| 791 | /* Return the address of range table of charset P. But not the start | ||
| 792 | of table itself, but the before where the number of ranges is | ||
| 793 | stored. `2 +' means to skip re_opcode_t and size of bitmap, | ||
| 794 | and the 2 bytes of flags at the start of the range table. */ | ||
| 795 | #define CHARSET_RANGE_TABLE(p) (&(p)[4 + CHARSET_BITMAP_SIZE (p)]) | ||
| 796 | |||
| 797 | /* Extract the bit flags that start a range table. */ | ||
| 798 | #define CHARSET_RANGE_TABLE_BITS(p) \ | ||
| 799 | ((p)[2 + CHARSET_BITMAP_SIZE (p)] \ | ||
| 800 | + (p)[3 + CHARSET_BITMAP_SIZE (p)] * 0x100) | ||
| 801 | |||
| 802 | /* Test if C is listed in the bitmap of charset P. */ | ||
| 803 | #define CHARSET_LOOKUP_BITMAP(p, c) \ | ||
| 804 | ((c) < CHARSET_BITMAP_SIZE (p) * BYTEWIDTH \ | ||
| 805 | && (p)[2 + (c) / BYTEWIDTH] & (1 << ((c) % BYTEWIDTH))) | ||
| 806 | |||
| 807 | /* Return the address of end of RANGE_TABLE. COUNT is number of | ||
| 808 | ranges (which is a pair of (start, end)) in the RANGE_TABLE. `* 2' | ||
| 809 | is start of range and end of range. `* 3' is size of each start | ||
| 810 | and end. */ | ||
| 811 | #define CHARSET_RANGE_TABLE_END(range_table, count) \ | ||
| 812 | ((range_table) + (count) * 2 * 3) | ||
| 813 | |||
| 814 | /* Test if C is in RANGE_TABLE. A flag NOT is negated if C is in. | ||
| 815 | COUNT is number of ranges in RANGE_TABLE. */ | ||
| 816 | #define CHARSET_LOOKUP_RANGE_TABLE_RAW(not, c, range_table, count) \ | ||
| 817 | do \ | ||
| 818 | { \ | ||
| 819 | re_wchar_t range_start, range_end; \ | ||
| 820 | re_char *p; \ | ||
| 821 | re_char *range_table_end \ | ||
| 822 | = CHARSET_RANGE_TABLE_END ((range_table), (count)); \ | ||
| 823 | \ | ||
| 824 | for (p = (range_table); p < range_table_end; p += 2 * 3) \ | ||
| 825 | { \ | ||
| 826 | EXTRACT_CHARACTER (range_start, p); \ | ||
| 827 | EXTRACT_CHARACTER (range_end, p + 3); \ | ||
| 828 | \ | ||
| 829 | if (range_start <= (c) && (c) <= range_end) \ | ||
| 830 | { \ | ||
| 831 | (not) = !(not); \ | ||
| 832 | break; \ | ||
| 833 | } \ | ||
| 834 | } \ | ||
| 835 | } \ | ||
| 836 | while (0) | ||
| 837 | |||
| 838 | /* Test if C is in range table of CHARSET. The flag NOT is negated if | ||
| 839 | C is listed in it. */ | ||
| 840 | #define CHARSET_LOOKUP_RANGE_TABLE(not, c, charset) \ | ||
| 841 | do \ | ||
| 842 | { \ | ||
| 843 | /* Number of ranges in range table. */ \ | ||
| 844 | int count; \ | ||
| 845 | re_char *range_table = CHARSET_RANGE_TABLE (charset); \ | ||
| 846 | \ | ||
| 847 | EXTRACT_NUMBER_AND_INCR (count, range_table); \ | ||
| 848 | CHARSET_LOOKUP_RANGE_TABLE_RAW ((not), (c), range_table, count); \ | ||
| 849 | } \ | ||
| 850 | while (0) | ||
| 851 | |||
| 852 | /* If DEBUG is defined, Regex prints many voluminous messages about what | ||
| 853 | it is doing (if the variable `debug' is nonzero). If linked with the | ||
| 854 | main program in `iregex.c', you can enter patterns and strings | ||
| 855 | interactively. And if linked with the main program in `main.c' and | ||
| 856 | the other test files, you can run the already-written tests. */ | ||
| 857 | |||
| 858 | #ifdef DEBUG | ||
| 859 | |||
| 860 | /* We use standard I/O for debugging. */ | ||
| 861 | # include <stdio.h> | ||
| 862 | |||
| 863 | /* It is useful to test things that ``must'' be true when debugging. */ | ||
| 864 | # include <assert.h> | ||
| 865 | |||
| 866 | static int debug = -100000; | ||
| 867 | |||
| 868 | # define DEBUG_STATEMENT(e) e | ||
| 869 | # define DEBUG_PRINT1(x) if (debug > 0) printf (x) | ||
| 870 | # define DEBUG_PRINT2(x1, x2) if (debug > 0) printf (x1, x2) | ||
| 871 | # define DEBUG_PRINT3(x1, x2, x3) if (debug > 0) printf (x1, x2, x3) | ||
| 872 | # define DEBUG_PRINT4(x1, x2, x3, x4) if (debug > 0) printf (x1, x2, x3, x4) | ||
| 873 | # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \ | ||
| 874 | if (debug > 0) print_partial_compiled_pattern (s, e) | ||
| 875 | # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \ | ||
| 876 | if (debug > 0) print_double_string (w, s1, sz1, s2, sz2) | ||
| 877 | |||
| 878 | |||
| 879 | /* Print the fastmap in human-readable form. */ | ||
| 880 | |||
| 881 | void | ||
| 882 | print_fastmap (fastmap) | ||
| 883 | char *fastmap; | ||
| 884 | { | ||
| 885 | unsigned was_a_range = 0; | ||
| 886 | unsigned i = 0; | ||
| 887 | |||
| 888 | while (i < (1 << BYTEWIDTH)) | ||
| 889 | { | ||
| 890 | if (fastmap[i++]) | ||
| 891 | { | ||
| 892 | was_a_range = 0; | ||
| 893 | putchar (i - 1); | ||
| 894 | while (i < (1 << BYTEWIDTH) && fastmap[i]) | ||
| 895 | { | ||
| 896 | was_a_range = 1; | ||
| 897 | i++; | ||
| 898 | } | ||
| 899 | if (was_a_range) | ||
| 900 | { | ||
| 901 | printf ("-"); | ||
| 902 | putchar (i - 1); | ||
| 903 | } | ||
| 904 | } | ||
| 905 | } | ||
| 906 | putchar ('\n'); | ||
| 907 | } | ||
| 908 | |||
| 909 | |||
| 910 | /* Print a compiled pattern string in human-readable form, starting at | ||
| 911 | the START pointer into it and ending just before the pointer END. */ | ||
| 912 | |||
| 913 | void | ||
| 914 | print_partial_compiled_pattern (start, end) | ||
| 915 | re_char *start; | ||
| 916 | re_char *end; | ||
| 917 | { | ||
| 918 | int mcnt, mcnt2; | ||
| 919 | re_char *p = start; | ||
| 920 | re_char *pend = end; | ||
| 921 | |||
| 922 | if (start == NULL) | ||
| 923 | { | ||
| 924 | fprintf (stderr, "(null)\n"); | ||
| 925 | return; | ||
| 926 | } | ||
| 927 | |||
| 928 | /* Loop over pattern commands. */ | ||
| 929 | while (p < pend) | ||
| 930 | { | ||
| 931 | fprintf (stderr, "%d:\t", p - start); | ||
| 932 | |||
| 933 | switch ((re_opcode_t) *p++) | ||
| 934 | { | ||
| 935 | case no_op: | ||
| 936 | fprintf (stderr, "/no_op"); | ||
| 937 | break; | ||
| 938 | |||
| 939 | case succeed: | ||
| 940 | fprintf (stderr, "/succeed"); | ||
| 941 | break; | ||
| 942 | |||
| 943 | case exactn: | ||
| 944 | mcnt = *p++; | ||
| 945 | fprintf (stderr, "/exactn/%d", mcnt); | ||
| 946 | do | ||
| 947 | { | ||
| 948 | fprintf (stderr, "/%c", *p++); | ||
| 949 | } | ||
| 950 | while (--mcnt); | ||
| 951 | break; | ||
| 952 | |||
| 953 | case start_memory: | ||
| 954 | fprintf (stderr, "/start_memory/%d", *p++); | ||
| 955 | break; | ||
| 956 | |||
| 957 | case stop_memory: | ||
| 958 | fprintf (stderr, "/stop_memory/%d", *p++); | ||
| 959 | break; | ||
| 960 | |||
| 961 | case duplicate: | ||
| 962 | fprintf (stderr, "/duplicate/%d", *p++); | ||
| 963 | break; | ||
| 964 | |||
| 965 | case anychar: | ||
| 966 | fprintf (stderr, "/anychar"); | ||
| 967 | break; | ||
| 968 | |||
| 969 | case charset: | ||
| 970 | case charset_not: | ||
| 971 | { | ||
| 972 | register int c, last = -100; | ||
| 973 | register int in_range = 0; | ||
| 974 | int length = CHARSET_BITMAP_SIZE (p - 1); | ||
| 975 | int has_range_table = CHARSET_RANGE_TABLE_EXISTS_P (p - 1); | ||
| 976 | |||
| 977 | fprintf (stderr, "/charset [%s", | ||
| 978 | (re_opcode_t) *(p - 1) == charset_not ? "^" : ""); | ||
| 979 | |||
| 980 | if (p + *p >= pend) | ||
| 981 | fprintf (stderr, " !extends past end of pattern! "); | ||
| 982 | |||
| 983 | for (c = 0; c < 256; c++) | ||
| 984 | if (c / 8 < length | ||
| 985 | && (p[1 + (c/8)] & (1 << (c % 8)))) | ||
| 986 | { | ||
| 987 | /* Are we starting a range? */ | ||
| 988 | if (last + 1 == c && ! in_range) | ||
| 989 | { | ||
| 990 | fprintf (stderr, "-"); | ||
| 991 | in_range = 1; | ||
| 992 | } | ||
| 993 | /* Have we broken a range? */ | ||
| 994 | else if (last + 1 != c && in_range) | ||
| 995 | { | ||
| 996 | fprintf (stderr, "%c", last); | ||
| 997 | in_range = 0; | ||
| 998 | } | ||
| 999 | |||
| 1000 | if (! in_range) | ||
| 1001 | fprintf (stderr, "%c", c); | ||
| 1002 | |||
| 1003 | last = c; | ||
| 1004 | } | ||
| 1005 | |||
| 1006 | if (in_range) | ||
| 1007 | fprintf (stderr, "%c", last); | ||
| 1008 | |||
| 1009 | fprintf (stderr, "]"); | ||
| 1010 | |||
| 1011 | p += 1 + length; | ||
| 1012 | |||
| 1013 | if (has_range_table) | ||
| 1014 | { | ||
| 1015 | int count; | ||
| 1016 | fprintf (stderr, "has-range-table"); | ||
| 1017 | |||
| 1018 | /* ??? Should print the range table; for now, just skip it. */ | ||
| 1019 | p += 2; /* skip range table bits */ | ||
| 1020 | EXTRACT_NUMBER_AND_INCR (count, p); | ||
| 1021 | p = CHARSET_RANGE_TABLE_END (p, count); | ||
| 1022 | } | ||
| 1023 | } | ||
| 1024 | break; | ||
| 1025 | |||
| 1026 | case begline: | ||
| 1027 | fprintf (stderr, "/begline"); | ||
| 1028 | break; | ||
| 1029 | |||
| 1030 | case endline: | ||
| 1031 | fprintf (stderr, "/endline"); | ||
| 1032 | break; | ||
| 1033 | |||
| 1034 | case on_failure_jump: | ||
| 1035 | extract_number_and_incr (&mcnt, &p); | ||
| 1036 | fprintf (stderr, "/on_failure_jump to %d", p + mcnt - start); | ||
| 1037 | break; | ||
| 1038 | |||
| 1039 | case on_failure_keep_string_jump: | ||
| 1040 | extract_number_and_incr (&mcnt, &p); | ||
| 1041 | fprintf (stderr, "/on_failure_keep_string_jump to %d", p + mcnt - start); | ||
| 1042 | break; | ||
| 1043 | |||
| 1044 | case on_failure_jump_nastyloop: | ||
| 1045 | extract_number_and_incr (&mcnt, &p); | ||
| 1046 | fprintf (stderr, "/on_failure_jump_nastyloop to %d", p + mcnt - start); | ||
| 1047 | break; | ||
| 1048 | |||
| 1049 | case on_failure_jump_loop: | ||
| 1050 | extract_number_and_incr (&mcnt, &p); | ||
| 1051 | fprintf (stderr, "/on_failure_jump_loop to %d", p + mcnt - start); | ||
| 1052 | break; | ||
| 1053 | |||
| 1054 | case on_failure_jump_smart: | ||
| 1055 | extract_number_and_incr (&mcnt, &p); | ||
| 1056 | fprintf (stderr, "/on_failure_jump_smart to %d", p + mcnt - start); | ||
| 1057 | break; | ||
| 1058 | |||
| 1059 | case jump: | ||
| 1060 | extract_number_and_incr (&mcnt, &p); | ||
| 1061 | fprintf (stderr, "/jump to %d", p + mcnt - start); | ||
| 1062 | break; | ||
| 1063 | |||
| 1064 | case succeed_n: | ||
| 1065 | extract_number_and_incr (&mcnt, &p); | ||
| 1066 | extract_number_and_incr (&mcnt2, &p); | ||
| 1067 | fprintf (stderr, "/succeed_n to %d, %d times", p - 2 + mcnt - start, mcnt2); | ||
| 1068 | break; | ||
| 1069 | |||
| 1070 | case jump_n: | ||
| 1071 | extract_number_and_incr (&mcnt, &p); | ||
| 1072 | extract_number_and_incr (&mcnt2, &p); | ||
| 1073 | fprintf (stderr, "/jump_n to %d, %d times", p - 2 + mcnt - start, mcnt2); | ||
| 1074 | break; | ||
| 1075 | |||
| 1076 | case set_number_at: | ||
| 1077 | extract_number_and_incr (&mcnt, &p); | ||
| 1078 | extract_number_and_incr (&mcnt2, &p); | ||
| 1079 | fprintf (stderr, "/set_number_at location %d to %d", p - 2 + mcnt - start, mcnt2); | ||
| 1080 | break; | ||
| 1081 | |||
| 1082 | case wordbound: | ||
| 1083 | fprintf (stderr, "/wordbound"); | ||
| 1084 | break; | ||
| 1085 | |||
| 1086 | case notwordbound: | ||
| 1087 | fprintf (stderr, "/notwordbound"); | ||
| 1088 | break; | ||
| 1089 | |||
| 1090 | case wordbeg: | ||
| 1091 | fprintf (stderr, "/wordbeg"); | ||
| 1092 | break; | ||
| 1093 | |||
| 1094 | case wordend: | ||
| 1095 | fprintf (stderr, "/wordend"); | ||
| 1096 | |||
| 1097 | case syntaxspec: | ||
| 1098 | fprintf (stderr, "/syntaxspec"); | ||
| 1099 | mcnt = *p++; | ||
| 1100 | fprintf (stderr, "/%d", mcnt); | ||
| 1101 | break; | ||
| 1102 | |||
| 1103 | case notsyntaxspec: | ||
| 1104 | fprintf (stderr, "/notsyntaxspec"); | ||
| 1105 | mcnt = *p++; | ||
| 1106 | fprintf (stderr, "/%d", mcnt); | ||
| 1107 | break; | ||
| 1108 | |||
| 1109 | # ifdef emacs | ||
| 1110 | case before_dot: | ||
| 1111 | fprintf (stderr, "/before_dot"); | ||
| 1112 | break; | ||
| 1113 | |||
| 1114 | case at_dot: | ||
| 1115 | fprintf (stderr, "/at_dot"); | ||
| 1116 | break; | ||
| 1117 | |||
| 1118 | case after_dot: | ||
| 1119 | fprintf (stderr, "/after_dot"); | ||
| 1120 | break; | ||
| 1121 | |||
| 1122 | case categoryspec: | ||
| 1123 | fprintf (stderr, "/categoryspec"); | ||
| 1124 | mcnt = *p++; | ||
| 1125 | fprintf (stderr, "/%d", mcnt); | ||
| 1126 | break; | ||
| 1127 | |||
| 1128 | case notcategoryspec: | ||
| 1129 | fprintf (stderr, "/notcategoryspec"); | ||
| 1130 | mcnt = *p++; | ||
| 1131 | fprintf (stderr, "/%d", mcnt); | ||
| 1132 | break; | ||
| 1133 | # endif /* emacs */ | ||
| 1134 | |||
| 1135 | case begbuf: | ||
| 1136 | fprintf (stderr, "/begbuf"); | ||
| 1137 | break; | ||
| 1138 | |||
| 1139 | case endbuf: | ||
| 1140 | fprintf (stderr, "/endbuf"); | ||
| 1141 | break; | ||
| 1142 | |||
| 1143 | default: | ||
| 1144 | fprintf (stderr, "?%d", *(p-1)); | ||
| 1145 | } | ||
| 1146 | |||
| 1147 | fprintf (stderr, "\n"); | ||
| 1148 | } | ||
| 1149 | |||
| 1150 | fprintf (stderr, "%d:\tend of pattern.\n", p - start); | ||
| 1151 | } | ||
| 1152 | |||
| 1153 | |||
| 1154 | void | ||
| 1155 | print_compiled_pattern (bufp) | ||
| 1156 | struct re_pattern_buffer *bufp; | ||
| 1157 | { | ||
| 1158 | re_char *buffer = bufp->buffer; | ||
| 1159 | |||
| 1160 | print_partial_compiled_pattern (buffer, buffer + bufp->used); | ||
| 1161 | printf ("%ld bytes used/%ld bytes allocated.\n", | ||
| 1162 | bufp->used, bufp->allocated); | ||
| 1163 | |||
| 1164 | if (bufp->fastmap_accurate && bufp->fastmap) | ||
| 1165 | { | ||
| 1166 | printf ("fastmap: "); | ||
| 1167 | print_fastmap (bufp->fastmap); | ||
| 1168 | } | ||
| 1169 | |||
| 1170 | printf ("re_nsub: %d\t", bufp->re_nsub); | ||
| 1171 | printf ("regs_alloc: %d\t", bufp->regs_allocated); | ||
| 1172 | printf ("can_be_null: %d\t", bufp->can_be_null); | ||
| 1173 | printf ("no_sub: %d\t", bufp->no_sub); | ||
| 1174 | printf ("not_bol: %d\t", bufp->not_bol); | ||
| 1175 | printf ("not_eol: %d\t", bufp->not_eol); | ||
| 1176 | printf ("syntax: %lx\n", bufp->syntax); | ||
| 1177 | fflush (stdout); | ||
| 1178 | /* Perhaps we should print the translate table? */ | ||
| 1179 | } | ||
| 1180 | |||
| 1181 | |||
| 1182 | void | ||
| 1183 | print_double_string (where, string1, size1, string2, size2) | ||
| 1184 | re_char *where; | ||
| 1185 | re_char *string1; | ||
| 1186 | re_char *string2; | ||
| 1187 | int size1; | ||
| 1188 | int size2; | ||
| 1189 | { | ||
| 1190 | int this_char; | ||
| 1191 | |||
| 1192 | if (where == NULL) | ||
| 1193 | printf ("(null)"); | ||
| 1194 | else | ||
| 1195 | { | ||
| 1196 | if (FIRST_STRING_P (where)) | ||
| 1197 | { | ||
| 1198 | for (this_char = where - string1; this_char < size1; this_char++) | ||
| 1199 | putchar (string1[this_char]); | ||
| 1200 | |||
| 1201 | where = string2; | ||
| 1202 | } | ||
| 1203 | |||
| 1204 | for (this_char = where - string2; this_char < size2; this_char++) | ||
| 1205 | putchar (string2[this_char]); | ||
| 1206 | } | ||
| 1207 | } | ||
| 1208 | |||
| 1209 | #else /* not DEBUG */ | ||
| 1210 | |||
| 1211 | # undef assert | ||
| 1212 | # define assert(e) | ||
| 1213 | |||
| 1214 | # define DEBUG_STATEMENT(e) | ||
| 1215 | # define DEBUG_PRINT1(x) | ||
| 1216 | # define DEBUG_PRINT2(x1, x2) | ||
| 1217 | # define DEBUG_PRINT3(x1, x2, x3) | ||
| 1218 | # define DEBUG_PRINT4(x1, x2, x3, x4) | ||
| 1219 | # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) | ||
| 1220 | # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) | ||
| 1221 | |||
| 1222 | #endif /* not DEBUG */ | ||
| 1223 | |||
| 1224 | /* Set by `re_set_syntax' to the current regexp syntax to recognize. Can | ||
| 1225 | also be assigned to arbitrarily: each pattern buffer stores its own | ||
| 1226 | syntax, so it can be changed between regex compilations. */ | ||
| 1227 | /* This has no initializer because initialized variables in Emacs | ||
| 1228 | become read-only after dumping. */ | ||
| 1229 | reg_syntax_t re_syntax_options; | ||
| 1230 | |||
| 1231 | |||
| 1232 | /* Specify the precise syntax of regexps for compilation. This provides | ||
| 1233 | for compatibility for various utilities which historically have | ||
| 1234 | different, incompatible syntaxes. | ||
| 1235 | |||
| 1236 | The argument SYNTAX is a bit mask comprised of the various bits | ||
| 1237 | defined in regex.h. We return the old syntax. */ | ||
| 1238 | |||
| 1239 | reg_syntax_t | ||
| 1240 | re_set_syntax (syntax) | ||
| 1241 | reg_syntax_t syntax; | ||
| 1242 | { | ||
| 1243 | reg_syntax_t ret = re_syntax_options; | ||
| 1244 | |||
| 1245 | re_syntax_options = syntax; | ||
| 1246 | return ret; | ||
| 1247 | } | ||
| 1248 | WEAK_ALIAS (__re_set_syntax, re_set_syntax) | ||
| 1249 | |||
| 1250 | /* This table gives an error message for each of the error codes listed | ||
| 1251 | in regex.h. Obviously the order here has to be same as there. | ||
| 1252 | POSIX doesn't require that we do anything for REG_NOERROR, | ||
| 1253 | but why not be nice? */ | ||
| 1254 | |||
| 1255 | static const char *re_error_msgid[] = | ||
| 1256 | { | ||
| 1257 | gettext_noop ("Success"), /* REG_NOERROR */ | ||
| 1258 | gettext_noop ("No match"), /* REG_NOMATCH */ | ||
| 1259 | gettext_noop ("Invalid regular expression"), /* REG_BADPAT */ | ||
| 1260 | gettext_noop ("Invalid collation character"), /* REG_ECOLLATE */ | ||
| 1261 | gettext_noop ("Invalid character class name"), /* REG_ECTYPE */ | ||
| 1262 | gettext_noop ("Trailing backslash"), /* REG_EESCAPE */ | ||
| 1263 | gettext_noop ("Invalid back reference"), /* REG_ESUBREG */ | ||
| 1264 | gettext_noop ("Unmatched [ or [^"), /* REG_EBRACK */ | ||
| 1265 | gettext_noop ("Unmatched ( or \\("), /* REG_EPAREN */ | ||
| 1266 | gettext_noop ("Unmatched \\{"), /* REG_EBRACE */ | ||
| 1267 | gettext_noop ("Invalid content of \\{\\}"), /* REG_BADBR */ | ||
| 1268 | gettext_noop ("Invalid range end"), /* REG_ERANGE */ | ||
| 1269 | gettext_noop ("Memory exhausted"), /* REG_ESPACE */ | ||
| 1270 | gettext_noop ("Invalid preceding regular expression"), /* REG_BADRPT */ | ||
| 1271 | gettext_noop ("Premature end of regular expression"), /* REG_EEND */ | ||
| 1272 | gettext_noop ("Regular expression too big"), /* REG_ESIZE */ | ||
| 1273 | gettext_noop ("Unmatched ) or \\)"), /* REG_ERPAREN */ | ||
| 1274 | }; | ||
| 1275 | |||
| 1276 | /* Avoiding alloca during matching, to placate r_alloc. */ | ||
| 1277 | |||
| 1278 | /* Define MATCH_MAY_ALLOCATE unless we need to make sure that the | ||
| 1279 | searching and matching functions should not call alloca. On some | ||
| 1280 | systems, alloca is implemented in terms of malloc, and if we're | ||
| 1281 | using the relocating allocator routines, then malloc could cause a | ||
| 1282 | relocation, which might (if the strings being searched are in the | ||
| 1283 | ralloc heap) shift the data out from underneath the regexp | ||
| 1284 | routines. | ||
| 1285 | |||
| 1286 | Here's another reason to avoid allocation: Emacs | ||
| 1287 | processes input from X in a signal handler; processing X input may | ||
| 1288 | call malloc; if input arrives while a matching routine is calling | ||
| 1289 | malloc, then we're scrod. But Emacs can't just block input while | ||
| 1290 | calling matching routines; then we don't notice interrupts when | ||
| 1291 | they come in. So, Emacs blocks input around all regexp calls | ||
| 1292 | except the matching calls, which it leaves unprotected, in the | ||
| 1293 | faith that they will not malloc. */ | ||
| 1294 | |||
| 1295 | /* Normally, this is fine. */ | ||
| 1296 | #define MATCH_MAY_ALLOCATE | ||
| 1297 | |||
| 1298 | /* When using GNU C, we are not REALLY using the C alloca, no matter | ||
| 1299 | what config.h may say. So don't take precautions for it. */ | ||
| 1300 | #ifdef __GNUC__ | ||
| 1301 | # undef C_ALLOCA | ||
| 1302 | #endif | ||
| 1303 | |||
| 1304 | /* The match routines may not allocate if (1) they would do it with malloc | ||
| 1305 | and (2) it's not safe for them to use malloc. | ||
| 1306 | Note that if REL_ALLOC is defined, matching would not use malloc for the | ||
| 1307 | failure stack, but we would still use it for the register vectors; | ||
| 1308 | so REL_ALLOC should not affect this. */ | ||
| 1309 | #if (defined C_ALLOCA || defined REGEX_MALLOC) && defined emacs | ||
| 1310 | # undef MATCH_MAY_ALLOCATE | ||
| 1311 | #endif | ||
| 1312 | |||
| 1313 | |||
| 1314 | /* Failure stack declarations and macros; both re_compile_fastmap and | ||
| 1315 | re_match_2 use a failure stack. These have to be macros because of | ||
| 1316 | REGEX_ALLOCATE_STACK. */ | ||
| 1317 | |||
| 1318 | |||
| 1319 | /* Approximate number of failure points for which to initially allocate space | ||
| 1320 | when matching. If this number is exceeded, we allocate more | ||
| 1321 | space, so it is not a hard limit. */ | ||
| 1322 | #ifndef INIT_FAILURE_ALLOC | ||
| 1323 | # define INIT_FAILURE_ALLOC 20 | ||
| 1324 | #endif | ||
| 1325 | |||
| 1326 | /* Roughly the maximum number of failure points on the stack. Would be | ||
| 1327 | exactly that if always used TYPICAL_FAILURE_SIZE items each time we failed. | ||
| 1328 | This is a variable only so users of regex can assign to it; we never | ||
| 1329 | change it ourselves. We always multiply it by TYPICAL_FAILURE_SIZE | ||
| 1330 | before using it, so it should probably be a byte-count instead. */ | ||
| 1331 | # if defined MATCH_MAY_ALLOCATE | ||
| 1332 | /* Note that 4400 was enough to cause a crash on Alpha OSF/1, | ||
| 1333 | whose default stack limit is 2mb. In order for a larger | ||
| 1334 | value to work reliably, you have to try to make it accord | ||
| 1335 | with the process stack limit. */ | ||
| 1336 | size_t re_max_failures = 40000; | ||
| 1337 | # else | ||
| 1338 | size_t re_max_failures = 4000; | ||
| 1339 | # endif | ||
| 1340 | |||
| 1341 | union fail_stack_elt | ||
| 1342 | { | ||
| 1343 | re_char *pointer; | ||
| 1344 | /* This should be the biggest `int' that's no bigger than a pointer. */ | ||
| 1345 | long integer; | ||
| 1346 | }; | ||
| 1347 | |||
| 1348 | typedef union fail_stack_elt fail_stack_elt_t; | ||
| 1349 | |||
| 1350 | typedef struct | ||
| 1351 | { | ||
| 1352 | fail_stack_elt_t *stack; | ||
| 1353 | size_t size; | ||
| 1354 | size_t avail; /* Offset of next open position. */ | ||
| 1355 | size_t frame; /* Offset of the cur constructed frame. */ | ||
| 1356 | } fail_stack_type; | ||
| 1357 | |||
| 1358 | #define FAIL_STACK_EMPTY() (fail_stack.frame == 0) | ||
| 1359 | #define FAIL_STACK_FULL() (fail_stack.avail == fail_stack.size) | ||
| 1360 | |||
| 1361 | |||
| 1362 | /* Define macros to initialize and free the failure stack. | ||
| 1363 | Do `return -2' if the alloc fails. */ | ||
| 1364 | |||
| 1365 | #ifdef MATCH_MAY_ALLOCATE | ||
| 1366 | # define INIT_FAIL_STACK() \ | ||
| 1367 | do { \ | ||
| 1368 | fail_stack.stack = (fail_stack_elt_t *) \ | ||
| 1369 | REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * TYPICAL_FAILURE_SIZE \ | ||
| 1370 | * sizeof (fail_stack_elt_t)); \ | ||
| 1371 | \ | ||
| 1372 | if (fail_stack.stack == NULL) \ | ||
| 1373 | return -2; \ | ||
| 1374 | \ | ||
| 1375 | fail_stack.size = INIT_FAILURE_ALLOC; \ | ||
| 1376 | fail_stack.avail = 0; \ | ||
| 1377 | fail_stack.frame = 0; \ | ||
| 1378 | } while (0) | ||
| 1379 | |||
| 1380 | # define RESET_FAIL_STACK() REGEX_FREE_STACK (fail_stack.stack) | ||
| 1381 | #else | ||
| 1382 | # define INIT_FAIL_STACK() \ | ||
| 1383 | do { \ | ||
| 1384 | fail_stack.avail = 0; \ | ||
| 1385 | fail_stack.frame = 0; \ | ||
| 1386 | } while (0) | ||
| 1387 | |||
| 1388 | # define RESET_FAIL_STACK() ((void)0) | ||
| 1389 | #endif | ||
| 1390 | |||
| 1391 | |||
| 1392 | /* Double the size of FAIL_STACK, up to a limit | ||
| 1393 | which allows approximately `re_max_failures' items. | ||
| 1394 | |||
| 1395 | Return 1 if succeeds, and 0 if either ran out of memory | ||
| 1396 | allocating space for it or it was already too large. | ||
| 1397 | |||
| 1398 | REGEX_REALLOCATE_STACK requires `destination' be declared. */ | ||
| 1399 | |||
| 1400 | /* Factor to increase the failure stack size by | ||
| 1401 | when we increase it. | ||
| 1402 | This used to be 2, but 2 was too wasteful | ||
| 1403 | because the old discarded stacks added up to as much space | ||
| 1404 | were as ultimate, maximum-size stack. */ | ||
| 1405 | #define FAIL_STACK_GROWTH_FACTOR 4 | ||
| 1406 | |||
| 1407 | #define GROW_FAIL_STACK(fail_stack) \ | ||
| 1408 | (((fail_stack).size * sizeof (fail_stack_elt_t) \ | ||
| 1409 | >= re_max_failures * TYPICAL_FAILURE_SIZE) \ | ||
| 1410 | ? 0 \ | ||
| 1411 | : ((fail_stack).stack \ | ||
| 1412 | = (fail_stack_elt_t *) \ | ||
| 1413 | REGEX_REALLOCATE_STACK ((fail_stack).stack, \ | ||
| 1414 | (fail_stack).size * sizeof (fail_stack_elt_t), \ | ||
| 1415 | MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \ | ||
| 1416 | ((fail_stack).size * sizeof (fail_stack_elt_t) \ | ||
| 1417 | * FAIL_STACK_GROWTH_FACTOR))), \ | ||
| 1418 | \ | ||
| 1419 | (fail_stack).stack == NULL \ | ||
| 1420 | ? 0 \ | ||
| 1421 | : ((fail_stack).size \ | ||
| 1422 | = (MIN (re_max_failures * TYPICAL_FAILURE_SIZE, \ | ||
| 1423 | ((fail_stack).size * sizeof (fail_stack_elt_t) \ | ||
| 1424 | * FAIL_STACK_GROWTH_FACTOR)) \ | ||
| 1425 | / sizeof (fail_stack_elt_t)), \ | ||
| 1426 | 1))) | ||
| 1427 | |||
| 1428 | |||
| 1429 | /* Push a pointer value onto the failure stack. | ||
| 1430 | Assumes the variable `fail_stack'. Probably should only | ||
| 1431 | be called from within `PUSH_FAILURE_POINT'. */ | ||
| 1432 | #define PUSH_FAILURE_POINTER(item) \ | ||
| 1433 | fail_stack.stack[fail_stack.avail++].pointer = (item) | ||
| 1434 | |||
| 1435 | /* This pushes an integer-valued item onto the failure stack. | ||
| 1436 | Assumes the variable `fail_stack'. Probably should only | ||
| 1437 | be called from within `PUSH_FAILURE_POINT'. */ | ||
| 1438 | #define PUSH_FAILURE_INT(item) \ | ||
| 1439 | fail_stack.stack[fail_stack.avail++].integer = (item) | ||
| 1440 | |||
| 1441 | /* Push a fail_stack_elt_t value onto the failure stack. | ||
| 1442 | Assumes the variable `fail_stack'. Probably should only | ||
| 1443 | be called from within `PUSH_FAILURE_POINT'. */ | ||
| 1444 | #define PUSH_FAILURE_ELT(item) \ | ||
| 1445 | fail_stack.stack[fail_stack.avail++] = (item) | ||
| 1446 | |||
| 1447 | /* These three POP... operations complement the three PUSH... operations. | ||
| 1448 | All assume that `fail_stack' is nonempty. */ | ||
| 1449 | #define POP_FAILURE_POINTER() fail_stack.stack[--fail_stack.avail].pointer | ||
| 1450 | #define POP_FAILURE_INT() fail_stack.stack[--fail_stack.avail].integer | ||
| 1451 | #define POP_FAILURE_ELT() fail_stack.stack[--fail_stack.avail] | ||
| 1452 | |||
| 1453 | /* Individual items aside from the registers. */ | ||
| 1454 | #define NUM_NONREG_ITEMS 3 | ||
| 1455 | |||
| 1456 | /* Used to examine the stack (to detect infinite loops). */ | ||
| 1457 | #define FAILURE_PAT(h) fail_stack.stack[(h) - 1].pointer | ||
| 1458 | #define FAILURE_STR(h) (fail_stack.stack[(h) - 2].pointer) | ||
| 1459 | #define NEXT_FAILURE_HANDLE(h) fail_stack.stack[(h) - 3].integer | ||
| 1460 | #define TOP_FAILURE_HANDLE() fail_stack.frame | ||
| 1461 | |||
| 1462 | |||
| 1463 | #define ENSURE_FAIL_STACK(space) \ | ||
| 1464 | while (REMAINING_AVAIL_SLOTS <= space) { \ | ||
| 1465 | if (!GROW_FAIL_STACK (fail_stack)) \ | ||
| 1466 | return -2; \ | ||
| 1467 | DEBUG_PRINT2 ("\n Doubled stack; size now: %d\n", (fail_stack).size);\ | ||
| 1468 | DEBUG_PRINT2 (" slots available: %d\n", REMAINING_AVAIL_SLOTS);\ | ||
| 1469 | } | ||
| 1470 | |||
| 1471 | /* Push register NUM onto the stack. */ | ||
| 1472 | #define PUSH_FAILURE_REG(num) \ | ||
| 1473 | do { \ | ||
| 1474 | char *destination; \ | ||
| 1475 | ENSURE_FAIL_STACK(3); \ | ||
| 1476 | DEBUG_PRINT4 (" Push reg %d (spanning %p -> %p)\n", \ | ||
| 1477 | num, regstart[num], regend[num]); \ | ||
| 1478 | PUSH_FAILURE_POINTER (regstart[num]); \ | ||
| 1479 | PUSH_FAILURE_POINTER (regend[num]); \ | ||
| 1480 | PUSH_FAILURE_INT (num); \ | ||
| 1481 | } while (0) | ||
| 1482 | |||
| 1483 | /* Change the counter's value to VAL, but make sure that it will | ||
| 1484 | be reset when backtracking. */ | ||
| 1485 | #define PUSH_NUMBER(ptr,val) \ | ||
| 1486 | do { \ | ||
| 1487 | char *destination; \ | ||
| 1488 | int c; \ | ||
| 1489 | ENSURE_FAIL_STACK(3); \ | ||
| 1490 | EXTRACT_NUMBER (c, ptr); \ | ||
| 1491 | DEBUG_PRINT4 (" Push number %p = %d -> %d\n", ptr, c, val); \ | ||
| 1492 | PUSH_FAILURE_INT (c); \ | ||
| 1493 | PUSH_FAILURE_POINTER (ptr); \ | ||
| 1494 | PUSH_FAILURE_INT (-1); \ | ||
| 1495 | STORE_NUMBER (ptr, val); \ | ||
| 1496 | } while (0) | ||
| 1497 | |||
| 1498 | /* Pop a saved register off the stack. */ | ||
| 1499 | #define POP_FAILURE_REG_OR_COUNT() \ | ||
| 1500 | do { \ | ||
| 1501 | int reg = POP_FAILURE_INT (); \ | ||
| 1502 | if (reg == -1) \ | ||
| 1503 | { \ | ||
| 1504 | /* It's a counter. */ \ | ||
| 1505 | /* Here, we discard `const', making re_match non-reentrant. */ \ | ||
| 1506 | unsigned char *ptr = (unsigned char*) POP_FAILURE_POINTER (); \ | ||
| 1507 | reg = POP_FAILURE_INT (); \ | ||
| 1508 | STORE_NUMBER (ptr, reg); \ | ||
| 1509 | DEBUG_PRINT3 (" Pop counter %p = %d\n", ptr, reg); \ | ||
| 1510 | } \ | ||
| 1511 | else \ | ||
| 1512 | { \ | ||
| 1513 | regend[reg] = POP_FAILURE_POINTER (); \ | ||
| 1514 | regstart[reg] = POP_FAILURE_POINTER (); \ | ||
| 1515 | DEBUG_PRINT4 (" Pop reg %d (spanning %p -> %p)\n", \ | ||
| 1516 | reg, regstart[reg], regend[reg]); \ | ||
| 1517 | } \ | ||
| 1518 | } while (0) | ||
| 1519 | |||
| 1520 | /* Check that we are not stuck in an infinite loop. */ | ||
| 1521 | #define CHECK_INFINITE_LOOP(pat_cur, string_place) \ | ||
| 1522 | do { \ | ||
| 1523 | int failure = TOP_FAILURE_HANDLE (); \ | ||
| 1524 | /* Check for infinite matching loops */ \ | ||
| 1525 | while (failure > 0 \ | ||
| 1526 | && (FAILURE_STR (failure) == string_place \ | ||
| 1527 | || FAILURE_STR (failure) == NULL)) \ | ||
| 1528 | { \ | ||
| 1529 | assert (FAILURE_PAT (failure) >= bufp->buffer \ | ||
| 1530 | && FAILURE_PAT (failure) <= bufp->buffer + bufp->used); \ | ||
| 1531 | if (FAILURE_PAT (failure) == pat_cur) \ | ||
| 1532 | { \ | ||
| 1533 | cycle = 1; \ | ||
| 1534 | break; \ | ||
| 1535 | } \ | ||
| 1536 | DEBUG_PRINT2 (" Other pattern: %p\n", FAILURE_PAT (failure)); \ | ||
| 1537 | failure = NEXT_FAILURE_HANDLE(failure); \ | ||
| 1538 | } \ | ||
| 1539 | DEBUG_PRINT2 (" Other string: %p\n", FAILURE_STR (failure)); \ | ||
| 1540 | } while (0) | ||
| 1541 | |||
| 1542 | /* Push the information about the state we will need | ||
| 1543 | if we ever fail back to it. | ||
| 1544 | |||
| 1545 | Requires variables fail_stack, regstart, regend and | ||
| 1546 | num_regs be declared. GROW_FAIL_STACK requires `destination' be | ||
| 1547 | declared. | ||
| 1548 | |||
| 1549 | Does `return FAILURE_CODE' if runs out of memory. */ | ||
| 1550 | |||
| 1551 | #define PUSH_FAILURE_POINT(pattern, string_place) \ | ||
| 1552 | do { \ | ||
| 1553 | char *destination; \ | ||
| 1554 | /* Must be int, so when we don't save any registers, the arithmetic \ | ||
| 1555 | of 0 + -1 isn't done as unsigned. */ \ | ||
| 1556 | \ | ||
| 1557 | DEBUG_STATEMENT (nfailure_points_pushed++); \ | ||
| 1558 | DEBUG_PRINT1 ("\nPUSH_FAILURE_POINT:\n"); \ | ||
| 1559 | DEBUG_PRINT2 (" Before push, next avail: %d\n", (fail_stack).avail); \ | ||
| 1560 | DEBUG_PRINT2 (" size: %d\n", (fail_stack).size);\ | ||
| 1561 | \ | ||
| 1562 | ENSURE_FAIL_STACK (NUM_NONREG_ITEMS); \ | ||
| 1563 | \ | ||
| 1564 | DEBUG_PRINT1 ("\n"); \ | ||
| 1565 | \ | ||
| 1566 | DEBUG_PRINT2 (" Push frame index: %d\n", fail_stack.frame); \ | ||
| 1567 | PUSH_FAILURE_INT (fail_stack.frame); \ | ||
| 1568 | \ | ||
| 1569 | DEBUG_PRINT2 (" Push string %p: `", string_place); \ | ||
| 1570 | DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, size2);\ | ||
| 1571 | DEBUG_PRINT1 ("'\n"); \ | ||
| 1572 | PUSH_FAILURE_POINTER (string_place); \ | ||
| 1573 | \ | ||
| 1574 | DEBUG_PRINT2 (" Push pattern %p: ", pattern); \ | ||
| 1575 | DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern, pend); \ | ||
| 1576 | PUSH_FAILURE_POINTER (pattern); \ | ||
| 1577 | \ | ||
| 1578 | /* Close the frame by moving the frame pointer past it. */ \ | ||
| 1579 | fail_stack.frame = fail_stack.avail; \ | ||
| 1580 | } while (0) | ||
| 1581 | |||
| 1582 | /* Estimate the size of data pushed by a typical failure stack entry. | ||
| 1583 | An estimate is all we need, because all we use this for | ||
| 1584 | is to choose a limit for how big to make the failure stack. */ | ||
| 1585 | /* BEWARE, the value `20' is hard-coded in emacs.c:main(). */ | ||
| 1586 | #define TYPICAL_FAILURE_SIZE 20 | ||
| 1587 | |||
| 1588 | /* How many items can still be added to the stack without overflowing it. */ | ||
| 1589 | #define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail) | ||
| 1590 | |||
| 1591 | |||
| 1592 | /* Pops what PUSH_FAIL_STACK pushes. | ||
| 1593 | |||
| 1594 | We restore into the parameters, all of which should be lvalues: | ||
| 1595 | STR -- the saved data position. | ||
| 1596 | PAT -- the saved pattern position. | ||
| 1597 | REGSTART, REGEND -- arrays of string positions. | ||
| 1598 | |||
| 1599 | Also assumes the variables `fail_stack' and (if debugging), `bufp', | ||
| 1600 | `pend', `string1', `size1', `string2', and `size2'. */ | ||
| 1601 | |||
| 1602 | #define POP_FAILURE_POINT(str, pat) \ | ||
| 1603 | do { \ | ||
| 1604 | assert (!FAIL_STACK_EMPTY ()); \ | ||
| 1605 | \ | ||
| 1606 | /* Remove failure points and point to how many regs pushed. */ \ | ||
| 1607 | DEBUG_PRINT1 ("POP_FAILURE_POINT:\n"); \ | ||
| 1608 | DEBUG_PRINT2 (" Before pop, next avail: %d\n", fail_stack.avail); \ | ||
| 1609 | DEBUG_PRINT2 (" size: %d\n", fail_stack.size); \ | ||
| 1610 | \ | ||
| 1611 | /* Pop the saved registers. */ \ | ||
| 1612 | while (fail_stack.frame < fail_stack.avail) \ | ||
| 1613 | POP_FAILURE_REG_OR_COUNT (); \ | ||
| 1614 | \ | ||
| 1615 | pat = POP_FAILURE_POINTER (); \ | ||
| 1616 | DEBUG_PRINT2 (" Popping pattern %p: ", pat); \ | ||
| 1617 | DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \ | ||
| 1618 | \ | ||
| 1619 | /* If the saved string location is NULL, it came from an \ | ||
| 1620 | on_failure_keep_string_jump opcode, and we want to throw away the \ | ||
| 1621 | saved NULL, thus retaining our current position in the string. */ \ | ||
| 1622 | str = POP_FAILURE_POINTER (); \ | ||
| 1623 | DEBUG_PRINT2 (" Popping string %p: `", str); \ | ||
| 1624 | DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \ | ||
| 1625 | DEBUG_PRINT1 ("'\n"); \ | ||
| 1626 | \ | ||
| 1627 | fail_stack.frame = POP_FAILURE_INT (); \ | ||
| 1628 | DEBUG_PRINT2 (" Popping frame index: %d\n", fail_stack.frame); \ | ||
| 1629 | \ | ||
| 1630 | assert (fail_stack.avail >= 0); \ | ||
| 1631 | assert (fail_stack.frame <= fail_stack.avail); \ | ||
| 1632 | \ | ||
| 1633 | DEBUG_STATEMENT (nfailure_points_popped++); \ | ||
| 1634 | } while (0) /* POP_FAILURE_POINT */ | ||
| 1635 | |||
| 1636 | |||
| 1637 | |||
| 1638 | /* Registers are set to a sentinel when they haven't yet matched. */ | ||
| 1639 | #define REG_UNSET(e) ((e) == NULL) | ||
| 1640 | |||
| 1641 | /* Subroutine declarations and macros for regex_compile. */ | ||
| 1642 | |||
| 1643 | static reg_errcode_t regex_compile _RE_ARGS ((re_char *pattern, size_t size, | ||
| 1644 | reg_syntax_t syntax, | ||
| 1645 | struct re_pattern_buffer *bufp)); | ||
| 1646 | static void store_op1 _RE_ARGS ((re_opcode_t op, unsigned char *loc, int arg)); | ||
| 1647 | static void store_op2 _RE_ARGS ((re_opcode_t op, unsigned char *loc, | ||
| 1648 | int arg1, int arg2)); | ||
| 1649 | static void insert_op1 _RE_ARGS ((re_opcode_t op, unsigned char *loc, | ||
| 1650 | int arg, unsigned char *end)); | ||
| 1651 | static void insert_op2 _RE_ARGS ((re_opcode_t op, unsigned char *loc, | ||
| 1652 | int arg1, int arg2, unsigned char *end)); | ||
| 1653 | static boolean at_begline_loc_p _RE_ARGS ((re_char *pattern, | ||
| 1654 | re_char *p, | ||
| 1655 | reg_syntax_t syntax)); | ||
| 1656 | static boolean at_endline_loc_p _RE_ARGS ((re_char *p, | ||
| 1657 | re_char *pend, | ||
| 1658 | reg_syntax_t syntax)); | ||
| 1659 | static re_char *skip_one_char _RE_ARGS ((re_char *p)); | ||
| 1660 | static int analyse_first _RE_ARGS ((re_char *p, re_char *pend, | ||
| 1661 | char *fastmap, const int multibyte)); | ||
| 1662 | |||
| 1663 | /* Fetch the next character in the uncompiled pattern, with no | ||
| 1664 | translation. */ | ||
| 1665 | #define PATFETCH(c) \ | ||
| 1666 | do { \ | ||
| 1667 | int len; \ | ||
| 1668 | if (p == pend) return REG_EEND; \ | ||
| 1669 | c = RE_STRING_CHAR_AND_LENGTH (p, pend - p, len); \ | ||
| 1670 | p += len; \ | ||
| 1671 | } while (0) | ||
| 1672 | |||
| 1673 | |||
| 1674 | /* If `translate' is non-null, return translate[D], else just D. We | ||
| 1675 | cast the subscript to translate because some data is declared as | ||
| 1676 | `char *', to avoid warnings when a string constant is passed. But | ||
| 1677 | when we use a character as a subscript we must make it unsigned. */ | ||
| 1678 | #ifndef TRANSLATE | ||
| 1679 | # define TRANSLATE(d) \ | ||
| 1680 | (RE_TRANSLATE_P (translate) ? RE_TRANSLATE (translate, (d)) : (d)) | ||
| 1681 | #endif | ||
| 1682 | |||
| 1683 | |||
| 1684 | /* Macros for outputting the compiled pattern into `buffer'. */ | ||
| 1685 | |||
| 1686 | /* If the buffer isn't allocated when it comes in, use this. */ | ||
| 1687 | #define INIT_BUF_SIZE 32 | ||
| 1688 | |||
| 1689 | /* Make sure we have at least N more bytes of space in buffer. */ | ||
| 1690 | #define GET_BUFFER_SPACE(n) \ | ||
| 1691 | while ((size_t) (b - bufp->buffer + (n)) > bufp->allocated) \ | ||
| 1692 | EXTEND_BUFFER () | ||
| 1693 | |||
| 1694 | /* Make sure we have one more byte of buffer space and then add C to it. */ | ||
| 1695 | #define BUF_PUSH(c) \ | ||
| 1696 | do { \ | ||
| 1697 | GET_BUFFER_SPACE (1); \ | ||
| 1698 | *b++ = (unsigned char) (c); \ | ||
| 1699 | } while (0) | ||
| 1700 | |||
| 1701 | |||
| 1702 | /* Ensure we have two more bytes of buffer space and then append C1 and C2. */ | ||
| 1703 | #define BUF_PUSH_2(c1, c2) \ | ||
| 1704 | do { \ | ||
| 1705 | GET_BUFFER_SPACE (2); \ | ||
| 1706 | *b++ = (unsigned char) (c1); \ | ||
| 1707 | *b++ = (unsigned char) (c2); \ | ||
| 1708 | } while (0) | ||
| 1709 | |||
| 1710 | |||
| 1711 | /* As with BUF_PUSH_2, except for three bytes. */ | ||
| 1712 | #define BUF_PUSH_3(c1, c2, c3) \ | ||
| 1713 | do { \ | ||
| 1714 | GET_BUFFER_SPACE (3); \ | ||
| 1715 | *b++ = (unsigned char) (c1); \ | ||
| 1716 | *b++ = (unsigned char) (c2); \ | ||
| 1717 | *b++ = (unsigned char) (c3); \ | ||
| 1718 | } while (0) | ||
| 1719 | |||
| 1720 | |||
| 1721 | /* Store a jump with opcode OP at LOC to location TO. We store a | ||
| 1722 | relative address offset by the three bytes the jump itself occupies. */ | ||
| 1723 | #define STORE_JUMP(op, loc, to) \ | ||
| 1724 | store_op1 (op, loc, (to) - (loc) - 3) | ||
| 1725 | |||
| 1726 | /* Likewise, for a two-argument jump. */ | ||
| 1727 | #define STORE_JUMP2(op, loc, to, arg) \ | ||
| 1728 | store_op2 (op, loc, (to) - (loc) - 3, arg) | ||
| 1729 | |||
| 1730 | /* Like `STORE_JUMP', but for inserting. Assume `b' is the buffer end. */ | ||
| 1731 | #define INSERT_JUMP(op, loc, to) \ | ||
| 1732 | insert_op1 (op, loc, (to) - (loc) - 3, b) | ||
| 1733 | |||
| 1734 | /* Like `STORE_JUMP2', but for inserting. Assume `b' is the buffer end. */ | ||
| 1735 | #define INSERT_JUMP2(op, loc, to, arg) \ | ||
| 1736 | insert_op2 (op, loc, (to) - (loc) - 3, arg, b) | ||
| 1737 | |||
| 1738 | |||
| 1739 | /* This is not an arbitrary limit: the arguments which represent offsets | ||
| 1740 | into the pattern are two bytes long. So if 2^15 bytes turns out to | ||
| 1741 | be too small, many things would have to change. */ | ||
| 1742 | # define MAX_BUF_SIZE (1L << 15) | ||
| 1743 | |||
| 1744 | #if 0 /* This is when we thought it could be 2^16 bytes. */ | ||
| 1745 | /* Any other compiler which, like MSC, has allocation limit below 2^16 | ||
| 1746 | bytes will have to use approach similar to what was done below for | ||
| 1747 | MSC and drop MAX_BUF_SIZE a bit. Otherwise you may end up | ||
| 1748 | reallocating to 0 bytes. Such thing is not going to work too well. | ||
| 1749 | You have been warned!! */ | ||
| 1750 | #if defined _MSC_VER && !defined WIN32 | ||
| 1751 | /* Microsoft C 16-bit versions limit malloc to approx 65512 bytes. */ | ||
| 1752 | # define MAX_BUF_SIZE 65500L | ||
| 1753 | #else | ||
| 1754 | # define MAX_BUF_SIZE (1L << 16) | ||
| 1755 | #endif | ||
| 1756 | #endif /* 0 */ | ||
| 1757 | |||
| 1758 | /* Extend the buffer by twice its current size via realloc and | ||
| 1759 | reset the pointers that pointed into the old block to point to the | ||
| 1760 | correct places in the new one. If extending the buffer results in it | ||
| 1761 | being larger than MAX_BUF_SIZE, then flag memory exhausted. */ | ||
| 1762 | #if __BOUNDED_POINTERS__ | ||
| 1763 | # define SET_HIGH_BOUND(P) (__ptrhigh (P) = __ptrlow (P) + bufp->allocated) | ||
| 1764 | # define MOVE_BUFFER_POINTER(P) \ | ||
| 1765 | (__ptrlow (P) += incr, SET_HIGH_BOUND (P), __ptrvalue (P) += incr) | ||
| 1766 | # define ELSE_EXTEND_BUFFER_HIGH_BOUND \ | ||
| 1767 | else \ | ||
| 1768 | { \ | ||
| 1769 | SET_HIGH_BOUND (b); \ | ||
| 1770 | SET_HIGH_BOUND (begalt); \ | ||
| 1771 | if (fixup_alt_jump) \ | ||
| 1772 | SET_HIGH_BOUND (fixup_alt_jump); \ | ||
| 1773 | if (laststart) \ | ||
| 1774 | SET_HIGH_BOUND (laststart); \ | ||
| 1775 | if (pending_exact) \ | ||
| 1776 | SET_HIGH_BOUND (pending_exact); \ | ||
| 1777 | } | ||
| 1778 | #else | ||
| 1779 | # define MOVE_BUFFER_POINTER(P) (P) += incr | ||
| 1780 | # define ELSE_EXTEND_BUFFER_HIGH_BOUND | ||
| 1781 | #endif | ||
| 1782 | #define EXTEND_BUFFER() \ | ||
| 1783 | do { \ | ||
| 1784 | re_char *old_buffer = bufp->buffer; \ | ||
| 1785 | if (bufp->allocated == MAX_BUF_SIZE) \ | ||
| 1786 | return REG_ESIZE; \ | ||
| 1787 | bufp->allocated <<= 1; \ | ||
| 1788 | if (bufp->allocated > MAX_BUF_SIZE) \ | ||
| 1789 | bufp->allocated = MAX_BUF_SIZE; \ | ||
| 1790 | RETALLOC (bufp->buffer, bufp->allocated, unsigned char); \ | ||
| 1791 | if (bufp->buffer == NULL) \ | ||
| 1792 | return REG_ESPACE; \ | ||
| 1793 | /* If the buffer moved, move all the pointers into it. */ \ | ||
| 1794 | if (old_buffer != bufp->buffer) \ | ||
| 1795 | { \ | ||
| 1796 | int incr = bufp->buffer - old_buffer; \ | ||
| 1797 | MOVE_BUFFER_POINTER (b); \ | ||
| 1798 | MOVE_BUFFER_POINTER (begalt); \ | ||
| 1799 | if (fixup_alt_jump) \ | ||
| 1800 | MOVE_BUFFER_POINTER (fixup_alt_jump); \ | ||
| 1801 | if (laststart) \ | ||
| 1802 | MOVE_BUFFER_POINTER (laststart); \ | ||
| 1803 | if (pending_exact) \ | ||
| 1804 | MOVE_BUFFER_POINTER (pending_exact); \ | ||
| 1805 | } \ | ||
| 1806 | ELSE_EXTEND_BUFFER_HIGH_BOUND \ | ||
| 1807 | } while (0) | ||
| 1808 | |||
| 1809 | |||
| 1810 | /* Since we have one byte reserved for the register number argument to | ||
| 1811 | {start,stop}_memory, the maximum number of groups we can report | ||
| 1812 | things about is what fits in that byte. */ | ||
| 1813 | #define MAX_REGNUM 255 | ||
| 1814 | |||
| 1815 | /* But patterns can have more than `MAX_REGNUM' registers. We just | ||
| 1816 | ignore the excess. */ | ||
| 1817 | typedef int regnum_t; | ||
| 1818 | |||
| 1819 | |||
| 1820 | /* Macros for the compile stack. */ | ||
| 1821 | |||
| 1822 | /* Since offsets can go either forwards or backwards, this type needs to | ||
| 1823 | be able to hold values from -(MAX_BUF_SIZE - 1) to MAX_BUF_SIZE - 1. */ | ||
| 1824 | /* int may be not enough when sizeof(int) == 2. */ | ||
| 1825 | typedef long pattern_offset_t; | ||
| 1826 | |||
| 1827 | typedef struct | ||
| 1828 | { | ||
| 1829 | pattern_offset_t begalt_offset; | ||
| 1830 | pattern_offset_t fixup_alt_jump; | ||
| 1831 | pattern_offset_t laststart_offset; | ||
| 1832 | regnum_t regnum; | ||
| 1833 | } compile_stack_elt_t; | ||
| 1834 | |||
| 1835 | |||
| 1836 | typedef struct | ||
| 1837 | { | ||
| 1838 | compile_stack_elt_t *stack; | ||
| 1839 | unsigned size; | ||
| 1840 | unsigned avail; /* Offset of next open position. */ | ||
| 1841 | } compile_stack_type; | ||
| 1842 | |||
| 1843 | |||
| 1844 | #define INIT_COMPILE_STACK_SIZE 32 | ||
| 1845 | |||
| 1846 | #define COMPILE_STACK_EMPTY (compile_stack.avail == 0) | ||
| 1847 | #define COMPILE_STACK_FULL (compile_stack.avail == compile_stack.size) | ||
| 1848 | |||
| 1849 | /* The next available element. */ | ||
| 1850 | #define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail]) | ||
| 1851 | |||
| 1852 | /* Explicit quit checking is only used on NTemacs. */ | ||
| 1853 | #if defined WINDOWSNT && defined emacs && defined QUIT | ||
| 1854 | extern int immediate_quit; | ||
| 1855 | # define IMMEDIATE_QUIT_CHECK \ | ||
| 1856 | do { \ | ||
| 1857 | if (immediate_quit) QUIT; \ | ||
| 1858 | } while (0) | ||
| 1859 | #else | ||
| 1860 | # define IMMEDIATE_QUIT_CHECK ((void)0) | ||
| 1861 | #endif | ||
| 1862 | |||
| 1863 | /* Structure to manage work area for range table. */ | ||
| 1864 | struct range_table_work_area | ||
| 1865 | { | ||
| 1866 | int *table; /* actual work area. */ | ||
| 1867 | int allocated; /* allocated size for work area in bytes. */ | ||
| 1868 | int used; /* actually used size in words. */ | ||
| 1869 | int bits; /* flag to record character classes */ | ||
| 1870 | }; | ||
| 1871 | |||
| 1872 | /* Make sure that WORK_AREA can hold more N multibyte characters. | ||
| 1873 | This is used only in set_image_of_range and set_image_of_range_1. | ||
| 1874 | It expects WORK_AREA to be a pointer. | ||
| 1875 | If it can't get the space, it returns from the surrounding function. */ | ||
| 1876 | |||
| 1877 | #define EXTEND_RANGE_TABLE(work_area, n) \ | ||
| 1878 | do { \ | ||
| 1879 | if (((work_area)->used + (n)) * sizeof (int) > (work_area)->allocated) \ | ||
| 1880 | { \ | ||
| 1881 | extend_range_table_work_area (work_area); \ | ||
| 1882 | if ((work_area)->table == 0) \ | ||
| 1883 | return (REG_ESPACE); \ | ||
| 1884 | } \ | ||
| 1885 | } while (0) | ||
| 1886 | |||
| 1887 | #define SET_RANGE_TABLE_WORK_AREA_BIT(work_area, bit) \ | ||
| 1888 | (work_area).bits |= (bit) | ||
| 1889 | |||
| 1890 | /* Bits used to implement the multibyte-part of the various character classes | ||
| 1891 | such as [:alnum:] in a charset's range table. */ | ||
| 1892 | #define BIT_WORD 0x1 | ||
| 1893 | #define BIT_LOWER 0x2 | ||
| 1894 | #define BIT_PUNCT 0x4 | ||
| 1895 | #define BIT_SPACE 0x8 | ||
| 1896 | #define BIT_UPPER 0x10 | ||
| 1897 | #define BIT_MULTIBYTE 0x20 | ||
| 1898 | |||
| 1899 | /* Set a range START..END to WORK_AREA. | ||
| 1900 | The range is passed through TRANSLATE, so START and END | ||
| 1901 | should be untranslated. */ | ||
| 1902 | #define SET_RANGE_TABLE_WORK_AREA(work_area, start, end) \ | ||
| 1903 | do { \ | ||
| 1904 | int tem; \ | ||
| 1905 | tem = set_image_of_range (&work_area, start, end, translate); \ | ||
| 1906 | if (tem > 0) \ | ||
| 1907 | FREE_STACK_RETURN (tem); \ | ||
| 1908 | } while (0) | ||
| 1909 | |||
| 1910 | /* Free allocated memory for WORK_AREA. */ | ||
| 1911 | #define FREE_RANGE_TABLE_WORK_AREA(work_area) \ | ||
| 1912 | do { \ | ||
| 1913 | if ((work_area).table) \ | ||
| 1914 | free ((work_area).table); \ | ||
| 1915 | } while (0) | ||
| 1916 | |||
| 1917 | #define CLEAR_RANGE_TABLE_WORK_USED(work_area) ((work_area).used = 0, (work_area).bits = 0) | ||
| 1918 | #define RANGE_TABLE_WORK_USED(work_area) ((work_area).used) | ||
| 1919 | #define RANGE_TABLE_WORK_BITS(work_area) ((work_area).bits) | ||
| 1920 | #define RANGE_TABLE_WORK_ELT(work_area, i) ((work_area).table[i]) | ||
| 1921 | |||
| 1922 | |||
| 1923 | /* Set the bit for character C in a list. */ | ||
| 1924 | #define SET_LIST_BIT(c) (b[((c)) / BYTEWIDTH] |= 1 << ((c) % BYTEWIDTH)) | ||
| 1925 | |||
| 1926 | |||
| 1927 | /* Get the next unsigned number in the uncompiled pattern. */ | ||
| 1928 | #define GET_UNSIGNED_NUMBER(num) \ | ||
| 1929 | do { if (p != pend) \ | ||
| 1930 | { \ | ||
| 1931 | PATFETCH (c); \ | ||
| 1932 | if (c == ' ') \ | ||
| 1933 | FREE_STACK_RETURN (REG_BADBR); \ | ||
| 1934 | while ('0' <= c && c <= '9') \ | ||
| 1935 | { \ | ||
| 1936 | int prev; \ | ||
| 1937 | if (num < 0) \ | ||
| 1938 | num = 0; \ | ||
| 1939 | prev = num; \ | ||
| 1940 | num = num * 10 + c - '0'; \ | ||
| 1941 | if (num / 10 != prev) \ | ||
| 1942 | FREE_STACK_RETURN (REG_BADBR); \ | ||
| 1943 | if (p == pend) \ | ||
| 1944 | break; \ | ||
| 1945 | PATFETCH (c); \ | ||
| 1946 | } \ | ||
| 1947 | if (c == ' ') \ | ||
| 1948 | FREE_STACK_RETURN (REG_BADBR); \ | ||
| 1949 | } \ | ||
| 1950 | } while (0) | ||
| 1951 | |||
| 1952 | #if WIDE_CHAR_SUPPORT | ||
| 1953 | /* The GNU C library provides support for user-defined character classes | ||
| 1954 | and the functions from ISO C amendement 1. */ | ||
| 1955 | # ifdef CHARCLASS_NAME_MAX | ||
| 1956 | # define CHAR_CLASS_MAX_LENGTH CHARCLASS_NAME_MAX | ||
| 1957 | # else | ||
| 1958 | /* This shouldn't happen but some implementation might still have this | ||
| 1959 | problem. Use a reasonable default value. */ | ||
| 1960 | # define CHAR_CLASS_MAX_LENGTH 256 | ||
| 1961 | # endif | ||
| 1962 | typedef wctype_t re_wctype_t; | ||
| 1963 | typedef wchar_t re_wchar_t; | ||
| 1964 | # define re_wctype wctype | ||
| 1965 | # define re_iswctype iswctype | ||
| 1966 | # define re_wctype_to_bit(cc) 0 | ||
| 1967 | #else | ||
| 1968 | # define CHAR_CLASS_MAX_LENGTH 9 /* Namely, `multibyte'. */ | ||
| 1969 | # define btowc(c) c | ||
| 1970 | |||
| 1971 | /* Character classes. */ | ||
| 1972 | typedef enum { RECC_ERROR = 0, | ||
| 1973 | RECC_ALNUM, RECC_ALPHA, RECC_WORD, | ||
| 1974 | RECC_GRAPH, RECC_PRINT, | ||
| 1975 | RECC_LOWER, RECC_UPPER, | ||
| 1976 | RECC_PUNCT, RECC_CNTRL, | ||
| 1977 | RECC_DIGIT, RECC_XDIGIT, | ||
| 1978 | RECC_BLANK, RECC_SPACE, | ||
| 1979 | RECC_MULTIBYTE, RECC_NONASCII, | ||
| 1980 | RECC_ASCII, RECC_UNIBYTE | ||
| 1981 | } re_wctype_t; | ||
| 1982 | |||
| 1983 | typedef int re_wchar_t; | ||
| 1984 | |||
| 1985 | /* Map a string to the char class it names (if any). */ | ||
| 1986 | static re_wctype_t | ||
| 1987 | re_wctype (str) | ||
| 1988 | re_char *str; | ||
| 1989 | { | ||
| 1990 | const char *string = str; | ||
| 1991 | if (STREQ (string, "alnum")) return RECC_ALNUM; | ||
| 1992 | else if (STREQ (string, "alpha")) return RECC_ALPHA; | ||
| 1993 | else if (STREQ (string, "word")) return RECC_WORD; | ||
| 1994 | else if (STREQ (string, "ascii")) return RECC_ASCII; | ||
| 1995 | else if (STREQ (string, "nonascii")) return RECC_NONASCII; | ||
| 1996 | else if (STREQ (string, "graph")) return RECC_GRAPH; | ||
| 1997 | else if (STREQ (string, "lower")) return RECC_LOWER; | ||
| 1998 | else if (STREQ (string, "print")) return RECC_PRINT; | ||
| 1999 | else if (STREQ (string, "punct")) return RECC_PUNCT; | ||
| 2000 | else if (STREQ (string, "space")) return RECC_SPACE; | ||
| 2001 | else if (STREQ (string, "upper")) return RECC_UPPER; | ||
| 2002 | else if (STREQ (string, "unibyte")) return RECC_UNIBYTE; | ||
| 2003 | else if (STREQ (string, "multibyte")) return RECC_MULTIBYTE; | ||
| 2004 | else if (STREQ (string, "digit")) return RECC_DIGIT; | ||
| 2005 | else if (STREQ (string, "xdigit")) return RECC_XDIGIT; | ||
| 2006 | else if (STREQ (string, "cntrl")) return RECC_CNTRL; | ||
| 2007 | else if (STREQ (string, "blank")) return RECC_BLANK; | ||
| 2008 | else return 0; | ||
| 2009 | } | ||
| 2010 | |||
| 2011 | /* True iff CH is in the char class CC. */ | ||
| 2012 | static boolean | ||
| 2013 | re_iswctype (ch, cc) | ||
| 2014 | int ch; | ||
| 2015 | re_wctype_t cc; | ||
| 2016 | { | ||
| 2017 | switch (cc) | ||
| 2018 | { | ||
| 2019 | case RECC_ALNUM: return ISALNUM (ch); | ||
| 2020 | case RECC_ALPHA: return ISALPHA (ch); | ||
| 2021 | case RECC_BLANK: return ISBLANK (ch); | ||
| 2022 | case RECC_CNTRL: return ISCNTRL (ch); | ||
| 2023 | case RECC_DIGIT: return ISDIGIT (ch); | ||
| 2024 | case RECC_GRAPH: return ISGRAPH (ch); | ||
| 2025 | case RECC_LOWER: return ISLOWER (ch); | ||
| 2026 | case RECC_PRINT: return ISPRINT (ch); | ||
| 2027 | case RECC_PUNCT: return ISPUNCT (ch); | ||
| 2028 | case RECC_SPACE: return ISSPACE (ch); | ||
| 2029 | case RECC_UPPER: return ISUPPER (ch); | ||
| 2030 | case RECC_XDIGIT: return ISXDIGIT (ch); | ||
| 2031 | case RECC_ASCII: return IS_REAL_ASCII (ch); | ||
| 2032 | case RECC_NONASCII: return !IS_REAL_ASCII (ch); | ||
| 2033 | case RECC_UNIBYTE: return ISUNIBYTE (ch); | ||
| 2034 | case RECC_MULTIBYTE: return !ISUNIBYTE (ch); | ||
| 2035 | case RECC_WORD: return ISWORD (ch); | ||
| 2036 | case RECC_ERROR: return false; | ||
| 2037 | default: | ||
| 2038 | abort(); | ||
| 2039 | } | ||
| 2040 | } | ||
| 2041 | |||
| 2042 | /* Return a bit-pattern to use in the range-table bits to match multibyte | ||
| 2043 | chars of class CC. */ | ||
| 2044 | static int | ||
| 2045 | re_wctype_to_bit (cc) | ||
| 2046 | re_wctype_t cc; | ||
| 2047 | { | ||
| 2048 | switch (cc) | ||
| 2049 | { | ||
| 2050 | case RECC_NONASCII: case RECC_PRINT: case RECC_GRAPH: | ||
| 2051 | case RECC_MULTIBYTE: return BIT_MULTIBYTE; | ||
| 2052 | case RECC_ALPHA: case RECC_ALNUM: case RECC_WORD: return BIT_WORD; | ||
| 2053 | case RECC_LOWER: return BIT_LOWER; | ||
| 2054 | case RECC_UPPER: return BIT_UPPER; | ||
| 2055 | case RECC_PUNCT: return BIT_PUNCT; | ||
| 2056 | case RECC_SPACE: return BIT_SPACE; | ||
| 2057 | case RECC_ASCII: case RECC_DIGIT: case RECC_XDIGIT: case RECC_CNTRL: | ||
| 2058 | case RECC_BLANK: case RECC_UNIBYTE: case RECC_ERROR: return 0; | ||
| 2059 | default: | ||
| 2060 | abort(); | ||
| 2061 | } | ||
| 2062 | } | ||
| 2063 | #endif | ||
| 2064 | |||
| 2065 | /* Filling in the work area of a range. */ | ||
| 2066 | |||
| 2067 | /* Actually extend the space in WORK_AREA. */ | ||
| 2068 | |||
| 2069 | static void | ||
| 2070 | extend_range_table_work_area (work_area) | ||
| 2071 | struct range_table_work_area *work_area; | ||
| 2072 | { | ||
| 2073 | work_area->allocated += 16 * sizeof (int); | ||
| 2074 | if (work_area->table) | ||
| 2075 | work_area->table | ||
| 2076 | = (int *) realloc (work_area->table, work_area->allocated); | ||
| 2077 | else | ||
| 2078 | work_area->table | ||
| 2079 | = (int *) malloc (work_area->allocated); | ||
| 2080 | } | ||
| 2081 | |||
| 2082 | #ifdef emacs | ||
| 2083 | |||
| 2084 | /* Carefully find the ranges of codes that are equivalent | ||
| 2085 | under case conversion to the range start..end when passed through | ||
| 2086 | TRANSLATE. Handle the case where non-letters can come in between | ||
| 2087 | two upper-case letters (which happens in Latin-1). | ||
| 2088 | Also handle the case of groups of more than 2 case-equivalent chars. | ||
| 2089 | |||
| 2090 | The basic method is to look at consecutive characters and see | ||
| 2091 | if they can form a run that can be handled as one. | ||
| 2092 | |||
| 2093 | Returns -1 if successful, REG_ESPACE if ran out of space. */ | ||
| 2094 | |||
| 2095 | static int | ||
| 2096 | set_image_of_range_1 (work_area, start, end, translate) | ||
| 2097 | RE_TRANSLATE_TYPE translate; | ||
| 2098 | struct range_table_work_area *work_area; | ||
| 2099 | re_wchar_t start, end; | ||
| 2100 | { | ||
| 2101 | /* `one_case' indicates a character, or a run of characters, | ||
| 2102 | each of which is an isolate (no case-equivalents). | ||
| 2103 | This includes all ASCII non-letters. | ||
| 2104 | |||
| 2105 | `two_case' indicates a character, or a run of characters, | ||
| 2106 | each of which has two case-equivalent forms. | ||
| 2107 | This includes all ASCII letters. | ||
| 2108 | |||
| 2109 | `strange' indicates a character that has more than one | ||
| 2110 | case-equivalent. */ | ||
| 2111 | |||
| 2112 | enum case_type {one_case, two_case, strange}; | ||
| 2113 | |||
| 2114 | /* Describe the run that is in progress, | ||
| 2115 | which the next character can try to extend. | ||
| 2116 | If run_type is strange, that means there really is no run. | ||
| 2117 | If run_type is one_case, then run_start...run_end is the run. | ||
| 2118 | If run_type is two_case, then the run is run_start...run_end, | ||
| 2119 | and the case-equivalents end at run_eqv_end. */ | ||
| 2120 | |||
| 2121 | enum case_type run_type = strange; | ||
| 2122 | int run_start, run_end, run_eqv_end; | ||
| 2123 | |||
| 2124 | Lisp_Object eqv_table; | ||
| 2125 | |||
| 2126 | if (!RE_TRANSLATE_P (translate)) | ||
| 2127 | { | ||
| 2128 | EXTEND_RANGE_TABLE (work_area, 2); | ||
| 2129 | work_area->table[work_area->used++] = (start); | ||
| 2130 | work_area->table[work_area->used++] = (end); | ||
| 2131 | return -1; | ||
| 2132 | } | ||
| 2133 | |||
| 2134 | eqv_table = XCHAR_TABLE (translate)->extras[2]; | ||
| 2135 | |||
| 2136 | for (; start <= end; start++) | ||
| 2137 | { | ||
| 2138 | enum case_type this_type; | ||
| 2139 | int eqv = RE_TRANSLATE (eqv_table, start); | ||
| 2140 | int minchar, maxchar; | ||
| 2141 | |||
| 2142 | /* Classify this character */ | ||
| 2143 | if (eqv == start) | ||
| 2144 | this_type = one_case; | ||
| 2145 | else if (RE_TRANSLATE (eqv_table, eqv) == start) | ||
| 2146 | this_type = two_case; | ||
| 2147 | else | ||
| 2148 | this_type = strange; | ||
| 2149 | |||
| 2150 | if (start < eqv) | ||
| 2151 | minchar = start, maxchar = eqv; | ||
| 2152 | else | ||
| 2153 | minchar = eqv, maxchar = start; | ||
| 2154 | |||
| 2155 | /* Can this character extend the run in progress? */ | ||
| 2156 | if (this_type == strange || this_type != run_type | ||
| 2157 | || !(minchar == run_end + 1 | ||
| 2158 | && (run_type == two_case | ||
| 2159 | ? maxchar == run_eqv_end + 1 : 1))) | ||
| 2160 | { | ||
| 2161 | /* No, end the run. | ||
| 2162 | Record each of its equivalent ranges. */ | ||
| 2163 | if (run_type == one_case) | ||
| 2164 | { | ||
| 2165 | EXTEND_RANGE_TABLE (work_area, 2); | ||
| 2166 | work_area->table[work_area->used++] = run_start; | ||
| 2167 | work_area->table[work_area->used++] = run_end; | ||
| 2168 | } | ||
| 2169 | else if (run_type == two_case) | ||
| 2170 | { | ||
| 2171 | EXTEND_RANGE_TABLE (work_area, 4); | ||
| 2172 | work_area->table[work_area->used++] = run_start; | ||
| 2173 | work_area->table[work_area->used++] = run_end; | ||
| 2174 | work_area->table[work_area->used++] | ||
| 2175 | = RE_TRANSLATE (eqv_table, run_start); | ||
| 2176 | work_area->table[work_area->used++] | ||
| 2177 | = RE_TRANSLATE (eqv_table, run_end); | ||
| 2178 | } | ||
| 2179 | run_type = strange; | ||
| 2180 | } | ||
| 2181 | |||
| 2182 | if (this_type == strange) | ||
| 2183 | { | ||
| 2184 | /* For a strange character, add each of its equivalents, one | ||
| 2185 | by one. Don't start a range. */ | ||
| 2186 | do | ||
| 2187 | { | ||
| 2188 | EXTEND_RANGE_TABLE (work_area, 2); | ||
| 2189 | work_area->table[work_area->used++] = eqv; | ||
| 2190 | work_area->table[work_area->used++] = eqv; | ||
| 2191 | eqv = RE_TRANSLATE (eqv_table, eqv); | ||
| 2192 | } | ||
| 2193 | while (eqv != start); | ||
| 2194 | } | ||
| 2195 | |||
| 2196 | /* Add this char to the run, or start a new run. */ | ||
| 2197 | else if (run_type == strange) | ||
| 2198 | { | ||
| 2199 | /* Initialize a new range. */ | ||
| 2200 | run_type = this_type; | ||
| 2201 | run_start = start; | ||
| 2202 | run_end = start; | ||
| 2203 | run_eqv_end = RE_TRANSLATE (eqv_table, run_end); | ||
| 2204 | } | ||
| 2205 | else | ||
| 2206 | { | ||
| 2207 | /* Extend a running range. */ | ||
| 2208 | run_end = minchar; | ||
| 2209 | run_eqv_end = RE_TRANSLATE (eqv_table, run_end); | ||
| 2210 | } | ||
| 2211 | } | ||
| 2212 | |||
| 2213 | /* If a run is still in progress at the end, finish it now | ||
| 2214 | by recording its equivalent ranges. */ | ||
| 2215 | if (run_type == one_case) | ||
| 2216 | { | ||
| 2217 | EXTEND_RANGE_TABLE (work_area, 2); | ||
| 2218 | work_area->table[work_area->used++] = run_start; | ||
| 2219 | work_area->table[work_area->used++] = run_end; | ||
| 2220 | } | ||
| 2221 | else if (run_type == two_case) | ||
| 2222 | { | ||
| 2223 | EXTEND_RANGE_TABLE (work_area, 4); | ||
| 2224 | work_area->table[work_area->used++] = run_start; | ||
| 2225 | work_area->table[work_area->used++] = run_end; | ||
| 2226 | work_area->table[work_area->used++] | ||
| 2227 | = RE_TRANSLATE (eqv_table, run_start); | ||
| 2228 | work_area->table[work_area->used++] | ||
| 2229 | = RE_TRANSLATE (eqv_table, run_end); | ||
| 2230 | } | ||
| 2231 | |||
| 2232 | return -1; | ||
| 2233 | } | ||
| 2234 | |||
| 2235 | #endif /* emacs */ | ||
| 2236 | |||
| 2237 | /* Record the the image of the range start..end when passed through | ||
| 2238 | TRANSLATE. This is not necessarily TRANSLATE(start)..TRANSLATE(end) | ||
| 2239 | and is not even necessarily contiguous. | ||
| 2240 | Normally we approximate it with the smallest contiguous range that contains | ||
| 2241 | all the chars we need. However, for Latin-1 we go to extra effort | ||
| 2242 | to do a better job. | ||
| 2243 | |||
| 2244 | This function is not called for ASCII ranges. | ||
| 2245 | |||
| 2246 | Returns -1 if successful, REG_ESPACE if ran out of space. */ | ||
| 2247 | |||
| 2248 | static int | ||
| 2249 | set_image_of_range (work_area, start, end, translate) | ||
| 2250 | RE_TRANSLATE_TYPE translate; | ||
| 2251 | struct range_table_work_area *work_area; | ||
| 2252 | re_wchar_t start, end; | ||
| 2253 | { | ||
| 2254 | re_wchar_t cmin, cmax; | ||
| 2255 | |||
| 2256 | #ifdef emacs | ||
| 2257 | /* For Latin-1 ranges, use set_image_of_range_1 | ||
| 2258 | to get proper handling of ranges that include letters and nonletters. | ||
| 2259 | For a range that includes the whole of Latin-1, this is not necessary. | ||
| 2260 | For other character sets, we don't bother to get this right. */ | ||
| 2261 | if (RE_TRANSLATE_P (translate) && start < 04400 | ||
| 2262 | && !(start < 04200 && end >= 04377)) | ||
| 2263 | { | ||
| 2264 | int newend; | ||
| 2265 | int tem; | ||
| 2266 | newend = end; | ||
| 2267 | if (newend > 04377) | ||
| 2268 | newend = 04377; | ||
| 2269 | tem = set_image_of_range_1 (work_area, start, newend, translate); | ||
| 2270 | if (tem > 0) | ||
| 2271 | return tem; | ||
| 2272 | |||
| 2273 | start = 04400; | ||
| 2274 | if (end < 04400) | ||
| 2275 | return -1; | ||
| 2276 | } | ||
| 2277 | #endif | ||
| 2278 | |||
| 2279 | EXTEND_RANGE_TABLE (work_area, 2); | ||
| 2280 | work_area->table[work_area->used++] = (start); | ||
| 2281 | work_area->table[work_area->used++] = (end); | ||
| 2282 | |||
| 2283 | cmin = -1, cmax = -1; | ||
| 2284 | |||
| 2285 | if (RE_TRANSLATE_P (translate)) | ||
| 2286 | { | ||
| 2287 | int ch; | ||
| 2288 | |||
| 2289 | for (ch = start; ch <= end; ch++) | ||
| 2290 | { | ||
| 2291 | re_wchar_t c = TRANSLATE (ch); | ||
| 2292 | if (! (start <= c && c <= end)) | ||
| 2293 | { | ||
| 2294 | if (cmin == -1) | ||
| 2295 | cmin = c, cmax = c; | ||
| 2296 | else | ||
| 2297 | { | ||
| 2298 | cmin = MIN (cmin, c); | ||
| 2299 | cmax = MAX (cmax, c); | ||
| 2300 | } | ||
| 2301 | } | ||
| 2302 | } | ||
| 2303 | |||
| 2304 | if (cmin != -1) | ||
| 2305 | { | ||
| 2306 | EXTEND_RANGE_TABLE (work_area, 2); | ||
| 2307 | work_area->table[work_area->used++] = (cmin); | ||
| 2308 | work_area->table[work_area->used++] = (cmax); | ||
| 2309 | } | ||
| 2310 | } | ||
| 2311 | |||
| 2312 | return -1; | ||
| 2313 | } | ||
| 2314 | |||
| 2315 | #ifndef MATCH_MAY_ALLOCATE | ||
| 2316 | |||
| 2317 | /* If we cannot allocate large objects within re_match_2_internal, | ||
| 2318 | we make the fail stack and register vectors global. | ||
| 2319 | The fail stack, we grow to the maximum size when a regexp | ||
| 2320 | is compiled. | ||
| 2321 | The register vectors, we adjust in size each time we | ||
| 2322 | compile a regexp, according to the number of registers it needs. */ | ||
| 2323 | |||
| 2324 | static fail_stack_type fail_stack; | ||
| 2325 | |||
| 2326 | /* Size with which the following vectors are currently allocated. | ||
| 2327 | That is so we can make them bigger as needed, | ||
| 2328 | but never make them smaller. */ | ||
| 2329 | static int regs_allocated_size; | ||
| 2330 | |||
| 2331 | static re_char ** regstart, ** regend; | ||
| 2332 | static re_char **best_regstart, **best_regend; | ||
| 2333 | |||
| 2334 | /* Make the register vectors big enough for NUM_REGS registers, | ||
| 2335 | but don't make them smaller. */ | ||
| 2336 | |||
| 2337 | static | ||
| 2338 | regex_grow_registers (num_regs) | ||
| 2339 | int num_regs; | ||
| 2340 | { | ||
| 2341 | if (num_regs > regs_allocated_size) | ||
| 2342 | { | ||
| 2343 | RETALLOC_IF (regstart, num_regs, re_char *); | ||
| 2344 | RETALLOC_IF (regend, num_regs, re_char *); | ||
| 2345 | RETALLOC_IF (best_regstart, num_regs, re_char *); | ||
| 2346 | RETALLOC_IF (best_regend, num_regs, re_char *); | ||
| 2347 | |||
| 2348 | regs_allocated_size = num_regs; | ||
| 2349 | } | ||
| 2350 | } | ||
| 2351 | |||
| 2352 | #endif /* not MATCH_MAY_ALLOCATE */ | ||
| 2353 | |||
| 2354 | static boolean group_in_compile_stack _RE_ARGS ((compile_stack_type | ||
| 2355 | compile_stack, | ||
| 2356 | regnum_t regnum)); | ||
| 2357 | |||
| 2358 | /* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX. | ||
| 2359 | Returns one of error codes defined in `regex.h', or zero for success. | ||
| 2360 | |||
| 2361 | Assumes the `allocated' (and perhaps `buffer') and `translate' | ||
| 2362 | fields are set in BUFP on entry. | ||
| 2363 | |||
| 2364 | If it succeeds, results are put in BUFP (if it returns an error, the | ||
| 2365 | contents of BUFP are undefined): | ||
| 2366 | `buffer' is the compiled pattern; | ||
| 2367 | `syntax' is set to SYNTAX; | ||
| 2368 | `used' is set to the length of the compiled pattern; | ||
| 2369 | `fastmap_accurate' is zero; | ||
| 2370 | `re_nsub' is the number of subexpressions in PATTERN; | ||
| 2371 | `not_bol' and `not_eol' are zero; | ||
| 2372 | |||
| 2373 | The `fastmap' field is neither examined nor set. */ | ||
| 2374 | |||
| 2375 | /* Insert the `jump' from the end of last alternative to "here". | ||
| 2376 | The space for the jump has already been allocated. */ | ||
| 2377 | #define FIXUP_ALT_JUMP() \ | ||
| 2378 | do { \ | ||
| 2379 | if (fixup_alt_jump) \ | ||
| 2380 | STORE_JUMP (jump, fixup_alt_jump, b); \ | ||
| 2381 | } while (0) | ||
| 2382 | |||
| 2383 | |||
| 2384 | /* Return, freeing storage we allocated. */ | ||
| 2385 | #define FREE_STACK_RETURN(value) \ | ||
| 2386 | do { \ | ||
| 2387 | FREE_RANGE_TABLE_WORK_AREA (range_table_work); \ | ||
| 2388 | free (compile_stack.stack); \ | ||
| 2389 | return value; \ | ||
| 2390 | } while (0) | ||
| 2391 | |||
| 2392 | static reg_errcode_t | ||
| 2393 | regex_compile (pattern, size, syntax, bufp) | ||
| 2394 | re_char *pattern; | ||
| 2395 | size_t size; | ||
| 2396 | reg_syntax_t syntax; | ||
| 2397 | struct re_pattern_buffer *bufp; | ||
| 2398 | { | ||
| 2399 | /* We fetch characters from PATTERN here. */ | ||
| 2400 | register re_wchar_t c, c1; | ||
| 2401 | |||
| 2402 | /* A random temporary spot in PATTERN. */ | ||
| 2403 | re_char *p1; | ||
| 2404 | |||
| 2405 | /* Points to the end of the buffer, where we should append. */ | ||
| 2406 | register unsigned char *b; | ||
| 2407 | |||
| 2408 | /* Keeps track of unclosed groups. */ | ||
| 2409 | compile_stack_type compile_stack; | ||
| 2410 | |||
| 2411 | /* Points to the current (ending) position in the pattern. */ | ||
| 2412 | #ifdef AIX | ||
| 2413 | /* `const' makes AIX compiler fail. */ | ||
| 2414 | unsigned char *p = pattern; | ||
| 2415 | #else | ||
| 2416 | re_char *p = pattern; | ||
| 2417 | #endif | ||
| 2418 | re_char *pend = pattern + size; | ||
| 2419 | |||
| 2420 | /* How to translate the characters in the pattern. */ | ||
| 2421 | RE_TRANSLATE_TYPE translate = bufp->translate; | ||
| 2422 | |||
| 2423 | /* Address of the count-byte of the most recently inserted `exactn' | ||
| 2424 | command. This makes it possible to tell if a new exact-match | ||
| 2425 | character can be added to that command or if the character requires | ||
| 2426 | a new `exactn' command. */ | ||
| 2427 | unsigned char *pending_exact = 0; | ||
| 2428 | |||
| 2429 | /* Address of start of the most recently finished expression. | ||
| 2430 | This tells, e.g., postfix * where to find the start of its | ||
| 2431 | operand. Reset at the beginning of groups and alternatives. */ | ||
| 2432 | unsigned char *laststart = 0; | ||
| 2433 | |||
| 2434 | /* Address of beginning of regexp, or inside of last group. */ | ||
| 2435 | unsigned char *begalt; | ||
| 2436 | |||
| 2437 | /* Place in the uncompiled pattern (i.e., the {) to | ||
| 2438 | which to go back if the interval is invalid. */ | ||
| 2439 | re_char *beg_interval; | ||
| 2440 | |||
| 2441 | /* Address of the place where a forward jump should go to the end of | ||
| 2442 | the containing expression. Each alternative of an `or' -- except the | ||
| 2443 | last -- ends with a forward jump of this sort. */ | ||
| 2444 | unsigned char *fixup_alt_jump = 0; | ||
| 2445 | |||
| 2446 | /* Counts open-groups as they are encountered. Remembered for the | ||
| 2447 | matching close-group on the compile stack, so the same register | ||
| 2448 | number is put in the stop_memory as the start_memory. */ | ||
| 2449 | regnum_t regnum = 0; | ||
| 2450 | |||
| 2451 | /* Work area for range table of charset. */ | ||
| 2452 | struct range_table_work_area range_table_work; | ||
| 2453 | |||
| 2454 | /* If the object matched can contain multibyte characters. */ | ||
| 2455 | const boolean multibyte = RE_MULTIBYTE_P (bufp); | ||
| 2456 | |||
| 2457 | #ifdef DEBUG | ||
| 2458 | debug++; | ||
| 2459 | DEBUG_PRINT1 ("\nCompiling pattern: "); | ||
| 2460 | if (debug > 0) | ||
| 2461 | { | ||
| 2462 | unsigned debug_count; | ||
| 2463 | |||
| 2464 | for (debug_count = 0; debug_count < size; debug_count++) | ||
| 2465 | putchar (pattern[debug_count]); | ||
| 2466 | putchar ('\n'); | ||
| 2467 | } | ||
| 2468 | #endif /* DEBUG */ | ||
| 2469 | |||
| 2470 | /* Initialize the compile stack. */ | ||
| 2471 | compile_stack.stack = TALLOC (INIT_COMPILE_STACK_SIZE, compile_stack_elt_t); | ||
| 2472 | if (compile_stack.stack == NULL) | ||
| 2473 | return REG_ESPACE; | ||
| 2474 | |||
| 2475 | compile_stack.size = INIT_COMPILE_STACK_SIZE; | ||
| 2476 | compile_stack.avail = 0; | ||
| 2477 | |||
| 2478 | range_table_work.table = 0; | ||
| 2479 | range_table_work.allocated = 0; | ||
| 2480 | |||
| 2481 | /* Initialize the pattern buffer. */ | ||
| 2482 | bufp->syntax = syntax; | ||
| 2483 | bufp->fastmap_accurate = 0; | ||
| 2484 | bufp->not_bol = bufp->not_eol = 0; | ||
| 2485 | |||
| 2486 | /* Set `used' to zero, so that if we return an error, the pattern | ||
| 2487 | printer (for debugging) will think there's no pattern. We reset it | ||
| 2488 | at the end. */ | ||
| 2489 | bufp->used = 0; | ||
| 2490 | |||
| 2491 | /* Always count groups, whether or not bufp->no_sub is set. */ | ||
| 2492 | bufp->re_nsub = 0; | ||
| 2493 | |||
| 2494 | #if !defined emacs && !defined SYNTAX_TABLE | ||
| 2495 | /* Initialize the syntax table. */ | ||
| 2496 | init_syntax_once (); | ||
| 2497 | #endif | ||
| 2498 | |||
| 2499 | if (bufp->allocated == 0) | ||
| 2500 | { | ||
| 2501 | if (bufp->buffer) | ||
| 2502 | { /* If zero allocated, but buffer is non-null, try to realloc | ||
| 2503 | enough space. This loses if buffer's address is bogus, but | ||
| 2504 | that is the user's responsibility. */ | ||
| 2505 | RETALLOC (bufp->buffer, INIT_BUF_SIZE, unsigned char); | ||
| 2506 | } | ||
| 2507 | else | ||
| 2508 | { /* Caller did not allocate a buffer. Do it for them. */ | ||
| 2509 | bufp->buffer = TALLOC (INIT_BUF_SIZE, unsigned char); | ||
| 2510 | } | ||
| 2511 | if (!bufp->buffer) FREE_STACK_RETURN (REG_ESPACE); | ||
| 2512 | |||
| 2513 | bufp->allocated = INIT_BUF_SIZE; | ||
| 2514 | } | ||
| 2515 | |||
| 2516 | begalt = b = bufp->buffer; | ||
| 2517 | |||
| 2518 | /* Loop through the uncompiled pattern until we're at the end. */ | ||
| 2519 | while (p != pend) | ||
| 2520 | { | ||
| 2521 | PATFETCH (c); | ||
| 2522 | |||
| 2523 | switch (c) | ||
| 2524 | { | ||
| 2525 | case '^': | ||
| 2526 | { | ||
| 2527 | if ( /* If at start of pattern, it's an operator. */ | ||
| 2528 | p == pattern + 1 | ||
| 2529 | /* If context independent, it's an operator. */ | ||
| 2530 | || syntax & RE_CONTEXT_INDEP_ANCHORS | ||
| 2531 | /* Otherwise, depends on what's come before. */ | ||
| 2532 | || at_begline_loc_p (pattern, p, syntax)) | ||
| 2533 | BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? begbuf : begline); | ||
| 2534 | else | ||
| 2535 | goto normal_char; | ||
| 2536 | } | ||
| 2537 | break; | ||
| 2538 | |||
| 2539 | |||
| 2540 | case '$': | ||
| 2541 | { | ||
| 2542 | if ( /* If at end of pattern, it's an operator. */ | ||
| 2543 | p == pend | ||
| 2544 | /* If context independent, it's an operator. */ | ||
| 2545 | || syntax & RE_CONTEXT_INDEP_ANCHORS | ||
| 2546 | /* Otherwise, depends on what's next. */ | ||
| 2547 | || at_endline_loc_p (p, pend, syntax)) | ||
| 2548 | BUF_PUSH ((syntax & RE_NO_NEWLINE_ANCHOR) ? endbuf : endline); | ||
| 2549 | else | ||
| 2550 | goto normal_char; | ||
| 2551 | } | ||
| 2552 | break; | ||
| 2553 | |||
| 2554 | |||
| 2555 | case '+': | ||
| 2556 | case '?': | ||
| 2557 | if ((syntax & RE_BK_PLUS_QM) | ||
| 2558 | || (syntax & RE_LIMITED_OPS)) | ||
| 2559 | goto normal_char; | ||
| 2560 | handle_plus: | ||
| 2561 | case '*': | ||
| 2562 | /* If there is no previous pattern... */ | ||
| 2563 | if (!laststart) | ||
| 2564 | { | ||
| 2565 | if (syntax & RE_CONTEXT_INVALID_OPS) | ||
| 2566 | FREE_STACK_RETURN (REG_BADRPT); | ||
| 2567 | else if (!(syntax & RE_CONTEXT_INDEP_OPS)) | ||
| 2568 | goto normal_char; | ||
| 2569 | } | ||
| 2570 | |||
| 2571 | { | ||
| 2572 | /* 1 means zero (many) matches is allowed. */ | ||
| 2573 | boolean zero_times_ok = 0, many_times_ok = 0; | ||
| 2574 | boolean greedy = 1; | ||
| 2575 | |||
| 2576 | /* If there is a sequence of repetition chars, collapse it | ||
| 2577 | down to just one (the right one). We can't combine | ||
| 2578 | interval operators with these because of, e.g., `a{2}*', | ||
| 2579 | which should only match an even number of `a's. */ | ||
| 2580 | |||
| 2581 | for (;;) | ||
| 2582 | { | ||
| 2583 | if ((syntax & RE_FRUGAL) | ||
| 2584 | && c == '?' && (zero_times_ok || many_times_ok)) | ||
| 2585 | greedy = 0; | ||
| 2586 | else | ||
| 2587 | { | ||
| 2588 | zero_times_ok |= c != '+'; | ||
| 2589 | many_times_ok |= c != '?'; | ||
| 2590 | } | ||
| 2591 | |||
| 2592 | if (p == pend) | ||
| 2593 | break; | ||
| 2594 | else if (*p == '*' | ||
| 2595 | || (!(syntax & RE_BK_PLUS_QM) | ||
| 2596 | && (*p == '+' || *p == '?'))) | ||
| 2597 | ; | ||
| 2598 | else if (syntax & RE_BK_PLUS_QM && *p == '\\') | ||
| 2599 | { | ||
| 2600 | if (p+1 == pend) | ||
| 2601 | FREE_STACK_RETURN (REG_EESCAPE); | ||
| 2602 | if (p[1] == '+' || p[1] == '?') | ||
| 2603 | PATFETCH (c); /* Gobble up the backslash. */ | ||
| 2604 | else | ||
| 2605 | break; | ||
| 2606 | } | ||
| 2607 | else | ||
| 2608 | break; | ||
| 2609 | /* If we get here, we found another repeat character. */ | ||
| 2610 | PATFETCH (c); | ||
| 2611 | } | ||
| 2612 | |||
| 2613 | /* Star, etc. applied to an empty pattern is equivalent | ||
| 2614 | to an empty pattern. */ | ||
| 2615 | if (!laststart || laststart == b) | ||
| 2616 | break; | ||
| 2617 | |||
| 2618 | /* Now we know whether or not zero matches is allowed | ||
| 2619 | and also whether or not two or more matches is allowed. */ | ||
| 2620 | if (greedy) | ||
| 2621 | { | ||
| 2622 | if (many_times_ok) | ||
| 2623 | { | ||
| 2624 | boolean simple = skip_one_char (laststart) == b; | ||
| 2625 | unsigned int startoffset = 0; | ||
| 2626 | re_opcode_t ofj = | ||
| 2627 | /* Check if the loop can match the empty string. */ | ||
| 2628 | (simple || !analyse_first (laststart, b, NULL, 0)) | ||
| 2629 | ? on_failure_jump : on_failure_jump_loop; | ||
| 2630 | assert (skip_one_char (laststart) <= b); | ||
| 2631 | |||
| 2632 | if (!zero_times_ok && simple) | ||
| 2633 | { /* Since simple * loops can be made faster by using | ||
| 2634 | on_failure_keep_string_jump, we turn simple P+ | ||
| 2635 | into PP* if P is simple. */ | ||
| 2636 | unsigned char *p1, *p2; | ||
| 2637 | startoffset = b - laststart; | ||
| 2638 | GET_BUFFER_SPACE (startoffset); | ||
| 2639 | p1 = b; p2 = laststart; | ||
| 2640 | while (p2 < p1) | ||
| 2641 | *b++ = *p2++; | ||
| 2642 | zero_times_ok = 1; | ||
| 2643 | } | ||
| 2644 | |||
| 2645 | GET_BUFFER_SPACE (6); | ||
| 2646 | if (!zero_times_ok) | ||
| 2647 | /* A + loop. */ | ||
| 2648 | STORE_JUMP (ofj, b, b + 6); | ||
| 2649 | else | ||
| 2650 | /* Simple * loops can use on_failure_keep_string_jump | ||
| 2651 | depending on what follows. But since we don't know | ||
| 2652 | that yet, we leave the decision up to | ||
| 2653 | on_failure_jump_smart. */ | ||
| 2654 | INSERT_JUMP (simple ? on_failure_jump_smart : ofj, | ||
| 2655 | laststart + startoffset, b + 6); | ||
| 2656 | b += 3; | ||
| 2657 | STORE_JUMP (jump, b, laststart + startoffset); | ||
| 2658 | b += 3; | ||
| 2659 | } | ||
| 2660 | else | ||
| 2661 | { | ||
| 2662 | /* A simple ? pattern. */ | ||
| 2663 | assert (zero_times_ok); | ||
| 2664 | GET_BUFFER_SPACE (3); | ||
| 2665 | INSERT_JUMP (on_failure_jump, laststart, b + 3); | ||
| 2666 | b += 3; | ||
| 2667 | } | ||
| 2668 | } | ||
| 2669 | else /* not greedy */ | ||
| 2670 | { /* I wish the greedy and non-greedy cases could be merged. */ | ||
| 2671 | |||
| 2672 | GET_BUFFER_SPACE (7); /* We might use less. */ | ||
| 2673 | if (many_times_ok) | ||
| 2674 | { | ||
| 2675 | boolean emptyp = analyse_first (laststart, b, NULL, 0); | ||
| 2676 | |||
| 2677 | /* The non-greedy multiple match looks like | ||
| 2678 | a repeat..until: we only need a conditional jump | ||
| 2679 | at the end of the loop. */ | ||
| 2680 | if (emptyp) BUF_PUSH (no_op); | ||
| 2681 | STORE_JUMP (emptyp ? on_failure_jump_nastyloop | ||
| 2682 | : on_failure_jump, b, laststart); | ||
| 2683 | b += 3; | ||
| 2684 | if (zero_times_ok) | ||
| 2685 | { | ||
| 2686 | /* The repeat...until naturally matches one or more. | ||
| 2687 | To also match zero times, we need to first jump to | ||
| 2688 | the end of the loop (its conditional jump). */ | ||
| 2689 | INSERT_JUMP (jump, laststart, b); | ||
| 2690 | b += 3; | ||
| 2691 | } | ||
| 2692 | } | ||
| 2693 | else | ||
| 2694 | { | ||
| 2695 | /* non-greedy a?? */ | ||
| 2696 | INSERT_JUMP (jump, laststart, b + 3); | ||
| 2697 | b += 3; | ||
| 2698 | INSERT_JUMP (on_failure_jump, laststart, laststart + 6); | ||
| 2699 | b += 3; | ||
| 2700 | } | ||
| 2701 | } | ||
| 2702 | } | ||
| 2703 | pending_exact = 0; | ||
| 2704 | break; | ||
| 2705 | |||
| 2706 | |||
| 2707 | case '.': | ||
| 2708 | laststart = b; | ||
| 2709 | BUF_PUSH (anychar); | ||
| 2710 | break; | ||
| 2711 | |||
| 2712 | |||
| 2713 | case '[': | ||
| 2714 | { | ||
| 2715 | CLEAR_RANGE_TABLE_WORK_USED (range_table_work); | ||
| 2716 | |||
| 2717 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | ||
| 2718 | |||
| 2719 | /* Ensure that we have enough space to push a charset: the | ||
| 2720 | opcode, the length count, and the bitset; 34 bytes in all. */ | ||
| 2721 | GET_BUFFER_SPACE (34); | ||
| 2722 | |||
| 2723 | laststart = b; | ||
| 2724 | |||
| 2725 | /* We test `*p == '^' twice, instead of using an if | ||
| 2726 | statement, so we only need one BUF_PUSH. */ | ||
| 2727 | BUF_PUSH (*p == '^' ? charset_not : charset); | ||
| 2728 | if (*p == '^') | ||
| 2729 | p++; | ||
| 2730 | |||
| 2731 | /* Remember the first position in the bracket expression. */ | ||
| 2732 | p1 = p; | ||
| 2733 | |||
| 2734 | /* Push the number of bytes in the bitmap. */ | ||
| 2735 | BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH); | ||
| 2736 | |||
| 2737 | /* Clear the whole map. */ | ||
| 2738 | bzero (b, (1 << BYTEWIDTH) / BYTEWIDTH); | ||
| 2739 | |||
| 2740 | /* charset_not matches newline according to a syntax bit. */ | ||
| 2741 | if ((re_opcode_t) b[-2] == charset_not | ||
| 2742 | && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) | ||
| 2743 | SET_LIST_BIT ('\n'); | ||
| 2744 | |||
| 2745 | /* Read in characters and ranges, setting map bits. */ | ||
| 2746 | for (;;) | ||
| 2747 | { | ||
| 2748 | boolean escaped_char = false; | ||
| 2749 | const unsigned char *p2 = p; | ||
| 2750 | |||
| 2751 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | ||
| 2752 | |||
| 2753 | /* Don't translate yet. The range TRANSLATE(X..Y) cannot | ||
| 2754 | always be determined from TRANSLATE(X) and TRANSLATE(Y) | ||
| 2755 | So the translation is done later in a loop. Example: | ||
| 2756 | (let ((case-fold-search t)) (string-match "[A-_]" "A")) */ | ||
| 2757 | PATFETCH (c); | ||
| 2758 | |||
| 2759 | /* \ might escape characters inside [...] and [^...]. */ | ||
| 2760 | if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\') | ||
| 2761 | { | ||
| 2762 | if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | ||
| 2763 | |||
| 2764 | PATFETCH (c); | ||
| 2765 | escaped_char = true; | ||
| 2766 | } | ||
| 2767 | else | ||
| 2768 | { | ||
| 2769 | /* Could be the end of the bracket expression. If it's | ||
| 2770 | not (i.e., when the bracket expression is `[]' so | ||
| 2771 | far), the ']' character bit gets set way below. */ | ||
| 2772 | if (c == ']' && p2 != p1) | ||
| 2773 | break; | ||
| 2774 | } | ||
| 2775 | |||
| 2776 | /* What should we do for the character which is | ||
| 2777 | greater than 0x7F, but not BASE_LEADING_CODE_P? | ||
| 2778 | XXX */ | ||
| 2779 | |||
| 2780 | /* See if we're at the beginning of a possible character | ||
| 2781 | class. */ | ||
| 2782 | |||
| 2783 | if (!escaped_char && | ||
| 2784 | syntax & RE_CHAR_CLASSES && c == '[' && *p == ':') | ||
| 2785 | { | ||
| 2786 | /* Leave room for the null. */ | ||
| 2787 | unsigned char str[CHAR_CLASS_MAX_LENGTH + 1]; | ||
| 2788 | const unsigned char *class_beg; | ||
| 2789 | |||
| 2790 | PATFETCH (c); | ||
| 2791 | c1 = 0; | ||
| 2792 | class_beg = p; | ||
| 2793 | |||
| 2794 | /* If pattern is `[[:'. */ | ||
| 2795 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | ||
| 2796 | |||
| 2797 | for (;;) | ||
| 2798 | { | ||
| 2799 | PATFETCH (c); | ||
| 2800 | if ((c == ':' && *p == ']') || p == pend) | ||
| 2801 | break; | ||
| 2802 | if (c1 < CHAR_CLASS_MAX_LENGTH) | ||
| 2803 | str[c1++] = c; | ||
| 2804 | else | ||
| 2805 | /* This is in any case an invalid class name. */ | ||
| 2806 | str[0] = '\0'; | ||
| 2807 | } | ||
| 2808 | str[c1] = '\0'; | ||
| 2809 | |||
| 2810 | /* If isn't a word bracketed by `[:' and `:]': | ||
| 2811 | undo the ending character, the letters, and | ||
| 2812 | leave the leading `:' and `[' (but set bits for | ||
| 2813 | them). */ | ||
| 2814 | if (c == ':' && *p == ']') | ||
| 2815 | { | ||
| 2816 | re_wchar_t ch; | ||
| 2817 | re_wctype_t cc; | ||
| 2818 | |||
| 2819 | cc = re_wctype (str); | ||
| 2820 | |||
| 2821 | if (cc == 0) | ||
| 2822 | FREE_STACK_RETURN (REG_ECTYPE); | ||
| 2823 | |||
| 2824 | /* Throw away the ] at the end of the character | ||
| 2825 | class. */ | ||
| 2826 | PATFETCH (c); | ||
| 2827 | |||
| 2828 | if (p == pend) FREE_STACK_RETURN (REG_EBRACK); | ||
| 2829 | |||
| 2830 | /* Most character classes in a multibyte match | ||
| 2831 | just set a flag. Exceptions are is_blank, | ||
| 2832 | is_digit, is_cntrl, and is_xdigit, since | ||
| 2833 | they can only match ASCII characters. We | ||
| 2834 | don't need to handle them for multibyte. | ||
| 2835 | They are distinguished by a negative wctype. */ | ||
| 2836 | |||
| 2837 | if (multibyte) | ||
| 2838 | SET_RANGE_TABLE_WORK_AREA_BIT (range_table_work, | ||
| 2839 | re_wctype_to_bit (cc)); | ||
| 2840 | |||
| 2841 | for (ch = 0; ch < 1 << BYTEWIDTH; ++ch) | ||
| 2842 | { | ||
| 2843 | int translated = TRANSLATE (ch); | ||
| 2844 | if (re_iswctype (btowc (ch), cc)) | ||
| 2845 | SET_LIST_BIT (translated); | ||
| 2846 | } | ||
| 2847 | |||
| 2848 | /* Repeat the loop. */ | ||
| 2849 | continue; | ||
| 2850 | } | ||
| 2851 | else | ||
| 2852 | { | ||
| 2853 | /* Go back to right after the "[:". */ | ||
| 2854 | p = class_beg; | ||
| 2855 | SET_LIST_BIT ('['); | ||
| 2856 | |||
| 2857 | /* Because the `:' may starts the range, we | ||
| 2858 | can't simply set bit and repeat the loop. | ||
| 2859 | Instead, just set it to C and handle below. */ | ||
| 2860 | c = ':'; | ||
| 2861 | } | ||
| 2862 | } | ||
| 2863 | |||
| 2864 | if (p < pend && p[0] == '-' && p[1] != ']') | ||
| 2865 | { | ||
| 2866 | |||
| 2867 | /* Discard the `-'. */ | ||
| 2868 | PATFETCH (c1); | ||
| 2869 | |||
| 2870 | /* Fetch the character which ends the range. */ | ||
| 2871 | PATFETCH (c1); | ||
| 2872 | |||
| 2873 | if (SINGLE_BYTE_CHAR_P (c)) | ||
| 2874 | { | ||
| 2875 | if (! SINGLE_BYTE_CHAR_P (c1)) | ||
| 2876 | { | ||
| 2877 | /* Handle a range starting with a | ||
| 2878 | character of less than 256, and ending | ||
| 2879 | with a character of not less than 256. | ||
| 2880 | Split that into two ranges, the low one | ||
| 2881 | ending at 0377, and the high one | ||
| 2882 | starting at the smallest character in | ||
| 2883 | the charset of C1 and ending at C1. */ | ||
| 2884 | int charset = CHAR_CHARSET (c1); | ||
| 2885 | re_wchar_t c2 = MAKE_CHAR (charset, 0, 0); | ||
| 2886 | |||
| 2887 | SET_RANGE_TABLE_WORK_AREA (range_table_work, | ||
| 2888 | c2, c1); | ||
| 2889 | c1 = 0377; | ||
| 2890 | } | ||
| 2891 | } | ||
| 2892 | else if (!SAME_CHARSET_P (c, c1)) | ||
| 2893 | FREE_STACK_RETURN (REG_ERANGE); | ||
| 2894 | } | ||
| 2895 | else | ||
| 2896 | /* Range from C to C. */ | ||
| 2897 | c1 = c; | ||
| 2898 | |||
| 2899 | /* Set the range ... */ | ||
| 2900 | if (SINGLE_BYTE_CHAR_P (c)) | ||
| 2901 | /* ... into bitmap. */ | ||
| 2902 | { | ||
| 2903 | re_wchar_t this_char; | ||
| 2904 | re_wchar_t range_start = c, range_end = c1; | ||
| 2905 | |||
| 2906 | /* If the start is after the end, the range is empty. */ | ||
| 2907 | if (range_start > range_end) | ||
| 2908 | { | ||
| 2909 | if (syntax & RE_NO_EMPTY_RANGES) | ||
| 2910 | FREE_STACK_RETURN (REG_ERANGE); | ||
| 2911 | /* Else, repeat the loop. */ | ||
| 2912 | } | ||
| 2913 | else | ||
| 2914 | { | ||
| 2915 | for (this_char = range_start; this_char <= range_end; | ||
| 2916 | this_char++) | ||
| 2917 | SET_LIST_BIT (TRANSLATE (this_char)); | ||
| 2918 | } | ||
| 2919 | } | ||
| 2920 | else | ||
| 2921 | /* ... into range table. */ | ||
| 2922 | SET_RANGE_TABLE_WORK_AREA (range_table_work, c, c1); | ||
| 2923 | } | ||
| 2924 | |||
| 2925 | /* Discard any (non)matching list bytes that are all 0 at the | ||
| 2926 | end of the map. Decrease the map-length byte too. */ | ||
| 2927 | while ((int) b[-1] > 0 && b[b[-1] - 1] == 0) | ||
| 2928 | b[-1]--; | ||
| 2929 | b += b[-1]; | ||
| 2930 | |||
| 2931 | /* Build real range table from work area. */ | ||
| 2932 | if (RANGE_TABLE_WORK_USED (range_table_work) | ||
| 2933 | || RANGE_TABLE_WORK_BITS (range_table_work)) | ||
| 2934 | { | ||
| 2935 | int i; | ||
| 2936 | int used = RANGE_TABLE_WORK_USED (range_table_work); | ||
| 2937 | |||
| 2938 | /* Allocate space for COUNT + RANGE_TABLE. Needs two | ||
| 2939 | bytes for flags, two for COUNT, and three bytes for | ||
| 2940 | each character. */ | ||
| 2941 | GET_BUFFER_SPACE (4 + used * 3); | ||
| 2942 | |||
| 2943 | /* Indicate the existence of range table. */ | ||
| 2944 | laststart[1] |= 0x80; | ||
| 2945 | |||
| 2946 | /* Store the character class flag bits into the range table. | ||
| 2947 | If not in emacs, these flag bits are always 0. */ | ||
| 2948 | *b++ = RANGE_TABLE_WORK_BITS (range_table_work) & 0xff; | ||
| 2949 | *b++ = RANGE_TABLE_WORK_BITS (range_table_work) >> 8; | ||
| 2950 | |||
| 2951 | STORE_NUMBER_AND_INCR (b, used / 2); | ||
| 2952 | for (i = 0; i < used; i++) | ||
| 2953 | STORE_CHARACTER_AND_INCR | ||
| 2954 | (b, RANGE_TABLE_WORK_ELT (range_table_work, i)); | ||
| 2955 | } | ||
| 2956 | } | ||
| 2957 | break; | ||
| 2958 | |||
| 2959 | |||
| 2960 | case '(': | ||
| 2961 | if (syntax & RE_NO_BK_PARENS) | ||
| 2962 | goto handle_open; | ||
| 2963 | else | ||
| 2964 | goto normal_char; | ||
| 2965 | |||
| 2966 | |||
| 2967 | case ')': | ||
| 2968 | if (syntax & RE_NO_BK_PARENS) | ||
| 2969 | goto handle_close; | ||
| 2970 | else | ||
| 2971 | goto normal_char; | ||
| 2972 | |||
| 2973 | |||
| 2974 | case '\n': | ||
| 2975 | if (syntax & RE_NEWLINE_ALT) | ||
| 2976 | goto handle_alt; | ||
| 2977 | else | ||
| 2978 | goto normal_char; | ||
| 2979 | |||
| 2980 | |||
| 2981 | case '|': | ||
| 2982 | if (syntax & RE_NO_BK_VBAR) | ||
| 2983 | goto handle_alt; | ||
| 2984 | else | ||
| 2985 | goto normal_char; | ||
| 2986 | |||
| 2987 | |||
| 2988 | case '{': | ||
| 2989 | if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES) | ||
| 2990 | goto handle_interval; | ||
| 2991 | else | ||
| 2992 | goto normal_char; | ||
| 2993 | |||
| 2994 | |||
| 2995 | case '\\': | ||
| 2996 | if (p == pend) FREE_STACK_RETURN (REG_EESCAPE); | ||
| 2997 | |||
| 2998 | /* Do not translate the character after the \, so that we can | ||
| 2999 | distinguish, e.g., \B from \b, even if we normally would | ||
| 3000 | translate, e.g., B to b. */ | ||
| 3001 | PATFETCH (c); | ||
| 3002 | |||
| 3003 | switch (c) | ||
| 3004 | { | ||
| 3005 | case '(': | ||
| 3006 | if (syntax & RE_NO_BK_PARENS) | ||
| 3007 | goto normal_backslash; | ||
| 3008 | |||
| 3009 | handle_open: | ||
| 3010 | { | ||
| 3011 | int shy = 0; | ||
| 3012 | if (p+1 < pend) | ||
| 3013 | { | ||
| 3014 | /* Look for a special (?...) construct */ | ||
| 3015 | if ((syntax & RE_SHY_GROUPS) && *p == '?') | ||
| 3016 | { | ||
| 3017 | PATFETCH (c); /* Gobble up the '?'. */ | ||
| 3018 | PATFETCH (c); | ||
| 3019 | switch (c) | ||
| 3020 | { | ||
| 3021 | case ':': shy = 1; break; | ||
| 3022 | default: | ||
| 3023 | /* Only (?:...) is supported right now. */ | ||
| 3024 | FREE_STACK_RETURN (REG_BADPAT); | ||
| 3025 | } | ||
| 3026 | } | ||
| 3027 | } | ||
| 3028 | |||
| 3029 | if (!shy) | ||
| 3030 | { | ||
| 3031 | bufp->re_nsub++; | ||
| 3032 | regnum++; | ||
| 3033 | } | ||
| 3034 | |||
| 3035 | if (COMPILE_STACK_FULL) | ||
| 3036 | { | ||
| 3037 | RETALLOC (compile_stack.stack, compile_stack.size << 1, | ||
| 3038 | compile_stack_elt_t); | ||
| 3039 | if (compile_stack.stack == NULL) return REG_ESPACE; | ||
| 3040 | |||
| 3041 | compile_stack.size <<= 1; | ||
| 3042 | } | ||
| 3043 | |||
| 3044 | /* These are the values to restore when we hit end of this | ||
| 3045 | group. They are all relative offsets, so that if the | ||
| 3046 | whole pattern moves because of realloc, they will still | ||
| 3047 | be valid. */ | ||
| 3048 | COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer; | ||
| 3049 | COMPILE_STACK_TOP.fixup_alt_jump | ||
| 3050 | = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0; | ||
| 3051 | COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer; | ||
| 3052 | COMPILE_STACK_TOP.regnum = shy ? -regnum : regnum; | ||
| 3053 | |||
| 3054 | /* Do not push a | ||
| 3055 | start_memory for groups beyond the last one we can | ||
| 3056 | represent in the compiled pattern. */ | ||
| 3057 | if (regnum <= MAX_REGNUM && !shy) | ||
| 3058 | BUF_PUSH_2 (start_memory, regnum); | ||
| 3059 | |||
| 3060 | compile_stack.avail++; | ||
| 3061 | |||
| 3062 | fixup_alt_jump = 0; | ||
| 3063 | laststart = 0; | ||
| 3064 | begalt = b; | ||
| 3065 | /* If we've reached MAX_REGNUM groups, then this open | ||
| 3066 | won't actually generate any code, so we'll have to | ||
| 3067 | clear pending_exact explicitly. */ | ||
| 3068 | pending_exact = 0; | ||
| 3069 | break; | ||
| 3070 | } | ||
| 3071 | |||
| 3072 | case ')': | ||
| 3073 | if (syntax & RE_NO_BK_PARENS) goto normal_backslash; | ||
| 3074 | |||
| 3075 | if (COMPILE_STACK_EMPTY) | ||
| 3076 | { | ||
| 3077 | if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | ||
| 3078 | goto normal_backslash; | ||
| 3079 | else | ||
| 3080 | FREE_STACK_RETURN (REG_ERPAREN); | ||
| 3081 | } | ||
| 3082 | |||
| 3083 | handle_close: | ||
| 3084 | FIXUP_ALT_JUMP (); | ||
| 3085 | |||
| 3086 | /* See similar code for backslashed left paren above. */ | ||
| 3087 | if (COMPILE_STACK_EMPTY) | ||
| 3088 | { | ||
| 3089 | if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) | ||
| 3090 | goto normal_char; | ||
| 3091 | else | ||
| 3092 | FREE_STACK_RETURN (REG_ERPAREN); | ||
| 3093 | } | ||
| 3094 | |||
| 3095 | /* Since we just checked for an empty stack above, this | ||
| 3096 | ``can't happen''. */ | ||
| 3097 | assert (compile_stack.avail != 0); | ||
| 3098 | { | ||
| 3099 | /* We don't just want to restore into `regnum', because | ||
| 3100 | later groups should continue to be numbered higher, | ||
| 3101 | as in `(ab)c(de)' -- the second group is #2. */ | ||
| 3102 | regnum_t this_group_regnum; | ||
| 3103 | |||
| 3104 | compile_stack.avail--; | ||
| 3105 | begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset; | ||
| 3106 | fixup_alt_jump | ||
| 3107 | = COMPILE_STACK_TOP.fixup_alt_jump | ||
| 3108 | ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1 | ||
| 3109 | : 0; | ||
| 3110 | laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset; | ||
| 3111 | this_group_regnum = COMPILE_STACK_TOP.regnum; | ||
| 3112 | /* If we've reached MAX_REGNUM groups, then this open | ||
| 3113 | won't actually generate any code, so we'll have to | ||
| 3114 | clear pending_exact explicitly. */ | ||
| 3115 | pending_exact = 0; | ||
| 3116 | |||
| 3117 | /* We're at the end of the group, so now we know how many | ||
| 3118 | groups were inside this one. */ | ||
| 3119 | if (this_group_regnum <= MAX_REGNUM && this_group_regnum > 0) | ||
| 3120 | BUF_PUSH_2 (stop_memory, this_group_regnum); | ||
| 3121 | } | ||
| 3122 | break; | ||
| 3123 | |||
| 3124 | |||
| 3125 | case '|': /* `\|'. */ | ||
| 3126 | if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR) | ||
| 3127 | goto normal_backslash; | ||
| 3128 | handle_alt: | ||
| 3129 | if (syntax & RE_LIMITED_OPS) | ||
| 3130 | goto normal_char; | ||
| 3131 | |||
| 3132 | /* Insert before the previous alternative a jump which | ||
| 3133 | jumps to this alternative if the former fails. */ | ||
| 3134 | GET_BUFFER_SPACE (3); | ||
| 3135 | INSERT_JUMP (on_failure_jump, begalt, b + 6); | ||
| 3136 | pending_exact = 0; | ||
| 3137 | b += 3; | ||
| 3138 | |||
| 3139 | /* The alternative before this one has a jump after it | ||
| 3140 | which gets executed if it gets matched. Adjust that | ||
| 3141 | jump so it will jump to this alternative's analogous | ||
| 3142 | jump (put in below, which in turn will jump to the next | ||
| 3143 | (if any) alternative's such jump, etc.). The last such | ||
| 3144 | jump jumps to the correct final destination. A picture: | ||
| 3145 | _____ _____ | ||
| 3146 | | | | | | ||
| 3147 | | v | v | ||
| 3148 | a | b | c | ||
| 3149 | |||
| 3150 | If we are at `b', then fixup_alt_jump right now points to a | ||
| 3151 | three-byte space after `a'. We'll put in the jump, set | ||
| 3152 | fixup_alt_jump to right after `b', and leave behind three | ||
| 3153 | bytes which we'll fill in when we get to after `c'. */ | ||
| 3154 | |||
| 3155 | FIXUP_ALT_JUMP (); | ||
| 3156 | |||
| 3157 | /* Mark and leave space for a jump after this alternative, | ||
| 3158 | to be filled in later either by next alternative or | ||
| 3159 | when know we're at the end of a series of alternatives. */ | ||
| 3160 | fixup_alt_jump = b; | ||
| 3161 | GET_BUFFER_SPACE (3); | ||
| 3162 | b += 3; | ||
| 3163 | |||
| 3164 | laststart = 0; | ||
| 3165 | begalt = b; | ||
| 3166 | break; | ||
| 3167 | |||
| 3168 | |||
| 3169 | case '{': | ||
| 3170 | /* If \{ is a literal. */ | ||
| 3171 | if (!(syntax & RE_INTERVALS) | ||
| 3172 | /* If we're at `\{' and it's not the open-interval | ||
| 3173 | operator. */ | ||
| 3174 | || (syntax & RE_NO_BK_BRACES)) | ||
| 3175 | goto normal_backslash; | ||
| 3176 | |||
| 3177 | handle_interval: | ||
| 3178 | { | ||
| 3179 | /* If got here, then the syntax allows intervals. */ | ||
| 3180 | |||
| 3181 | /* At least (most) this many matches must be made. */ | ||
| 3182 | int lower_bound = 0, upper_bound = -1; | ||
| 3183 | |||
| 3184 | beg_interval = p; | ||
| 3185 | |||
| 3186 | if (p == pend) | ||
| 3187 | FREE_STACK_RETURN (REG_EBRACE); | ||
| 3188 | |||
| 3189 | GET_UNSIGNED_NUMBER (lower_bound); | ||
| 3190 | |||
| 3191 | if (c == ',') | ||
| 3192 | GET_UNSIGNED_NUMBER (upper_bound); | ||
| 3193 | else | ||
| 3194 | /* Interval such as `{1}' => match exactly once. */ | ||
| 3195 | upper_bound = lower_bound; | ||
| 3196 | |||
| 3197 | if (lower_bound < 0 || upper_bound > RE_DUP_MAX | ||
| 3198 | || (upper_bound >= 0 && lower_bound > upper_bound)) | ||
| 3199 | FREE_STACK_RETURN (REG_BADBR); | ||
| 3200 | |||
| 3201 | if (!(syntax & RE_NO_BK_BRACES)) | ||
| 3202 | { | ||
| 3203 | if (c != '\\') | ||
| 3204 | FREE_STACK_RETURN (REG_BADBR); | ||
| 3205 | |||
| 3206 | PATFETCH (c); | ||
| 3207 | } | ||
| 3208 | |||
| 3209 | if (c != '}') | ||
| 3210 | FREE_STACK_RETURN (REG_BADBR); | ||
| 3211 | |||
| 3212 | /* We just parsed a valid interval. */ | ||
| 3213 | |||
| 3214 | /* If it's invalid to have no preceding re. */ | ||
| 3215 | if (!laststart) | ||
| 3216 | { | ||
| 3217 | if (syntax & RE_CONTEXT_INVALID_OPS) | ||
| 3218 | FREE_STACK_RETURN (REG_BADRPT); | ||
| 3219 | else if (syntax & RE_CONTEXT_INDEP_OPS) | ||
| 3220 | laststart = b; | ||
| 3221 | else | ||
| 3222 | goto unfetch_interval; | ||
| 3223 | } | ||
| 3224 | |||
| 3225 | if (upper_bound == 0) | ||
| 3226 | /* If the upper bound is zero, just drop the sub pattern | ||
| 3227 | altogether. */ | ||
| 3228 | b = laststart; | ||
| 3229 | else if (lower_bound == 1 && upper_bound == 1) | ||
| 3230 | /* Just match it once: nothing to do here. */ | ||
| 3231 | ; | ||
| 3232 | |||
| 3233 | /* Otherwise, we have a nontrivial interval. When | ||
| 3234 | we're all done, the pattern will look like: | ||
| 3235 | set_number_at <jump count> <upper bound> | ||
| 3236 | set_number_at <succeed_n count> <lower bound> | ||
| 3237 | succeed_n <after jump addr> <succeed_n count> | ||
| 3238 | <body of loop> | ||
| 3239 | jump_n <succeed_n addr> <jump count> | ||
| 3240 | (The upper bound and `jump_n' are omitted if | ||
| 3241 | `upper_bound' is 1, though.) */ | ||
| 3242 | else | ||
| 3243 | { /* If the upper bound is > 1, we need to insert | ||
| 3244 | more at the end of the loop. */ | ||
| 3245 | unsigned int nbytes = (upper_bound < 0 ? 3 | ||
| 3246 | : upper_bound > 1 ? 5 : 0); | ||
| 3247 | unsigned int startoffset = 0; | ||
| 3248 | |||
| 3249 | GET_BUFFER_SPACE (20); /* We might use less. */ | ||
| 3250 | |||
| 3251 | if (lower_bound == 0) | ||
| 3252 | { | ||
| 3253 | /* A succeed_n that starts with 0 is really a | ||
| 3254 | a simple on_failure_jump_loop. */ | ||
| 3255 | INSERT_JUMP (on_failure_jump_loop, laststart, | ||
| 3256 | b + 3 + nbytes); | ||
| 3257 | b += 3; | ||
| 3258 | } | ||
| 3259 | else | ||
| 3260 | { | ||
| 3261 | /* Initialize lower bound of the `succeed_n', even | ||
| 3262 | though it will be set during matching by its | ||
| 3263 | attendant `set_number_at' (inserted next), | ||
| 3264 | because `re_compile_fastmap' needs to know. | ||
| 3265 | Jump to the `jump_n' we might insert below. */ | ||
| 3266 | INSERT_JUMP2 (succeed_n, laststart, | ||
| 3267 | b + 5 + nbytes, | ||
| 3268 | lower_bound); | ||
| 3269 | b += 5; | ||
| 3270 | |||
| 3271 | /* Code to initialize the lower bound. Insert | ||
| 3272 | before the `succeed_n'. The `5' is the last two | ||
| 3273 | bytes of this `set_number_at', plus 3 bytes of | ||
| 3274 | the following `succeed_n'. */ | ||
| 3275 | insert_op2 (set_number_at, laststart, 5, lower_bound, b); | ||
| 3276 | b += 5; | ||
| 3277 | startoffset += 5; | ||
| 3278 | } | ||
| 3279 | |||
| 3280 | if (upper_bound < 0) | ||
| 3281 | { | ||
| 3282 | /* A negative upper bound stands for infinity, | ||
| 3283 | in which case it degenerates to a plain jump. */ | ||
| 3284 | STORE_JUMP (jump, b, laststart + startoffset); | ||
| 3285 | b += 3; | ||
| 3286 | } | ||
| 3287 | else if (upper_bound > 1) | ||
| 3288 | { /* More than one repetition is allowed, so | ||
| 3289 | append a backward jump to the `succeed_n' | ||
| 3290 | that starts this interval. | ||
| 3291 | |||
| 3292 | When we've reached this during matching, | ||
| 3293 | we'll have matched the interval once, so | ||
| 3294 | jump back only `upper_bound - 1' times. */ | ||
| 3295 | STORE_JUMP2 (jump_n, b, laststart + startoffset, | ||
| 3296 | upper_bound - 1); | ||
| 3297 | b += 5; | ||
| 3298 | |||
| 3299 | /* The location we want to set is the second | ||
| 3300 | parameter of the `jump_n'; that is `b-2' as | ||
| 3301 | an absolute address. `laststart' will be | ||
| 3302 | the `set_number_at' we're about to insert; | ||
| 3303 | `laststart+3' the number to set, the source | ||
| 3304 | for the relative address. But we are | ||
| 3305 | inserting into the middle of the pattern -- | ||
| 3306 | so everything is getting moved up by 5. | ||
| 3307 | Conclusion: (b - 2) - (laststart + 3) + 5, | ||
| 3308 | i.e., b - laststart. | ||
| 3309 | |||
| 3310 | We insert this at the beginning of the loop | ||
| 3311 | so that if we fail during matching, we'll | ||
| 3312 | reinitialize the bounds. */ | ||
| 3313 | insert_op2 (set_number_at, laststart, b - laststart, | ||
| 3314 | upper_bound - 1, b); | ||
| 3315 | b += 5; | ||
| 3316 | } | ||
| 3317 | } | ||
| 3318 | pending_exact = 0; | ||
| 3319 | beg_interval = NULL; | ||
| 3320 | } | ||
| 3321 | break; | ||
| 3322 | |||
| 3323 | unfetch_interval: | ||
| 3324 | /* If an invalid interval, match the characters as literals. */ | ||
| 3325 | assert (beg_interval); | ||
| 3326 | p = beg_interval; | ||
| 3327 | beg_interval = NULL; | ||
| 3328 | |||
| 3329 | /* normal_char and normal_backslash need `c'. */ | ||
| 3330 | c = '{'; | ||
| 3331 | |||
| 3332 | if (!(syntax & RE_NO_BK_BRACES)) | ||
| 3333 | { | ||
| 3334 | assert (p > pattern && p[-1] == '\\'); | ||
| 3335 | goto normal_backslash; | ||
| 3336 | } | ||
| 3337 | else | ||
| 3338 | goto normal_char; | ||
| 3339 | |||
| 3340 | #ifdef emacs | ||
| 3341 | /* There is no way to specify the before_dot and after_dot | ||
| 3342 | operators. rms says this is ok. --karl */ | ||
| 3343 | case '=': | ||
| 3344 | BUF_PUSH (at_dot); | ||
| 3345 | break; | ||
| 3346 | |||
| 3347 | case 's': | ||
| 3348 | laststart = b; | ||
| 3349 | PATFETCH (c); | ||
| 3350 | BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]); | ||
| 3351 | break; | ||
| 3352 | |||
| 3353 | case 'S': | ||
| 3354 | laststart = b; | ||
| 3355 | PATFETCH (c); | ||
| 3356 | BUF_PUSH_2 (notsyntaxspec, syntax_spec_code[c]); | ||
| 3357 | break; | ||
| 3358 | |||
| 3359 | case 'c': | ||
| 3360 | laststart = b; | ||
| 3361 | PATFETCH (c); | ||
| 3362 | BUF_PUSH_2 (categoryspec, c); | ||
| 3363 | break; | ||
| 3364 | |||
| 3365 | case 'C': | ||
| 3366 | laststart = b; | ||
| 3367 | PATFETCH (c); | ||
| 3368 | BUF_PUSH_2 (notcategoryspec, c); | ||
| 3369 | break; | ||
| 3370 | #endif /* emacs */ | ||
| 3371 | |||
| 3372 | |||
| 3373 | case 'w': | ||
| 3374 | if (syntax & RE_NO_GNU_OPS) | ||
| 3375 | goto normal_char; | ||
| 3376 | laststart = b; | ||
| 3377 | BUF_PUSH_2 (syntaxspec, Sword); | ||
| 3378 | break; | ||
| 3379 | |||
| 3380 | |||
| 3381 | case 'W': | ||
| 3382 | if (syntax & RE_NO_GNU_OPS) | ||
| 3383 | goto normal_char; | ||
| 3384 | laststart = b; | ||
| 3385 | BUF_PUSH_2 (notsyntaxspec, Sword); | ||
| 3386 | break; | ||
| 3387 | |||
| 3388 | |||
| 3389 | case '<': | ||
| 3390 | if (syntax & RE_NO_GNU_OPS) | ||
| 3391 | goto normal_char; | ||
| 3392 | BUF_PUSH (wordbeg); | ||
| 3393 | break; | ||
| 3394 | |||
| 3395 | case '>': | ||
| 3396 | if (syntax & RE_NO_GNU_OPS) | ||
| 3397 | goto normal_char; | ||
| 3398 | BUF_PUSH (wordend); | ||
| 3399 | break; | ||
| 3400 | |||
| 3401 | case 'b': | ||
| 3402 | if (syntax & RE_NO_GNU_OPS) | ||
| 3403 | goto normal_char; | ||
| 3404 | BUF_PUSH (wordbound); | ||
| 3405 | break; | ||
| 3406 | |||
| 3407 | case 'B': | ||
| 3408 | if (syntax & RE_NO_GNU_OPS) | ||
| 3409 | goto normal_char; | ||
| 3410 | BUF_PUSH (notwordbound); | ||
| 3411 | break; | ||
| 3412 | |||
| 3413 | case '`': | ||
| 3414 | if (syntax & RE_NO_GNU_OPS) | ||
| 3415 | goto normal_char; | ||
| 3416 | BUF_PUSH (begbuf); | ||
| 3417 | break; | ||
| 3418 | |||
| 3419 | case '\'': | ||
| 3420 | if (syntax & RE_NO_GNU_OPS) | ||
| 3421 | goto normal_char; | ||
| 3422 | BUF_PUSH (endbuf); | ||
| 3423 | break; | ||
| 3424 | |||
| 3425 | case '1': case '2': case '3': case '4': case '5': | ||
| 3426 | case '6': case '7': case '8': case '9': | ||
| 3427 | { | ||
| 3428 | regnum_t reg; | ||
| 3429 | |||
| 3430 | if (syntax & RE_NO_BK_REFS) | ||
| 3431 | goto normal_backslash; | ||
| 3432 | |||
| 3433 | reg = c - '0'; | ||
| 3434 | |||
| 3435 | /* Can't back reference to a subexpression before its end. */ | ||
| 3436 | if (reg > regnum || group_in_compile_stack (compile_stack, reg)) | ||
| 3437 | FREE_STACK_RETURN (REG_ESUBREG); | ||
| 3438 | |||
| 3439 | laststart = b; | ||
| 3440 | BUF_PUSH_2 (duplicate, reg); | ||
| 3441 | } | ||
| 3442 | break; | ||
| 3443 | |||
| 3444 | |||
| 3445 | case '+': | ||
| 3446 | case '?': | ||
| 3447 | if (syntax & RE_BK_PLUS_QM) | ||
| 3448 | goto handle_plus; | ||
| 3449 | else | ||
| 3450 | goto normal_backslash; | ||
| 3451 | |||
| 3452 | default: | ||
| 3453 | normal_backslash: | ||
| 3454 | /* You might think it would be useful for \ to mean | ||
| 3455 | not to translate; but if we don't translate it | ||
| 3456 | it will never match anything. */ | ||
| 3457 | goto normal_char; | ||
| 3458 | } | ||
| 3459 | break; | ||
| 3460 | |||
| 3461 | |||
| 3462 | default: | ||
| 3463 | /* Expects the character in `c'. */ | ||
| 3464 | normal_char: | ||
| 3465 | /* If no exactn currently being built. */ | ||
| 3466 | if (!pending_exact | ||
| 3467 | |||
| 3468 | /* If last exactn not at current position. */ | ||
| 3469 | || pending_exact + *pending_exact + 1 != b | ||
| 3470 | |||
| 3471 | /* We have only one byte following the exactn for the count. */ | ||
| 3472 | || *pending_exact >= (1 << BYTEWIDTH) - MAX_MULTIBYTE_LENGTH | ||
| 3473 | |||
| 3474 | /* If followed by a repetition operator. */ | ||
| 3475 | || (p != pend && (*p == '*' || *p == '^')) | ||
| 3476 | || ((syntax & RE_BK_PLUS_QM) | ||
| 3477 | ? p + 1 < pend && *p == '\\' && (p[1] == '+' || p[1] == '?') | ||
| 3478 | : p != pend && (*p == '+' || *p == '?')) | ||
| 3479 | || ((syntax & RE_INTERVALS) | ||
| 3480 | && ((syntax & RE_NO_BK_BRACES) | ||
| 3481 | ? p != pend && *p == '{' | ||
| 3482 | : p + 1 < pend && p[0] == '\\' && p[1] == '{'))) | ||
| 3483 | { | ||
| 3484 | /* Start building a new exactn. */ | ||
| 3485 | |||
| 3486 | laststart = b; | ||
| 3487 | |||
| 3488 | BUF_PUSH_2 (exactn, 0); | ||
| 3489 | pending_exact = b - 1; | ||
| 3490 | } | ||
| 3491 | |||
| 3492 | GET_BUFFER_SPACE (MAX_MULTIBYTE_LENGTH); | ||
| 3493 | { | ||
| 3494 | int len; | ||
| 3495 | |||
| 3496 | c = TRANSLATE (c); | ||
| 3497 | if (multibyte) | ||
| 3498 | len = CHAR_STRING (c, b); | ||
| 3499 | else | ||
| 3500 | *b = c, len = 1; | ||
| 3501 | b += len; | ||
| 3502 | (*pending_exact) += len; | ||
| 3503 | } | ||
| 3504 | |||
| 3505 | break; | ||
| 3506 | } /* switch (c) */ | ||
| 3507 | } /* while p != pend */ | ||
| 3508 | |||
| 3509 | |||
| 3510 | /* Through the pattern now. */ | ||
| 3511 | |||
| 3512 | FIXUP_ALT_JUMP (); | ||
| 3513 | |||
| 3514 | if (!COMPILE_STACK_EMPTY) | ||
| 3515 | FREE_STACK_RETURN (REG_EPAREN); | ||
| 3516 | |||
| 3517 | /* If we don't want backtracking, force success | ||
| 3518 | the first time we reach the end of the compiled pattern. */ | ||
| 3519 | if (syntax & RE_NO_POSIX_BACKTRACKING) | ||
| 3520 | BUF_PUSH (succeed); | ||
| 3521 | |||
| 3522 | /* We have succeeded; set the length of the buffer. */ | ||
| 3523 | bufp->used = b - bufp->buffer; | ||
| 3524 | |||
| 3525 | #ifdef DEBUG | ||
| 3526 | if (debug > 0) | ||
| 3527 | { | ||
| 3528 | re_compile_fastmap (bufp); | ||
| 3529 | DEBUG_PRINT1 ("\nCompiled pattern: \n"); | ||
| 3530 | print_compiled_pattern (bufp); | ||
| 3531 | } | ||
| 3532 | debug--; | ||
| 3533 | #endif /* DEBUG */ | ||
| 3534 | |||
| 3535 | #ifndef MATCH_MAY_ALLOCATE | ||
| 3536 | /* Initialize the failure stack to the largest possible stack. This | ||
| 3537 | isn't necessary unless we're trying to avoid calling alloca in | ||
| 3538 | the search and match routines. */ | ||
| 3539 | { | ||
| 3540 | int num_regs = bufp->re_nsub + 1; | ||
| 3541 | |||
| 3542 | if (fail_stack.size < re_max_failures * TYPICAL_FAILURE_SIZE) | ||
| 3543 | { | ||
| 3544 | fail_stack.size = re_max_failures * TYPICAL_FAILURE_SIZE; | ||
| 3545 | |||
| 3546 | if (! fail_stack.stack) | ||
| 3547 | fail_stack.stack | ||
| 3548 | = (fail_stack_elt_t *) malloc (fail_stack.size | ||
| 3549 | * sizeof (fail_stack_elt_t)); | ||
| 3550 | else | ||
| 3551 | fail_stack.stack | ||
| 3552 | = (fail_stack_elt_t *) realloc (fail_stack.stack, | ||
| 3553 | (fail_stack.size | ||
| 3554 | * sizeof (fail_stack_elt_t))); | ||
| 3555 | } | ||
| 3556 | |||
| 3557 | regex_grow_registers (num_regs); | ||
| 3558 | } | ||
| 3559 | #endif /* not MATCH_MAY_ALLOCATE */ | ||
| 3560 | |||
| 3561 | FREE_STACK_RETURN (REG_NOERROR); | ||
| 3562 | } /* regex_compile */ | ||
| 3563 | |||
| 3564 | /* Subroutines for `regex_compile'. */ | ||
| 3565 | |||
| 3566 | /* Store OP at LOC followed by two-byte integer parameter ARG. */ | ||
| 3567 | |||
| 3568 | static void | ||
| 3569 | store_op1 (op, loc, arg) | ||
| 3570 | re_opcode_t op; | ||
| 3571 | unsigned char *loc; | ||
| 3572 | int arg; | ||
| 3573 | { | ||
| 3574 | *loc = (unsigned char) op; | ||
| 3575 | STORE_NUMBER (loc + 1, arg); | ||
| 3576 | } | ||
| 3577 | |||
| 3578 | |||
| 3579 | /* Like `store_op1', but for two two-byte parameters ARG1 and ARG2. */ | ||
| 3580 | |||
| 3581 | static void | ||
| 3582 | store_op2 (op, loc, arg1, arg2) | ||
| 3583 | re_opcode_t op; | ||
| 3584 | unsigned char *loc; | ||
| 3585 | int arg1, arg2; | ||
| 3586 | { | ||
| 3587 | *loc = (unsigned char) op; | ||
| 3588 | STORE_NUMBER (loc + 1, arg1); | ||
| 3589 | STORE_NUMBER (loc + 3, arg2); | ||
| 3590 | } | ||
| 3591 | |||
| 3592 | |||
| 3593 | /* Copy the bytes from LOC to END to open up three bytes of space at LOC | ||
| 3594 | for OP followed by two-byte integer parameter ARG. */ | ||
| 3595 | |||
| 3596 | static void | ||
| 3597 | insert_op1 (op, loc, arg, end) | ||
| 3598 | re_opcode_t op; | ||
| 3599 | unsigned char *loc; | ||
| 3600 | int arg; | ||
| 3601 | unsigned char *end; | ||
| 3602 | { | ||
| 3603 | register unsigned char *pfrom = end; | ||
| 3604 | register unsigned char *pto = end + 3; | ||
| 3605 | |||
| 3606 | while (pfrom != loc) | ||
| 3607 | *--pto = *--pfrom; | ||
| 3608 | |||
| 3609 | store_op1 (op, loc, arg); | ||
| 3610 | } | ||
| 3611 | |||
| 3612 | |||
| 3613 | /* Like `insert_op1', but for two two-byte parameters ARG1 and ARG2. */ | ||
| 3614 | |||
| 3615 | static void | ||
| 3616 | insert_op2 (op, loc, arg1, arg2, end) | ||
| 3617 | re_opcode_t op; | ||
| 3618 | unsigned char *loc; | ||
| 3619 | int arg1, arg2; | ||
| 3620 | unsigned char *end; | ||
| 3621 | { | ||
| 3622 | register unsigned char *pfrom = end; | ||
| 3623 | register unsigned char *pto = end + 5; | ||
| 3624 | |||
| 3625 | while (pfrom != loc) | ||
| 3626 | *--pto = *--pfrom; | ||
| 3627 | |||
| 3628 | store_op2 (op, loc, arg1, arg2); | ||
| 3629 | } | ||
| 3630 | |||
| 3631 | |||
| 3632 | /* P points to just after a ^ in PATTERN. Return true if that ^ comes | ||
| 3633 | after an alternative or a begin-subexpression. We assume there is at | ||
| 3634 | least one character before the ^. */ | ||
| 3635 | |||
| 3636 | static boolean | ||
| 3637 | at_begline_loc_p (pattern, p, syntax) | ||
| 3638 | re_char *pattern, *p; | ||
| 3639 | reg_syntax_t syntax; | ||
| 3640 | { | ||
| 3641 | re_char *prev = p - 2; | ||
| 3642 | boolean prev_prev_backslash = prev > pattern && prev[-1] == '\\'; | ||
| 3643 | |||
| 3644 | return | ||
| 3645 | /* After a subexpression? */ | ||
| 3646 | (*prev == '(' && (syntax & RE_NO_BK_PARENS || prev_prev_backslash)) | ||
| 3647 | /* After an alternative? */ | ||
| 3648 | || (*prev == '|' && (syntax & RE_NO_BK_VBAR || prev_prev_backslash)) | ||
| 3649 | /* After a shy subexpression? */ | ||
| 3650 | || ((syntax & RE_SHY_GROUPS) && prev - 2 >= pattern | ||
| 3651 | && prev[-1] == '?' && prev[-2] == '(' | ||
| 3652 | && (syntax & RE_NO_BK_PARENS | ||
| 3653 | || (prev - 3 >= pattern && prev[-3] == '\\'))); | ||
| 3654 | } | ||
| 3655 | |||
| 3656 | |||
| 3657 | /* The dual of at_begline_loc_p. This one is for $. We assume there is | ||
| 3658 | at least one character after the $, i.e., `P < PEND'. */ | ||
| 3659 | |||
| 3660 | static boolean | ||
| 3661 | at_endline_loc_p (p, pend, syntax) | ||
| 3662 | re_char *p, *pend; | ||
| 3663 | reg_syntax_t syntax; | ||
| 3664 | { | ||
| 3665 | re_char *next = p; | ||
| 3666 | boolean next_backslash = *next == '\\'; | ||
| 3667 | re_char *next_next = p + 1 < pend ? p + 1 : 0; | ||
| 3668 | |||
| 3669 | return | ||
| 3670 | /* Before a subexpression? */ | ||
| 3671 | (syntax & RE_NO_BK_PARENS ? *next == ')' | ||
| 3672 | : next_backslash && next_next && *next_next == ')') | ||
| 3673 | /* Before an alternative? */ | ||
| 3674 | || (syntax & RE_NO_BK_VBAR ? *next == '|' | ||
| 3675 | : next_backslash && next_next && *next_next == '|'); | ||
| 3676 | } | ||
| 3677 | |||
| 3678 | |||
| 3679 | /* Returns true if REGNUM is in one of COMPILE_STACK's elements and | ||
| 3680 | false if it's not. */ | ||
| 3681 | |||
| 3682 | static boolean | ||
| 3683 | group_in_compile_stack (compile_stack, regnum) | ||
| 3684 | compile_stack_type compile_stack; | ||
| 3685 | regnum_t regnum; | ||
| 3686 | { | ||
| 3687 | int this_element; | ||
| 3688 | |||
| 3689 | for (this_element = compile_stack.avail - 1; | ||
| 3690 | this_element >= 0; | ||
| 3691 | this_element--) | ||
| 3692 | if (compile_stack.stack[this_element].regnum == regnum) | ||
| 3693 | return true; | ||
| 3694 | |||
| 3695 | return false; | ||
| 3696 | } | ||
| 3697 | |||
| 3698 | /* analyse_first. | ||
| 3699 | If fastmap is non-NULL, go through the pattern and fill fastmap | ||
| 3700 | with all the possible leading chars. If fastmap is NULL, don't | ||
| 3701 | bother filling it up (obviously) and only return whether the | ||
| 3702 | pattern could potentially match the empty string. | ||
| 3703 | |||
| 3704 | Return 1 if p..pend might match the empty string. | ||
| 3705 | Return 0 if p..pend matches at least one char. | ||
| 3706 | Return -1 if fastmap was not updated accurately. */ | ||
| 3707 | |||
| 3708 | static int | ||
| 3709 | analyse_first (p, pend, fastmap, multibyte) | ||
| 3710 | re_char *p, *pend; | ||
| 3711 | char *fastmap; | ||
| 3712 | const int multibyte; | ||
| 3713 | { | ||
| 3714 | int j, k; | ||
| 3715 | boolean not; | ||
| 3716 | |||
| 3717 | /* If all elements for base leading-codes in fastmap is set, this | ||
| 3718 | flag is set true. */ | ||
| 3719 | boolean match_any_multibyte_characters = false; | ||
| 3720 | |||
| 3721 | assert (p); | ||
| 3722 | |||
| 3723 | /* The loop below works as follows: | ||
| 3724 | - It has a working-list kept in the PATTERN_STACK and which basically | ||
| 3725 | starts by only containing a pointer to the first operation. | ||
| 3726 | - If the opcode we're looking at is a match against some set of | ||
| 3727 | chars, then we add those chars to the fastmap and go on to the | ||
| 3728 | next work element from the worklist (done via `break'). | ||
| 3729 | - If the opcode is a control operator on the other hand, we either | ||
| 3730 | ignore it (if it's meaningless at this point, such as `start_memory') | ||
| 3731 | or execute it (if it's a jump). If the jump has several destinations | ||
| 3732 | (i.e. `on_failure_jump'), then we push the other destination onto the | ||
| 3733 | worklist. | ||
| 3734 | We guarantee termination by ignoring backward jumps (more or less), | ||
| 3735 | so that `p' is monotonically increasing. More to the point, we | ||
| 3736 | never set `p' (or push) anything `<= p1'. */ | ||
| 3737 | |||
| 3738 | while (p < pend) | ||
| 3739 | { | ||
| 3740 | /* `p1' is used as a marker of how far back a `on_failure_jump' | ||
| 3741 | can go without being ignored. It is normally equal to `p' | ||
| 3742 | (which prevents any backward `on_failure_jump') except right | ||
| 3743 | after a plain `jump', to allow patterns such as: | ||
| 3744 | 0: jump 10 | ||
| 3745 | 3..9: <body> | ||
| 3746 | 10: on_failure_jump 3 | ||
| 3747 | as used for the *? operator. */ | ||
| 3748 | re_char *p1 = p; | ||
| 3749 | |||
| 3750 | switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++)) | ||
| 3751 | { | ||
| 3752 | case succeed: | ||
| 3753 | return 1; | ||
| 3754 | continue; | ||
| 3755 | |||
| 3756 | case duplicate: | ||
| 3757 | /* If the first character has to match a backreference, that means | ||
| 3758 | that the group was empty (since it already matched). Since this | ||
| 3759 | is the only case that interests us here, we can assume that the | ||
| 3760 | backreference must match the empty string. */ | ||
| 3761 | p++; | ||
| 3762 | continue; | ||
| 3763 | |||
| 3764 | |||
| 3765 | /* Following are the cases which match a character. These end | ||
| 3766 | with `break'. */ | ||
| 3767 | |||
| 3768 | case exactn: | ||
| 3769 | if (fastmap) | ||
| 3770 | { | ||
| 3771 | int c = RE_STRING_CHAR (p + 1, pend - p); | ||
| 3772 | |||
| 3773 | if (SINGLE_BYTE_CHAR_P (c)) | ||
| 3774 | fastmap[c] = 1; | ||
| 3775 | else | ||
| 3776 | fastmap[p[1]] = 1; | ||
| 3777 | } | ||
| 3778 | break; | ||
| 3779 | |||
| 3780 | |||
| 3781 | case anychar: | ||
| 3782 | /* We could put all the chars except for \n (and maybe \0) | ||
| 3783 | but we don't bother since it is generally not worth it. */ | ||
| 3784 | if (!fastmap) break; | ||
| 3785 | return -1; | ||
| 3786 | |||
| 3787 | |||
| 3788 | case charset_not: | ||
| 3789 | /* Chars beyond end of bitmap are possible matches. | ||
| 3790 | All the single-byte codes can occur in multibyte buffers. | ||
| 3791 | So any that are not listed in the charset | ||
| 3792 | are possible matches, even in multibyte buffers. */ | ||
| 3793 | if (!fastmap) break; | ||
| 3794 | for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH; | ||
| 3795 | j < (1 << BYTEWIDTH); j++) | ||
| 3796 | fastmap[j] = 1; | ||
| 3797 | /* Fallthrough */ | ||
| 3798 | case charset: | ||
| 3799 | if (!fastmap) break; | ||
| 3800 | not = (re_opcode_t) *(p - 1) == charset_not; | ||
| 3801 | for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH - 1, p++; | ||
| 3802 | j >= 0; j--) | ||
| 3803 | if (!!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))) ^ not) | ||
| 3804 | fastmap[j] = 1; | ||
| 3805 | |||
| 3806 | if ((not && multibyte) | ||
| 3807 | /* Any character set can possibly contain a character | ||
| 3808 | which doesn't match the specified set of characters. */ | ||
| 3809 | || (CHARSET_RANGE_TABLE_EXISTS_P (&p[-2]) | ||
| 3810 | && CHARSET_RANGE_TABLE_BITS (&p[-2]) != 0)) | ||
| 3811 | /* If we can match a character class, we can match | ||
| 3812 | any character set. */ | ||
| 3813 | { | ||
| 3814 | set_fastmap_for_multibyte_characters: | ||
| 3815 | if (match_any_multibyte_characters == false) | ||
| 3816 | { | ||
| 3817 | for (j = 0x80; j < 0xA0; j++) /* XXX */ | ||
| 3818 | if (BASE_LEADING_CODE_P (j)) | ||
| 3819 | fastmap[j] = 1; | ||
| 3820 | match_any_multibyte_characters = true; | ||
| 3821 | } | ||
| 3822 | } | ||
| 3823 | |||
| 3824 | else if (!not && CHARSET_RANGE_TABLE_EXISTS_P (&p[-2]) | ||
| 3825 | && match_any_multibyte_characters == false) | ||
| 3826 | { | ||
| 3827 | /* Set fastmap[I] 1 where I is a base leading code of each | ||
| 3828 | multibyte character in the range table. */ | ||
| 3829 | int c, count; | ||
| 3830 | |||
| 3831 | /* Make P points the range table. `+ 2' is to skip flag | ||
| 3832 | bits for a character class. */ | ||
| 3833 | p += CHARSET_BITMAP_SIZE (&p[-2]) + 2; | ||
| 3834 | |||
| 3835 | /* Extract the number of ranges in range table into COUNT. */ | ||
| 3836 | EXTRACT_NUMBER_AND_INCR (count, p); | ||
| 3837 | for (; count > 0; count--, p += 2 * 3) /* XXX */ | ||
| 3838 | { | ||
| 3839 | /* Extract the start of each range. */ | ||
| 3840 | EXTRACT_CHARACTER (c, p); | ||
| 3841 | j = CHAR_CHARSET (c); | ||
| 3842 | fastmap[CHARSET_LEADING_CODE_BASE (j)] = 1; | ||
| 3843 | } | ||
| 3844 | } | ||
| 3845 | break; | ||
| 3846 | |||
| 3847 | case syntaxspec: | ||
| 3848 | case notsyntaxspec: | ||
| 3849 | if (!fastmap) break; | ||
| 3850 | #ifndef emacs | ||
| 3851 | not = (re_opcode_t)p[-1] == notsyntaxspec; | ||
| 3852 | k = *p++; | ||
| 3853 | for (j = 0; j < (1 << BYTEWIDTH); j++) | ||
| 3854 | if ((SYNTAX (j) == (enum syntaxcode) k) ^ not) | ||
| 3855 | fastmap[j] = 1; | ||
| 3856 | break; | ||
| 3857 | #else /* emacs */ | ||
| 3858 | /* This match depends on text properties. These end with | ||
| 3859 | aborting optimizations. */ | ||
| 3860 | return -1; | ||
| 3861 | |||
| 3862 | case categoryspec: | ||
| 3863 | case notcategoryspec: | ||
| 3864 | if (!fastmap) break; | ||
| 3865 | not = (re_opcode_t)p[-1] == notcategoryspec; | ||
| 3866 | k = *p++; | ||
| 3867 | for (j = 0; j < (1 << BYTEWIDTH); j++) | ||
| 3868 | if ((CHAR_HAS_CATEGORY (j, k)) ^ not) | ||
| 3869 | fastmap[j] = 1; | ||
| 3870 | |||
| 3871 | if (multibyte) | ||
| 3872 | /* Any character set can possibly contain a character | ||
| 3873 | whose category is K (or not). */ | ||
| 3874 | goto set_fastmap_for_multibyte_characters; | ||
| 3875 | break; | ||
| 3876 | |||
| 3877 | /* All cases after this match the empty string. These end with | ||
| 3878 | `continue'. */ | ||
| 3879 | |||
| 3880 | case before_dot: | ||
| 3881 | case at_dot: | ||
| 3882 | case after_dot: | ||
| 3883 | #endif /* !emacs */ | ||
| 3884 | case no_op: | ||
| 3885 | case begline: | ||
| 3886 | case endline: | ||
| 3887 | case begbuf: | ||
| 3888 | case endbuf: | ||
| 3889 | case wordbound: | ||
| 3890 | case notwordbound: | ||
| 3891 | case wordbeg: | ||
| 3892 | case wordend: | ||
| 3893 | continue; | ||
| 3894 | |||
| 3895 | |||
| 3896 | case jump: | ||
| 3897 | EXTRACT_NUMBER_AND_INCR (j, p); | ||
| 3898 | if (j < 0) | ||
| 3899 | /* Backward jumps can only go back to code that we've already | ||
| 3900 | visited. `re_compile' should make sure this is true. */ | ||
| 3901 | break; | ||
| 3902 | p += j; | ||
| 3903 | switch (SWITCH_ENUM_CAST ((re_opcode_t) *p)) | ||
| 3904 | { | ||
| 3905 | case on_failure_jump: | ||
| 3906 | case on_failure_keep_string_jump: | ||
| 3907 | case on_failure_jump_loop: | ||
| 3908 | case on_failure_jump_nastyloop: | ||
| 3909 | case on_failure_jump_smart: | ||
| 3910 | p++; | ||
| 3911 | break; | ||
| 3912 | default: | ||
| 3913 | continue; | ||
| 3914 | }; | ||
| 3915 | /* Keep `p1' to allow the `on_failure_jump' we are jumping to | ||
| 3916 | to jump back to "just after here". */ | ||
| 3917 | /* Fallthrough */ | ||
| 3918 | |||
| 3919 | case on_failure_jump: | ||
| 3920 | case on_failure_keep_string_jump: | ||
| 3921 | case on_failure_jump_nastyloop: | ||
| 3922 | case on_failure_jump_loop: | ||
| 3923 | case on_failure_jump_smart: | ||
| 3924 | EXTRACT_NUMBER_AND_INCR (j, p); | ||
| 3925 | if (p + j <= p1) | ||
| 3926 | ; /* Backward jump to be ignored. */ | ||
| 3927 | else | ||
| 3928 | { /* We have to look down both arms. | ||
| 3929 | We first go down the "straight" path so as to minimize | ||
| 3930 | stack usage when going through alternatives. */ | ||
| 3931 | int r = analyse_first (p, pend, fastmap, multibyte); | ||
| 3932 | if (r) return r; | ||
| 3933 | p += j; | ||
| 3934 | } | ||
| 3935 | continue; | ||
| 3936 | |||
| 3937 | |||
| 3938 | case jump_n: | ||
| 3939 | /* This code simply does not properly handle forward jump_n. */ | ||
| 3940 | DEBUG_STATEMENT (EXTRACT_NUMBER (j, p); assert (j < 0)); | ||
| 3941 | p += 4; | ||
| 3942 | /* jump_n can either jump or fall through. The (backward) jump | ||
| 3943 | case has already been handled, so we only need to look at the | ||
| 3944 | fallthrough case. */ | ||
| 3945 | continue; | ||
| 3946 | |||
| 3947 | case succeed_n: | ||
| 3948 | /* If N == 0, it should be an on_failure_jump_loop instead. */ | ||
| 3949 | DEBUG_STATEMENT (EXTRACT_NUMBER (j, p + 2); assert (j > 0)); | ||
| 3950 | p += 4; | ||
| 3951 | /* We only care about one iteration of the loop, so we don't | ||
| 3952 | need to consider the case where this behaves like an | ||
| 3953 | on_failure_jump. */ | ||
| 3954 | continue; | ||
| 3955 | |||
| 3956 | |||
| 3957 | case set_number_at: | ||
| 3958 | p += 4; | ||
| 3959 | continue; | ||
| 3960 | |||
| 3961 | |||
| 3962 | case start_memory: | ||
| 3963 | case stop_memory: | ||
| 3964 | p += 1; | ||
| 3965 | continue; | ||
| 3966 | |||
| 3967 | |||
| 3968 | default: | ||
| 3969 | abort (); /* We have listed all the cases. */ | ||
| 3970 | } /* switch *p++ */ | ||
| 3971 | |||
| 3972 | /* Getting here means we have found the possible starting | ||
| 3973 | characters for one path of the pattern -- and that the empty | ||
| 3974 | string does not match. We need not follow this path further. */ | ||
| 3975 | return 0; | ||
| 3976 | } /* while p */ | ||
| 3977 | |||
| 3978 | /* We reached the end without matching anything. */ | ||
| 3979 | return 1; | ||
| 3980 | |||
| 3981 | } /* analyse_first */ | ||
| 3982 | |||
| 3983 | /* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in | ||
| 3984 | BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible | ||
| 3985 | characters can start a string that matches the pattern. This fastmap | ||
| 3986 | is used by re_search to skip quickly over impossible starting points. | ||
| 3987 | |||
| 3988 | Character codes above (1 << BYTEWIDTH) are not represented in the | ||
| 3989 | fastmap, but the leading codes are represented. Thus, the fastmap | ||
| 3990 | indicates which character sets could start a match. | ||
| 3991 | |||
| 3992 | The caller must supply the address of a (1 << BYTEWIDTH)-byte data | ||
| 3993 | area as BUFP->fastmap. | ||
| 3994 | |||
| 3995 | We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in | ||
| 3996 | the pattern buffer. | ||
| 3997 | |||
| 3998 | Returns 0 if we succeed, -2 if an internal error. */ | ||
| 3999 | |||
| 4000 | int | ||
| 4001 | re_compile_fastmap (bufp) | ||
| 4002 | struct re_pattern_buffer *bufp; | ||
| 4003 | { | ||
| 4004 | char *fastmap = bufp->fastmap; | ||
| 4005 | int analysis; | ||
| 4006 | |||
| 4007 | assert (fastmap && bufp->buffer); | ||
| 4008 | |||
| 4009 | bzero (fastmap, 1 << BYTEWIDTH); /* Assume nothing's valid. */ | ||
| 4010 | bufp->fastmap_accurate = 1; /* It will be when we're done. */ | ||
| 4011 | |||
| 4012 | analysis = analyse_first (bufp->buffer, bufp->buffer + bufp->used, | ||
| 4013 | fastmap, RE_MULTIBYTE_P (bufp)); | ||
| 4014 | bufp->can_be_null = (analysis != 0); | ||
| 4015 | return 0; | ||
| 4016 | } /* re_compile_fastmap */ | ||
| 4017 | |||
| 4018 | /* Set REGS to hold NUM_REGS registers, storing them in STARTS and | ||
| 4019 | ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use | ||
| 4020 | this memory for recording register information. STARTS and ENDS | ||
| 4021 | must be allocated using the malloc library routine, and must each | ||
| 4022 | be at least NUM_REGS * sizeof (regoff_t) bytes long. | ||
| 4023 | |||
| 4024 | If NUM_REGS == 0, then subsequent matches should allocate their own | ||
| 4025 | register data. | ||
| 4026 | |||
| 4027 | Unless this function is called, the first search or match using | ||
| 4028 | PATTERN_BUFFER will allocate its own register data, without | ||
| 4029 | freeing the old data. */ | ||
| 4030 | |||
| 4031 | void | ||
| 4032 | re_set_registers (bufp, regs, num_regs, starts, ends) | ||
| 4033 | struct re_pattern_buffer *bufp; | ||
| 4034 | struct re_registers *regs; | ||
| 4035 | unsigned num_regs; | ||
| 4036 | regoff_t *starts, *ends; | ||
| 4037 | { | ||
| 4038 | if (num_regs) | ||
| 4039 | { | ||
| 4040 | bufp->regs_allocated = REGS_REALLOCATE; | ||
| 4041 | regs->num_regs = num_regs; | ||
| 4042 | regs->start = starts; | ||
| 4043 | regs->end = ends; | ||
| 4044 | } | ||
| 4045 | else | ||
| 4046 | { | ||
| 4047 | bufp->regs_allocated = REGS_UNALLOCATED; | ||
| 4048 | regs->num_regs = 0; | ||
| 4049 | regs->start = regs->end = (regoff_t *) 0; | ||
| 4050 | } | ||
| 4051 | } | ||
| 4052 | WEAK_ALIAS (__re_set_registers, re_set_registers) | ||
| 4053 | |||
| 4054 | /* Searching routines. */ | ||
| 4055 | |||
| 4056 | /* Like re_search_2, below, but only one string is specified, and | ||
| 4057 | doesn't let you say where to stop matching. */ | ||
| 4058 | |||
| 4059 | int | ||
| 4060 | re_search (bufp, string, size, startpos, range, regs) | ||
| 4061 | struct re_pattern_buffer *bufp; | ||
| 4062 | const char *string; | ||
| 4063 | int size, startpos, range; | ||
| 4064 | struct re_registers *regs; | ||
| 4065 | { | ||
| 4066 | return re_search_2 (bufp, NULL, 0, string, size, startpos, range, | ||
| 4067 | regs, size); | ||
| 4068 | } | ||
| 4069 | WEAK_ALIAS (__re_search, re_search) | ||
| 4070 | |||
| 4071 | /* Head address of virtual concatenation of string. */ | ||
| 4072 | #define HEAD_ADDR_VSTRING(P) \ | ||
| 4073 | (((P) >= size1 ? string2 : string1)) | ||
| 4074 | |||
| 4075 | /* End address of virtual concatenation of string. */ | ||
| 4076 | #define STOP_ADDR_VSTRING(P) \ | ||
| 4077 | (((P) >= size1 ? string2 + size2 : string1 + size1)) | ||
| 4078 | |||
| 4079 | /* Address of POS in the concatenation of virtual string. */ | ||
| 4080 | #define POS_ADDR_VSTRING(POS) \ | ||
| 4081 | (((POS) >= size1 ? string2 - size1 : string1) + (POS)) | ||
| 4082 | |||
| 4083 | /* Using the compiled pattern in BUFP->buffer, first tries to match the | ||
| 4084 | virtual concatenation of STRING1 and STRING2, starting first at index | ||
| 4085 | STARTPOS, then at STARTPOS + 1, and so on. | ||
| 4086 | |||
| 4087 | STRING1 and STRING2 have length SIZE1 and SIZE2, respectively. | ||
| 4088 | |||
| 4089 | RANGE is how far to scan while trying to match. RANGE = 0 means try | ||
| 4090 | only at STARTPOS; in general, the last start tried is STARTPOS + | ||
| 4091 | RANGE. | ||
| 4092 | |||
| 4093 | In REGS, return the indices of the virtual concatenation of STRING1 | ||
| 4094 | and STRING2 that matched the entire BUFP->buffer and its contained | ||
| 4095 | subexpressions. | ||
| 4096 | |||
| 4097 | Do not consider matching one past the index STOP in the virtual | ||
| 4098 | concatenation of STRING1 and STRING2. | ||
| 4099 | |||
| 4100 | We return either the position in the strings at which the match was | ||
| 4101 | found, -1 if no match, or -2 if error (such as failure | ||
| 4102 | stack overflow). */ | ||
| 4103 | |||
| 4104 | int | ||
| 4105 | re_search_2 (bufp, str1, size1, str2, size2, startpos, range, regs, stop) | ||
| 4106 | struct re_pattern_buffer *bufp; | ||
| 4107 | const char *str1, *str2; | ||
| 4108 | int size1, size2; | ||
| 4109 | int startpos; | ||
| 4110 | int range; | ||
| 4111 | struct re_registers *regs; | ||
| 4112 | int stop; | ||
| 4113 | { | ||
| 4114 | int val; | ||
| 4115 | re_char *string1 = (re_char*) str1; | ||
| 4116 | re_char *string2 = (re_char*) str2; | ||
| 4117 | register char *fastmap = bufp->fastmap; | ||
| 4118 | register RE_TRANSLATE_TYPE translate = bufp->translate; | ||
| 4119 | int total_size = size1 + size2; | ||
| 4120 | int endpos = startpos + range; | ||
| 4121 | boolean anchored_start; | ||
| 4122 | |||
| 4123 | /* Nonzero if we have to concern multibyte character. */ | ||
| 4124 | const boolean multibyte = RE_MULTIBYTE_P (bufp); | ||
| 4125 | |||
| 4126 | /* Check for out-of-range STARTPOS. */ | ||
| 4127 | if (startpos < 0 || startpos > total_size) | ||
| 4128 | return -1; | ||
| 4129 | |||
| 4130 | /* Fix up RANGE if it might eventually take us outside | ||
| 4131 | the virtual concatenation of STRING1 and STRING2. | ||
| 4132 | Make sure we won't move STARTPOS below 0 or above TOTAL_SIZE. */ | ||
| 4133 | if (endpos < 0) | ||
| 4134 | range = 0 - startpos; | ||
| 4135 | else if (endpos > total_size) | ||
| 4136 | range = total_size - startpos; | ||
| 4137 | |||
| 4138 | /* If the search isn't to be a backwards one, don't waste time in a | ||
| 4139 | search for a pattern anchored at beginning of buffer. */ | ||
| 4140 | if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf && range > 0) | ||
| 4141 | { | ||
| 4142 | if (startpos > 0) | ||
| 4143 | return -1; | ||
| 4144 | else | ||
| 4145 | range = 0; | ||
| 4146 | } | ||
| 4147 | |||
| 4148 | #ifdef emacs | ||
| 4149 | /* In a forward search for something that starts with \=. | ||
| 4150 | don't keep searching past point. */ | ||
| 4151 | if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == at_dot && range > 0) | ||
| 4152 | { | ||
| 4153 | range = PT_BYTE - BEGV_BYTE - startpos; | ||
| 4154 | if (range < 0) | ||
| 4155 | return -1; | ||
| 4156 | } | ||
| 4157 | #endif /* emacs */ | ||
| 4158 | |||
| 4159 | /* Update the fastmap now if not correct already. */ | ||
| 4160 | if (fastmap && !bufp->fastmap_accurate) | ||
| 4161 | re_compile_fastmap (bufp); | ||
| 4162 | |||
| 4163 | /* See whether the pattern is anchored. */ | ||
| 4164 | anchored_start = (bufp->buffer[0] == begline); | ||
| 4165 | |||
| 4166 | #ifdef emacs | ||
| 4167 | gl_state.object = re_match_object; | ||
| 4168 | { | ||
| 4169 | int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (startpos)); | ||
| 4170 | |||
| 4171 | SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1); | ||
| 4172 | } | ||
| 4173 | #endif | ||
| 4174 | |||
| 4175 | /* Loop through the string, looking for a place to start matching. */ | ||
| 4176 | for (;;) | ||
| 4177 | { | ||
| 4178 | /* If the pattern is anchored, | ||
| 4179 | skip quickly past places we cannot match. | ||
| 4180 | We don't bother to treat startpos == 0 specially | ||
| 4181 | because that case doesn't repeat. */ | ||
| 4182 | if (anchored_start && startpos > 0) | ||
| 4183 | { | ||
| 4184 | if (! ((startpos <= size1 ? string1[startpos - 1] | ||
| 4185 | : string2[startpos - size1 - 1]) | ||
| 4186 | == '\n')) | ||
| 4187 | goto advance; | ||
| 4188 | } | ||
| 4189 | |||
| 4190 | /* If a fastmap is supplied, skip quickly over characters that | ||
| 4191 | cannot be the start of a match. If the pattern can match the | ||
| 4192 | null string, however, we don't need to skip characters; we want | ||
| 4193 | the first null string. */ | ||
| 4194 | if (fastmap && startpos < total_size && !bufp->can_be_null) | ||
| 4195 | { | ||
| 4196 | register re_char *d; | ||
| 4197 | register re_wchar_t buf_ch; | ||
| 4198 | |||
| 4199 | d = POS_ADDR_VSTRING (startpos); | ||
| 4200 | |||
| 4201 | if (range > 0) /* Searching forwards. */ | ||
| 4202 | { | ||
| 4203 | register int lim = 0; | ||
| 4204 | int irange = range; | ||
| 4205 | |||
| 4206 | if (startpos < size1 && startpos + range >= size1) | ||
| 4207 | lim = range - (size1 - startpos); | ||
| 4208 | |||
| 4209 | /* Written out as an if-else to avoid testing `translate' | ||
| 4210 | inside the loop. */ | ||
| 4211 | if (RE_TRANSLATE_P (translate)) | ||
| 4212 | { | ||
| 4213 | if (multibyte) | ||
| 4214 | while (range > lim) | ||
| 4215 | { | ||
| 4216 | int buf_charlen; | ||
| 4217 | |||
| 4218 | buf_ch = STRING_CHAR_AND_LENGTH (d, range - lim, | ||
| 4219 | buf_charlen); | ||
| 4220 | |||
| 4221 | buf_ch = RE_TRANSLATE (translate, buf_ch); | ||
| 4222 | if (buf_ch >= 0400 | ||
| 4223 | || fastmap[buf_ch]) | ||
| 4224 | break; | ||
| 4225 | |||
| 4226 | range -= buf_charlen; | ||
| 4227 | d += buf_charlen; | ||
| 4228 | } | ||
| 4229 | else | ||
| 4230 | while (range > lim | ||
| 4231 | && !fastmap[RE_TRANSLATE (translate, *d)]) | ||
| 4232 | { | ||
| 4233 | d++; | ||
| 4234 | range--; | ||
| 4235 | } | ||
| 4236 | } | ||
| 4237 | else | ||
| 4238 | while (range > lim && !fastmap[*d]) | ||
| 4239 | { | ||
| 4240 | d++; | ||
| 4241 | range--; | ||
| 4242 | } | ||
| 4243 | |||
| 4244 | startpos += irange - range; | ||
| 4245 | } | ||
| 4246 | else /* Searching backwards. */ | ||
| 4247 | { | ||
| 4248 | int room = (startpos >= size1 | ||
| 4249 | ? size2 + size1 - startpos | ||
| 4250 | : size1 - startpos); | ||
| 4251 | buf_ch = RE_STRING_CHAR (d, room); | ||
| 4252 | buf_ch = TRANSLATE (buf_ch); | ||
| 4253 | |||
| 4254 | if (! (buf_ch >= 0400 | ||
| 4255 | || fastmap[buf_ch])) | ||
| 4256 | goto advance; | ||
| 4257 | } | ||
| 4258 | } | ||
| 4259 | |||
| 4260 | /* If can't match the null string, and that's all we have left, fail. */ | ||
| 4261 | if (range >= 0 && startpos == total_size && fastmap | ||
| 4262 | && !bufp->can_be_null) | ||
| 4263 | return -1; | ||
| 4264 | |||
| 4265 | val = re_match_2_internal (bufp, string1, size1, string2, size2, | ||
| 4266 | startpos, regs, stop); | ||
| 4267 | #ifndef REGEX_MALLOC | ||
| 4268 | # ifdef C_ALLOCA | ||
| 4269 | alloca (0); | ||
| 4270 | # endif | ||
| 4271 | #endif | ||
| 4272 | |||
| 4273 | if (val >= 0) | ||
| 4274 | return startpos; | ||
| 4275 | |||
| 4276 | if (val == -2) | ||
| 4277 | return -2; | ||
| 4278 | |||
| 4279 | advance: | ||
| 4280 | if (!range) | ||
| 4281 | break; | ||
| 4282 | else if (range > 0) | ||
| 4283 | { | ||
| 4284 | /* Update STARTPOS to the next character boundary. */ | ||
| 4285 | if (multibyte) | ||
| 4286 | { | ||
| 4287 | re_char *p = POS_ADDR_VSTRING (startpos); | ||
| 4288 | re_char *pend = STOP_ADDR_VSTRING (startpos); | ||
| 4289 | int len = MULTIBYTE_FORM_LENGTH (p, pend - p); | ||
| 4290 | |||
| 4291 | range -= len; | ||
| 4292 | if (range < 0) | ||
| 4293 | break; | ||
| 4294 | startpos += len; | ||
| 4295 | } | ||
| 4296 | else | ||
| 4297 | { | ||
| 4298 | range--; | ||
| 4299 | startpos++; | ||
| 4300 | } | ||
| 4301 | } | ||
| 4302 | else | ||
| 4303 | { | ||
| 4304 | range++; | ||
| 4305 | startpos--; | ||
| 4306 | |||
| 4307 | /* Update STARTPOS to the previous character boundary. */ | ||
| 4308 | if (multibyte) | ||
| 4309 | { | ||
| 4310 | re_char *p = POS_ADDR_VSTRING (startpos) + 1; | ||
| 4311 | re_char *p0 = p; | ||
| 4312 | re_char *phead = HEAD_ADDR_VSTRING (startpos); | ||
| 4313 | |||
| 4314 | /* Find the head of multibyte form. */ | ||
| 4315 | PREV_CHAR_BOUNDARY (p, phead); | ||
| 4316 | range += p0 - 1 - p; | ||
| 4317 | if (range > 0) | ||
| 4318 | break; | ||
| 4319 | |||
| 4320 | startpos -= p0 - 1 - p; | ||
| 4321 | } | ||
| 4322 | } | ||
| 4323 | } | ||
| 4324 | return -1; | ||
| 4325 | } /* re_search_2 */ | ||
| 4326 | WEAK_ALIAS (__re_search_2, re_search_2) | ||
| 4327 | |||
| 4328 | /* Declarations and macros for re_match_2. */ | ||
| 4329 | |||
| 4330 | static int bcmp_translate _RE_ARGS((re_char *s1, re_char *s2, | ||
| 4331 | register int len, | ||
| 4332 | RE_TRANSLATE_TYPE translate, | ||
| 4333 | const int multibyte)); | ||
| 4334 | |||
| 4335 | /* This converts PTR, a pointer into one of the search strings `string1' | ||
| 4336 | and `string2' into an offset from the beginning of that string. */ | ||
| 4337 | #define POINTER_TO_OFFSET(ptr) \ | ||
| 4338 | (FIRST_STRING_P (ptr) \ | ||
| 4339 | ? ((regoff_t) ((ptr) - string1)) \ | ||
| 4340 | : ((regoff_t) ((ptr) - string2 + size1))) | ||
| 4341 | |||
| 4342 | /* Call before fetching a character with *d. This switches over to | ||
| 4343 | string2 if necessary. | ||
| 4344 | Check re_match_2_internal for a discussion of why end_match_2 might | ||
| 4345 | not be within string2 (but be equal to end_match_1 instead). */ | ||
| 4346 | #define PREFETCH() \ | ||
| 4347 | while (d == dend) \ | ||
| 4348 | { \ | ||
| 4349 | /* End of string2 => fail. */ \ | ||
| 4350 | if (dend == end_match_2) \ | ||
| 4351 | goto fail; \ | ||
| 4352 | /* End of string1 => advance to string2. */ \ | ||
| 4353 | d = string2; \ | ||
| 4354 | dend = end_match_2; \ | ||
| 4355 | } | ||
| 4356 | |||
| 4357 | /* Call before fetching a char with *d if you already checked other limits. | ||
| 4358 | This is meant for use in lookahead operations like wordend, etc.. | ||
| 4359 | where we might need to look at parts of the string that might be | ||
| 4360 | outside of the LIMITs (i.e past `stop'). */ | ||
| 4361 | #define PREFETCH_NOLIMIT() \ | ||
| 4362 | if (d == end1) \ | ||
| 4363 | { \ | ||
| 4364 | d = string2; \ | ||
| 4365 | dend = end_match_2; \ | ||
| 4366 | } \ | ||
| 4367 | |||
| 4368 | /* Test if at very beginning or at very end of the virtual concatenation | ||
| 4369 | of `string1' and `string2'. If only one string, it's `string2'. */ | ||
| 4370 | #define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2) | ||
| 4371 | #define AT_STRINGS_END(d) ((d) == end2) | ||
| 4372 | |||
| 4373 | |||
| 4374 | /* Test if D points to a character which is word-constituent. We have | ||
| 4375 | two special cases to check for: if past the end of string1, look at | ||
| 4376 | the first character in string2; and if before the beginning of | ||
| 4377 | string2, look at the last character in string1. */ | ||
| 4378 | #define WORDCHAR_P(d) \ | ||
| 4379 | (SYNTAX ((d) == end1 ? *string2 \ | ||
| 4380 | : (d) == string2 - 1 ? *(end1 - 1) : *(d)) \ | ||
| 4381 | == Sword) | ||
| 4382 | |||
| 4383 | /* Disabled due to a compiler bug -- see comment at case wordbound */ | ||
| 4384 | |||
| 4385 | /* The comment at case wordbound is following one, but we don't use | ||
| 4386 | AT_WORD_BOUNDARY anymore to support multibyte form. | ||
| 4387 | |||
| 4388 | The DEC Alpha C compiler 3.x generates incorrect code for the | ||
| 4389 | test WORDCHAR_P (d - 1) != WORDCHAR_P (d) in the expansion of | ||
| 4390 | AT_WORD_BOUNDARY, so this code is disabled. Expanding the | ||
| 4391 | macro and introducing temporary variables works around the bug. */ | ||
| 4392 | |||
| 4393 | #if 0 | ||
| 4394 | /* Test if the character before D and the one at D differ with respect | ||
| 4395 | to being word-constituent. */ | ||
| 4396 | #define AT_WORD_BOUNDARY(d) \ | ||
| 4397 | (AT_STRINGS_BEG (d) || AT_STRINGS_END (d) \ | ||
| 4398 | || WORDCHAR_P (d - 1) != WORDCHAR_P (d)) | ||
| 4399 | #endif | ||
| 4400 | |||
| 4401 | /* Free everything we malloc. */ | ||
| 4402 | #ifdef MATCH_MAY_ALLOCATE | ||
| 4403 | # define FREE_VAR(var) if (var) { REGEX_FREE (var); var = NULL; } else | ||
| 4404 | # define FREE_VARIABLES() \ | ||
| 4405 | do { \ | ||
| 4406 | REGEX_FREE_STACK (fail_stack.stack); \ | ||
| 4407 | FREE_VAR (regstart); \ | ||
| 4408 | FREE_VAR (regend); \ | ||
| 4409 | FREE_VAR (best_regstart); \ | ||
| 4410 | FREE_VAR (best_regend); \ | ||
| 4411 | } while (0) | ||
| 4412 | #else | ||
| 4413 | # define FREE_VARIABLES() ((void)0) /* Do nothing! But inhibit gcc warning. */ | ||
| 4414 | #endif /* not MATCH_MAY_ALLOCATE */ | ||
| 4415 | |||
| 4416 | |||
| 4417 | /* Optimization routines. */ | ||
| 4418 | |||
| 4419 | /* If the operation is a match against one or more chars, | ||
| 4420 | return a pointer to the next operation, else return NULL. */ | ||
| 4421 | static re_char * | ||
| 4422 | skip_one_char (p) | ||
| 4423 | re_char *p; | ||
| 4424 | { | ||
| 4425 | switch (SWITCH_ENUM_CAST (*p++)) | ||
| 4426 | { | ||
| 4427 | case anychar: | ||
| 4428 | break; | ||
| 4429 | |||
| 4430 | case exactn: | ||
| 4431 | p += *p + 1; | ||
| 4432 | break; | ||
| 4433 | |||
| 4434 | case charset_not: | ||
| 4435 | case charset: | ||
| 4436 | if (CHARSET_RANGE_TABLE_EXISTS_P (p - 1)) | ||
| 4437 | { | ||
| 4438 | int mcnt; | ||
| 4439 | p = CHARSET_RANGE_TABLE (p - 1); | ||
| 4440 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | ||
| 4441 | p = CHARSET_RANGE_TABLE_END (p, mcnt); | ||
| 4442 | } | ||
| 4443 | else | ||
| 4444 | p += 1 + CHARSET_BITMAP_SIZE (p - 1); | ||
| 4445 | break; | ||
| 4446 | |||
| 4447 | case syntaxspec: | ||
| 4448 | case notsyntaxspec: | ||
| 4449 | #ifdef emacs | ||
| 4450 | case categoryspec: | ||
| 4451 | case notcategoryspec: | ||
| 4452 | #endif /* emacs */ | ||
| 4453 | p++; | ||
| 4454 | break; | ||
| 4455 | |||
| 4456 | default: | ||
| 4457 | p = NULL; | ||
| 4458 | } | ||
| 4459 | return p; | ||
| 4460 | } | ||
| 4461 | |||
| 4462 | |||
| 4463 | /* Jump over non-matching operations. */ | ||
| 4464 | static re_char * | ||
| 4465 | skip_noops (p, pend) | ||
| 4466 | re_char *p, *pend; | ||
| 4467 | { | ||
| 4468 | int mcnt; | ||
| 4469 | while (p < pend) | ||
| 4470 | { | ||
| 4471 | switch (SWITCH_ENUM_CAST ((re_opcode_t) *p)) | ||
| 4472 | { | ||
| 4473 | case start_memory: | ||
| 4474 | case stop_memory: | ||
| 4475 | p += 2; break; | ||
| 4476 | case no_op: | ||
| 4477 | p += 1; break; | ||
| 4478 | case jump: | ||
| 4479 | p += 1; | ||
| 4480 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | ||
| 4481 | p += mcnt; | ||
| 4482 | break; | ||
| 4483 | default: | ||
| 4484 | return p; | ||
| 4485 | } | ||
| 4486 | } | ||
| 4487 | assert (p == pend); | ||
| 4488 | return p; | ||
| 4489 | } | ||
| 4490 | |||
| 4491 | /* Non-zero if "p1 matches something" implies "p2 fails". */ | ||
| 4492 | static int | ||
| 4493 | mutually_exclusive_p (bufp, p1, p2) | ||
| 4494 | struct re_pattern_buffer *bufp; | ||
| 4495 | re_char *p1, *p2; | ||
| 4496 | { | ||
| 4497 | re_opcode_t op2; | ||
| 4498 | const boolean multibyte = RE_MULTIBYTE_P (bufp); | ||
| 4499 | unsigned char *pend = bufp->buffer + bufp->used; | ||
| 4500 | |||
| 4501 | assert (p1 >= bufp->buffer && p1 < pend | ||
| 4502 | && p2 >= bufp->buffer && p2 <= pend); | ||
| 4503 | |||
| 4504 | /* Skip over open/close-group commands. | ||
| 4505 | If what follows this loop is a ...+ construct, | ||
| 4506 | look at what begins its body, since we will have to | ||
| 4507 | match at least one of that. */ | ||
| 4508 | p2 = skip_noops (p2, pend); | ||
| 4509 | /* The same skip can be done for p1, except that this function | ||
| 4510 | is only used in the case where p1 is a simple match operator. */ | ||
| 4511 | /* p1 = skip_noops (p1, pend); */ | ||
| 4512 | |||
| 4513 | assert (p1 >= bufp->buffer && p1 < pend | ||
| 4514 | && p2 >= bufp->buffer && p2 <= pend); | ||
| 4515 | |||
| 4516 | op2 = p2 == pend ? succeed : *p2; | ||
| 4517 | |||
| 4518 | switch (SWITCH_ENUM_CAST (op2)) | ||
| 4519 | { | ||
| 4520 | case succeed: | ||
| 4521 | case endbuf: | ||
| 4522 | /* If we're at the end of the pattern, we can change. */ | ||
| 4523 | if (skip_one_char (p1)) | ||
| 4524 | { | ||
| 4525 | DEBUG_PRINT1 (" End of pattern: fast loop.\n"); | ||
| 4526 | return 1; | ||
| 4527 | } | ||
| 4528 | break; | ||
| 4529 | |||
| 4530 | case endline: | ||
| 4531 | case exactn: | ||
| 4532 | { | ||
| 4533 | register re_wchar_t c | ||
| 4534 | = (re_opcode_t) *p2 == endline ? '\n' | ||
| 4535 | : RE_STRING_CHAR (p2 + 2, pend - p2 - 2); | ||
| 4536 | |||
| 4537 | if ((re_opcode_t) *p1 == exactn) | ||
| 4538 | { | ||
| 4539 | if (c != RE_STRING_CHAR (p1 + 2, pend - p1 - 2)) | ||
| 4540 | { | ||
| 4541 | DEBUG_PRINT3 (" '%c' != '%c' => fast loop.\n", c, p1[2]); | ||
| 4542 | return 1; | ||
| 4543 | } | ||
| 4544 | } | ||
| 4545 | |||
| 4546 | else if ((re_opcode_t) *p1 == charset | ||
| 4547 | || (re_opcode_t) *p1 == charset_not) | ||
| 4548 | { | ||
| 4549 | int not = (re_opcode_t) *p1 == charset_not; | ||
| 4550 | |||
| 4551 | /* Test if C is listed in charset (or charset_not) | ||
| 4552 | at `p1'. */ | ||
| 4553 | if (SINGLE_BYTE_CHAR_P (c)) | ||
| 4554 | { | ||
| 4555 | if (c < CHARSET_BITMAP_SIZE (p1) * BYTEWIDTH | ||
| 4556 | && p1[2 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) | ||
| 4557 | not = !not; | ||
| 4558 | } | ||
| 4559 | else if (CHARSET_RANGE_TABLE_EXISTS_P (p1)) | ||
| 4560 | CHARSET_LOOKUP_RANGE_TABLE (not, c, p1); | ||
| 4561 | |||
| 4562 | /* `not' is equal to 1 if c would match, which means | ||
| 4563 | that we can't change to pop_failure_jump. */ | ||
| 4564 | if (!not) | ||
| 4565 | { | ||
| 4566 | DEBUG_PRINT1 (" No match => fast loop.\n"); | ||
| 4567 | return 1; | ||
| 4568 | } | ||
| 4569 | } | ||
| 4570 | else if ((re_opcode_t) *p1 == anychar | ||
| 4571 | && c == '\n') | ||
| 4572 | { | ||
| 4573 | DEBUG_PRINT1 (" . != \\n => fast loop.\n"); | ||
| 4574 | return 1; | ||
| 4575 | } | ||
| 4576 | } | ||
| 4577 | break; | ||
| 4578 | |||
| 4579 | case charset: | ||
| 4580 | { | ||
| 4581 | if ((re_opcode_t) *p1 == exactn) | ||
| 4582 | /* Reuse the code above. */ | ||
| 4583 | return mutually_exclusive_p (bufp, p2, p1); | ||
| 4584 | |||
| 4585 | /* It is hard to list up all the character in charset | ||
| 4586 | P2 if it includes multibyte character. Give up in | ||
| 4587 | such case. */ | ||
| 4588 | else if (!multibyte || !CHARSET_RANGE_TABLE_EXISTS_P (p2)) | ||
| 4589 | { | ||
| 4590 | /* Now, we are sure that P2 has no range table. | ||
| 4591 | So, for the size of bitmap in P2, `p2[1]' is | ||
| 4592 | enough. But P1 may have range table, so the | ||
| 4593 | size of bitmap table of P1 is extracted by | ||
| 4594 | using macro `CHARSET_BITMAP_SIZE'. | ||
| 4595 | |||
| 4596 | Since we know that all the character listed in | ||
| 4597 | P2 is ASCII, it is enough to test only bitmap | ||
| 4598 | table of P1. */ | ||
| 4599 | |||
| 4600 | if ((re_opcode_t) *p1 == charset) | ||
| 4601 | { | ||
| 4602 | int idx; | ||
| 4603 | /* We win if the charset inside the loop | ||
| 4604 | has no overlap with the one after the loop. */ | ||
| 4605 | for (idx = 0; | ||
| 4606 | (idx < (int) p2[1] | ||
| 4607 | && idx < CHARSET_BITMAP_SIZE (p1)); | ||
| 4608 | idx++) | ||
| 4609 | if ((p2[2 + idx] & p1[2 + idx]) != 0) | ||
| 4610 | break; | ||
| 4611 | |||
| 4612 | if (idx == p2[1] | ||
| 4613 | || idx == CHARSET_BITMAP_SIZE (p1)) | ||
| 4614 | { | ||
| 4615 | DEBUG_PRINT1 (" No match => fast loop.\n"); | ||
| 4616 | return 1; | ||
| 4617 | } | ||
| 4618 | } | ||
| 4619 | else if ((re_opcode_t) *p1 == charset_not) | ||
| 4620 | { | ||
| 4621 | int idx; | ||
| 4622 | /* We win if the charset_not inside the loop lists | ||
| 4623 | every character listed in the charset after. */ | ||
| 4624 | for (idx = 0; idx < (int) p2[1]; idx++) | ||
| 4625 | if (! (p2[2 + idx] == 0 | ||
| 4626 | || (idx < CHARSET_BITMAP_SIZE (p1) | ||
| 4627 | && ((p2[2 + idx] & ~ p1[2 + idx]) == 0)))) | ||
| 4628 | break; | ||
| 4629 | |||
| 4630 | if (idx == p2[1]) | ||
| 4631 | { | ||
| 4632 | DEBUG_PRINT1 (" No match => fast loop.\n"); | ||
| 4633 | return 1; | ||
| 4634 | } | ||
| 4635 | } | ||
| 4636 | } | ||
| 4637 | } | ||
| 4638 | break; | ||
| 4639 | |||
| 4640 | case charset_not: | ||
| 4641 | switch (SWITCH_ENUM_CAST (*p1)) | ||
| 4642 | { | ||
| 4643 | case exactn: | ||
| 4644 | case charset: | ||
| 4645 | /* Reuse the code above. */ | ||
| 4646 | return mutually_exclusive_p (bufp, p2, p1); | ||
| 4647 | case charset_not: | ||
| 4648 | /* When we have two charset_not, it's very unlikely that | ||
| 4649 | they don't overlap. The union of the two sets of excluded | ||
| 4650 | chars should cover all possible chars, which, as a matter of | ||
| 4651 | fact, is virtually impossible in multibyte buffers. */ | ||
| 4652 | break; | ||
| 4653 | } | ||
| 4654 | break; | ||
| 4655 | |||
| 4656 | case wordend: | ||
| 4657 | case notsyntaxspec: | ||
| 4658 | return ((re_opcode_t) *p1 == syntaxspec | ||
| 4659 | && p1[1] == (op2 == wordend ? Sword : p2[1])); | ||
| 4660 | |||
| 4661 | case wordbeg: | ||
| 4662 | case syntaxspec: | ||
| 4663 | return ((re_opcode_t) *p1 == notsyntaxspec | ||
| 4664 | && p1[1] == (op2 == wordend ? Sword : p2[1])); | ||
| 4665 | |||
| 4666 | case wordbound: | ||
| 4667 | return (((re_opcode_t) *p1 == notsyntaxspec | ||
| 4668 | || (re_opcode_t) *p1 == syntaxspec) | ||
| 4669 | && p1[1] == Sword); | ||
| 4670 | |||
| 4671 | #ifdef emacs | ||
| 4672 | case categoryspec: | ||
| 4673 | return ((re_opcode_t) *p1 == notcategoryspec && p1[1] == p2[1]); | ||
| 4674 | case notcategoryspec: | ||
| 4675 | return ((re_opcode_t) *p1 == categoryspec && p1[1] == p2[1]); | ||
| 4676 | #endif /* emacs */ | ||
| 4677 | |||
| 4678 | default: | ||
| 4679 | ; | ||
| 4680 | } | ||
| 4681 | |||
| 4682 | /* Safe default. */ | ||
| 4683 | return 0; | ||
| 4684 | } | ||
| 4685 | |||
| 4686 | |||
| 4687 | /* Matching routines. */ | ||
| 4688 | |||
| 4689 | #ifndef emacs /* Emacs never uses this. */ | ||
| 4690 | /* re_match is like re_match_2 except it takes only a single string. */ | ||
| 4691 | |||
| 4692 | int | ||
| 4693 | re_match (bufp, string, size, pos, regs) | ||
| 4694 | struct re_pattern_buffer *bufp; | ||
| 4695 | const char *string; | ||
| 4696 | int size, pos; | ||
| 4697 | struct re_registers *regs; | ||
| 4698 | { | ||
| 4699 | int result = re_match_2_internal (bufp, NULL, 0, (re_char*) string, size, | ||
| 4700 | pos, regs, size); | ||
| 4701 | # if defined C_ALLOCA && !defined REGEX_MALLOC | ||
| 4702 | alloca (0); | ||
| 4703 | # endif | ||
| 4704 | return result; | ||
| 4705 | } | ||
| 4706 | WEAK_ALIAS (__re_match, re_match) | ||
| 4707 | #endif /* not emacs */ | ||
| 4708 | |||
| 4709 | #ifdef emacs | ||
| 4710 | /* In Emacs, this is the string or buffer in which we | ||
| 4711 | are matching. It is used for looking up syntax properties. */ | ||
| 4712 | Lisp_Object re_match_object; | ||
| 4713 | #endif | ||
| 4714 | |||
| 4715 | /* re_match_2 matches the compiled pattern in BUFP against the | ||
| 4716 | the (virtual) concatenation of STRING1 and STRING2 (of length SIZE1 | ||
| 4717 | and SIZE2, respectively). We start matching at POS, and stop | ||
| 4718 | matching at STOP. | ||
| 4719 | |||
| 4720 | If REGS is non-null and the `no_sub' field of BUFP is nonzero, we | ||
| 4721 | store offsets for the substring each group matched in REGS. See the | ||
| 4722 | documentation for exactly how many groups we fill. | ||
| 4723 | |||
| 4724 | We return -1 if no match, -2 if an internal error (such as the | ||
| 4725 | failure stack overflowing). Otherwise, we return the length of the | ||
| 4726 | matched substring. */ | ||
| 4727 | |||
| 4728 | int | ||
| 4729 | re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop) | ||
| 4730 | struct re_pattern_buffer *bufp; | ||
| 4731 | const char *string1, *string2; | ||
| 4732 | int size1, size2; | ||
| 4733 | int pos; | ||
| 4734 | struct re_registers *regs; | ||
| 4735 | int stop; | ||
| 4736 | { | ||
| 4737 | int result; | ||
| 4738 | |||
| 4739 | #ifdef emacs | ||
| 4740 | int charpos; | ||
| 4741 | gl_state.object = re_match_object; | ||
| 4742 | charpos = SYNTAX_TABLE_BYTE_TO_CHAR (POS_AS_IN_BUFFER (pos)); | ||
| 4743 | SETUP_SYNTAX_TABLE_FOR_OBJECT (re_match_object, charpos, 1); | ||
| 4744 | #endif | ||
| 4745 | |||
| 4746 | result = re_match_2_internal (bufp, (re_char*) string1, size1, | ||
| 4747 | (re_char*) string2, size2, | ||
| 4748 | pos, regs, stop); | ||
| 4749 | #if defined C_ALLOCA && !defined REGEX_MALLOC | ||
| 4750 | alloca (0); | ||
| 4751 | #endif | ||
| 4752 | return result; | ||
| 4753 | } | ||
| 4754 | WEAK_ALIAS (__re_match_2, re_match_2) | ||
| 4755 | |||
| 4756 | /* This is a separate function so that we can force an alloca cleanup | ||
| 4757 | afterwards. */ | ||
| 4758 | static int | ||
| 4759 | re_match_2_internal (bufp, string1, size1, string2, size2, pos, regs, stop) | ||
| 4760 | struct re_pattern_buffer *bufp; | ||
| 4761 | re_char *string1, *string2; | ||
| 4762 | int size1, size2; | ||
| 4763 | int pos; | ||
| 4764 | struct re_registers *regs; | ||
| 4765 | int stop; | ||
| 4766 | { | ||
| 4767 | /* General temporaries. */ | ||
| 4768 | int mcnt; | ||
| 4769 | size_t reg; | ||
| 4770 | boolean not; | ||
| 4771 | |||
| 4772 | /* Just past the end of the corresponding string. */ | ||
| 4773 | re_char *end1, *end2; | ||
| 4774 | |||
| 4775 | /* Pointers into string1 and string2, just past the last characters in | ||
| 4776 | each to consider matching. */ | ||
| 4777 | re_char *end_match_1, *end_match_2; | ||
| 4778 | |||
| 4779 | /* Where we are in the data, and the end of the current string. */ | ||
| 4780 | re_char *d, *dend; | ||
| 4781 | |||
| 4782 | /* Used sometimes to remember where we were before starting matching | ||
| 4783 | an operator so that we can go back in case of failure. This "atomic" | ||
| 4784 | behavior of matching opcodes is indispensable to the correctness | ||
| 4785 | of the on_failure_keep_string_jump optimization. */ | ||
| 4786 | re_char *dfail; | ||
| 4787 | |||
| 4788 | /* Where we are in the pattern, and the end of the pattern. */ | ||
| 4789 | re_char *p = bufp->buffer; | ||
| 4790 | re_char *pend = p + bufp->used; | ||
| 4791 | |||
| 4792 | /* We use this to map every character in the string. */ | ||
| 4793 | RE_TRANSLATE_TYPE translate = bufp->translate; | ||
| 4794 | |||
| 4795 | /* Nonzero if we have to concern multibyte character. */ | ||
| 4796 | const boolean multibyte = RE_MULTIBYTE_P (bufp); | ||
| 4797 | |||
| 4798 | /* Failure point stack. Each place that can handle a failure further | ||
| 4799 | down the line pushes a failure point on this stack. It consists of | ||
| 4800 | regstart, and regend for all registers corresponding to | ||
| 4801 | the subexpressions we're currently inside, plus the number of such | ||
| 4802 | registers, and, finally, two char *'s. The first char * is where | ||
| 4803 | to resume scanning the pattern; the second one is where to resume | ||
| 4804 | scanning the strings. */ | ||
| 4805 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */ | ||
| 4806 | fail_stack_type fail_stack; | ||
| 4807 | #endif | ||
| 4808 | #ifdef DEBUG | ||
| 4809 | unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0; | ||
| 4810 | #endif | ||
| 4811 | |||
| 4812 | #if defined REL_ALLOC && defined REGEX_MALLOC | ||
| 4813 | /* This holds the pointer to the failure stack, when | ||
| 4814 | it is allocated relocatably. */ | ||
| 4815 | fail_stack_elt_t *failure_stack_ptr; | ||
| 4816 | #endif | ||
| 4817 | |||
| 4818 | /* We fill all the registers internally, independent of what we | ||
| 4819 | return, for use in backreferences. The number here includes | ||
| 4820 | an element for register zero. */ | ||
| 4821 | size_t num_regs = bufp->re_nsub + 1; | ||
| 4822 | |||
| 4823 | /* Information on the contents of registers. These are pointers into | ||
| 4824 | the input strings; they record just what was matched (on this | ||
| 4825 | attempt) by a subexpression part of the pattern, that is, the | ||
| 4826 | regnum-th regstart pointer points to where in the pattern we began | ||
| 4827 | matching and the regnum-th regend points to right after where we | ||
| 4828 | stopped matching the regnum-th subexpression. (The zeroth register | ||
| 4829 | keeps track of what the whole pattern matches.) */ | ||
| 4830 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | ||
| 4831 | re_char **regstart, **regend; | ||
| 4832 | #endif | ||
| 4833 | |||
| 4834 | /* The following record the register info as found in the above | ||
| 4835 | variables when we find a match better than any we've seen before. | ||
| 4836 | This happens as we backtrack through the failure points, which in | ||
| 4837 | turn happens only if we have not yet matched the entire string. */ | ||
| 4838 | unsigned best_regs_set = false; | ||
| 4839 | #ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */ | ||
| 4840 | re_char **best_regstart, **best_regend; | ||
| 4841 | #endif | ||
| 4842 | |||
| 4843 | /* Logically, this is `best_regend[0]'. But we don't want to have to | ||
| 4844 | allocate space for that if we're not allocating space for anything | ||
| 4845 | else (see below). Also, we never need info about register 0 for | ||
| 4846 | any of the other register vectors, and it seems rather a kludge to | ||
| 4847 | treat `best_regend' differently than the rest. So we keep track of | ||
| 4848 | the end of the best match so far in a separate variable. We | ||
| 4849 | initialize this to NULL so that when we backtrack the first time | ||
| 4850 | and need to test it, it's not garbage. */ | ||
| 4851 | re_char *match_end = NULL; | ||
| 4852 | |||
| 4853 | #ifdef DEBUG | ||
| 4854 | /* Counts the total number of registers pushed. */ | ||
| 4855 | unsigned num_regs_pushed = 0; | ||
| 4856 | #endif | ||
| 4857 | |||
| 4858 | DEBUG_PRINT1 ("\n\nEntering re_match_2.\n"); | ||
| 4859 | |||
| 4860 | INIT_FAIL_STACK (); | ||
| 4861 | |||
| 4862 | #ifdef MATCH_MAY_ALLOCATE | ||
| 4863 | /* Do not bother to initialize all the register variables if there are | ||
| 4864 | no groups in the pattern, as it takes a fair amount of time. If | ||
| 4865 | there are groups, we include space for register 0 (the whole | ||
| 4866 | pattern), even though we never use it, since it simplifies the | ||
| 4867 | array indexing. We should fix this. */ | ||
| 4868 | if (bufp->re_nsub) | ||
| 4869 | { | ||
| 4870 | regstart = REGEX_TALLOC (num_regs, re_char *); | ||
| 4871 | regend = REGEX_TALLOC (num_regs, re_char *); | ||
| 4872 | best_regstart = REGEX_TALLOC (num_regs, re_char *); | ||
| 4873 | best_regend = REGEX_TALLOC (num_regs, re_char *); | ||
| 4874 | |||
| 4875 | if (!(regstart && regend && best_regstart && best_regend)) | ||
| 4876 | { | ||
| 4877 | FREE_VARIABLES (); | ||
| 4878 | return -2; | ||
| 4879 | } | ||
| 4880 | } | ||
| 4881 | else | ||
| 4882 | { | ||
| 4883 | /* We must initialize all our variables to NULL, so that | ||
| 4884 | `FREE_VARIABLES' doesn't try to free them. */ | ||
| 4885 | regstart = regend = best_regstart = best_regend = NULL; | ||
| 4886 | } | ||
| 4887 | #endif /* MATCH_MAY_ALLOCATE */ | ||
| 4888 | |||
| 4889 | /* The starting position is bogus. */ | ||
| 4890 | if (pos < 0 || pos > size1 + size2) | ||
| 4891 | { | ||
| 4892 | FREE_VARIABLES (); | ||
| 4893 | return -1; | ||
| 4894 | } | ||
| 4895 | |||
| 4896 | /* Initialize subexpression text positions to -1 to mark ones that no | ||
| 4897 | start_memory/stop_memory has been seen for. Also initialize the | ||
| 4898 | register information struct. */ | ||
| 4899 | for (reg = 1; reg < num_regs; reg++) | ||
| 4900 | regstart[reg] = regend[reg] = NULL; | ||
| 4901 | |||
| 4902 | /* We move `string1' into `string2' if the latter's empty -- but not if | ||
| 4903 | `string1' is null. */ | ||
| 4904 | if (size2 == 0 && string1 != NULL) | ||
| 4905 | { | ||
| 4906 | string2 = string1; | ||
| 4907 | size2 = size1; | ||
| 4908 | string1 = 0; | ||
| 4909 | size1 = 0; | ||
| 4910 | } | ||
| 4911 | end1 = string1 + size1; | ||
| 4912 | end2 = string2 + size2; | ||
| 4913 | |||
| 4914 | /* `p' scans through the pattern as `d' scans through the data. | ||
| 4915 | `dend' is the end of the input string that `d' points within. `d' | ||
| 4916 | is advanced into the following input string whenever necessary, but | ||
| 4917 | this happens before fetching; therefore, at the beginning of the | ||
| 4918 | loop, `d' can be pointing at the end of a string, but it cannot | ||
| 4919 | equal `string2'. */ | ||
| 4920 | if (pos >= size1) | ||
| 4921 | { | ||
| 4922 | /* Only match within string2. */ | ||
| 4923 | d = string2 + pos - size1; | ||
| 4924 | dend = end_match_2 = string2 + stop - size1; | ||
| 4925 | end_match_1 = end1; /* Just to give it a value. */ | ||
| 4926 | } | ||
| 4927 | else | ||
| 4928 | { | ||
| 4929 | if (stop < size1) | ||
| 4930 | { | ||
| 4931 | /* Only match within string1. */ | ||
| 4932 | end_match_1 = string1 + stop; | ||
| 4933 | /* BEWARE! | ||
| 4934 | When we reach end_match_1, PREFETCH normally switches to string2. | ||
| 4935 | But in the present case, this means that just doing a PREFETCH | ||
| 4936 | makes us jump from `stop' to `gap' within the string. | ||
| 4937 | What we really want here is for the search to stop as | ||
| 4938 | soon as we hit end_match_1. That's why we set end_match_2 | ||
| 4939 | to end_match_1 (since PREFETCH fails as soon as we hit | ||
| 4940 | end_match_2). */ | ||
| 4941 | end_match_2 = end_match_1; | ||
| 4942 | } | ||
| 4943 | else | ||
| 4944 | { /* It's important to use this code when stop == size so that | ||
| 4945 | moving `d' from end1 to string2 will not prevent the d == dend | ||
| 4946 | check from catching the end of string. */ | ||
| 4947 | end_match_1 = end1; | ||
| 4948 | end_match_2 = string2 + stop - size1; | ||
| 4949 | } | ||
| 4950 | d = string1 + pos; | ||
| 4951 | dend = end_match_1; | ||
| 4952 | } | ||
| 4953 | |||
| 4954 | DEBUG_PRINT1 ("The compiled pattern is: "); | ||
| 4955 | DEBUG_PRINT_COMPILED_PATTERN (bufp, p, pend); | ||
| 4956 | DEBUG_PRINT1 ("The string to match is: `"); | ||
| 4957 | DEBUG_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2); | ||
| 4958 | DEBUG_PRINT1 ("'\n"); | ||
| 4959 | |||
| 4960 | /* This loops over pattern commands. It exits by returning from the | ||
| 4961 | function if the match is complete, or it drops through if the match | ||
| 4962 | fails at this starting point in the input data. */ | ||
| 4963 | for (;;) | ||
| 4964 | { | ||
| 4965 | DEBUG_PRINT2 ("\n%p: ", p); | ||
| 4966 | |||
| 4967 | if (p == pend) | ||
| 4968 | { /* End of pattern means we might have succeeded. */ | ||
| 4969 | DEBUG_PRINT1 ("end of pattern ... "); | ||
| 4970 | |||
| 4971 | /* If we haven't matched the entire string, and we want the | ||
| 4972 | longest match, try backtracking. */ | ||
| 4973 | if (d != end_match_2) | ||
| 4974 | { | ||
| 4975 | /* 1 if this match ends in the same string (string1 or string2) | ||
| 4976 | as the best previous match. */ | ||
| 4977 | boolean same_str_p = (FIRST_STRING_P (match_end) | ||
| 4978 | == FIRST_STRING_P (d)); | ||
| 4979 | /* 1 if this match is the best seen so far. */ | ||
| 4980 | boolean best_match_p; | ||
| 4981 | |||
| 4982 | /* AIX compiler got confused when this was combined | ||
| 4983 | with the previous declaration. */ | ||
| 4984 | if (same_str_p) | ||
| 4985 | best_match_p = d > match_end; | ||
| 4986 | else | ||
| 4987 | best_match_p = !FIRST_STRING_P (d); | ||
| 4988 | |||
| 4989 | DEBUG_PRINT1 ("backtracking.\n"); | ||
| 4990 | |||
| 4991 | if (!FAIL_STACK_EMPTY ()) | ||
| 4992 | { /* More failure points to try. */ | ||
| 4993 | |||
| 4994 | /* If exceeds best match so far, save it. */ | ||
| 4995 | if (!best_regs_set || best_match_p) | ||
| 4996 | { | ||
| 4997 | best_regs_set = true; | ||
| 4998 | match_end = d; | ||
| 4999 | |||
| 5000 | DEBUG_PRINT1 ("\nSAVING match as best so far.\n"); | ||
| 5001 | |||
| 5002 | for (reg = 1; reg < num_regs; reg++) | ||
| 5003 | { | ||
| 5004 | best_regstart[reg] = regstart[reg]; | ||
| 5005 | best_regend[reg] = regend[reg]; | ||
| 5006 | } | ||
| 5007 | } | ||
| 5008 | goto fail; | ||
| 5009 | } | ||
| 5010 | |||
| 5011 | /* If no failure points, don't restore garbage. And if | ||
| 5012 | last match is real best match, don't restore second | ||
| 5013 | best one. */ | ||
| 5014 | else if (best_regs_set && !best_match_p) | ||
| 5015 | { | ||
| 5016 | restore_best_regs: | ||
| 5017 | /* Restore best match. It may happen that `dend == | ||
| 5018 | end_match_1' while the restored d is in string2. | ||
| 5019 | For example, the pattern `x.*y.*z' against the | ||
| 5020 | strings `x-' and `y-z-', if the two strings are | ||
| 5021 | not consecutive in memory. */ | ||
| 5022 | DEBUG_PRINT1 ("Restoring best registers.\n"); | ||
| 5023 | |||
| 5024 | d = match_end; | ||
| 5025 | dend = ((d >= string1 && d <= end1) | ||
| 5026 | ? end_match_1 : end_match_2); | ||
| 5027 | |||
| 5028 | for (reg = 1; reg < num_regs; reg++) | ||
| 5029 | { | ||
| 5030 | regstart[reg] = best_regstart[reg]; | ||
| 5031 | regend[reg] = best_regend[reg]; | ||
| 5032 | } | ||
| 5033 | } | ||
| 5034 | } /* d != end_match_2 */ | ||
| 5035 | |||
| 5036 | succeed_label: | ||
| 5037 | DEBUG_PRINT1 ("Accepting match.\n"); | ||
| 5038 | |||
| 5039 | /* If caller wants register contents data back, do it. */ | ||
| 5040 | if (regs && !bufp->no_sub) | ||
| 5041 | { | ||
| 5042 | /* Have the register data arrays been allocated? */ | ||
| 5043 | if (bufp->regs_allocated == REGS_UNALLOCATED) | ||
| 5044 | { /* No. So allocate them with malloc. We need one | ||
| 5045 | extra element beyond `num_regs' for the `-1' marker | ||
| 5046 | GNU code uses. */ | ||
| 5047 | regs->num_regs = MAX (RE_NREGS, num_regs + 1); | ||
| 5048 | regs->start = TALLOC (regs->num_regs, regoff_t); | ||
| 5049 | regs->end = TALLOC (regs->num_regs, regoff_t); | ||
| 5050 | if (regs->start == NULL || regs->end == NULL) | ||
| 5051 | { | ||
| 5052 | FREE_VARIABLES (); | ||
| 5053 | return -2; | ||
| 5054 | } | ||
| 5055 | bufp->regs_allocated = REGS_REALLOCATE; | ||
| 5056 | } | ||
| 5057 | else if (bufp->regs_allocated == REGS_REALLOCATE) | ||
| 5058 | { /* Yes. If we need more elements than were already | ||
| 5059 | allocated, reallocate them. If we need fewer, just | ||
| 5060 | leave it alone. */ | ||
| 5061 | if (regs->num_regs < num_regs + 1) | ||
| 5062 | { | ||
| 5063 | regs->num_regs = num_regs + 1; | ||
| 5064 | RETALLOC (regs->start, regs->num_regs, regoff_t); | ||
| 5065 | RETALLOC (regs->end, regs->num_regs, regoff_t); | ||
| 5066 | if (regs->start == NULL || regs->end == NULL) | ||
| 5067 | { | ||
| 5068 | FREE_VARIABLES (); | ||
| 5069 | return -2; | ||
| 5070 | } | ||
| 5071 | } | ||
| 5072 | } | ||
| 5073 | else | ||
| 5074 | { | ||
| 5075 | /* These braces fend off a "empty body in an else-statement" | ||
| 5076 | warning under GCC when assert expands to nothing. */ | ||
| 5077 | assert (bufp->regs_allocated == REGS_FIXED); | ||
| 5078 | } | ||
| 5079 | |||
| 5080 | /* Convert the pointer data in `regstart' and `regend' to | ||
| 5081 | indices. Register zero has to be set differently, | ||
| 5082 | since we haven't kept track of any info for it. */ | ||
| 5083 | if (regs->num_regs > 0) | ||
| 5084 | { | ||
| 5085 | regs->start[0] = pos; | ||
| 5086 | regs->end[0] = POINTER_TO_OFFSET (d); | ||
| 5087 | } | ||
| 5088 | |||
| 5089 | /* Go through the first `min (num_regs, regs->num_regs)' | ||
| 5090 | registers, since that is all we initialized. */ | ||
| 5091 | for (reg = 1; reg < MIN (num_regs, regs->num_regs); reg++) | ||
| 5092 | { | ||
| 5093 | if (REG_UNSET (regstart[reg]) || REG_UNSET (regend[reg])) | ||
| 5094 | regs->start[reg] = regs->end[reg] = -1; | ||
| 5095 | else | ||
| 5096 | { | ||
| 5097 | regs->start[reg] | ||
| 5098 | = (regoff_t) POINTER_TO_OFFSET (regstart[reg]); | ||
| 5099 | regs->end[reg] | ||
| 5100 | = (regoff_t) POINTER_TO_OFFSET (regend[reg]); | ||
| 5101 | } | ||
| 5102 | } | ||
| 5103 | |||
| 5104 | /* If the regs structure we return has more elements than | ||
| 5105 | were in the pattern, set the extra elements to -1. If | ||
| 5106 | we (re)allocated the registers, this is the case, | ||
| 5107 | because we always allocate enough to have at least one | ||
| 5108 | -1 at the end. */ | ||
| 5109 | for (reg = num_regs; reg < regs->num_regs; reg++) | ||
| 5110 | regs->start[reg] = regs->end[reg] = -1; | ||
| 5111 | } /* regs && !bufp->no_sub */ | ||
| 5112 | |||
| 5113 | DEBUG_PRINT4 ("%u failure points pushed, %u popped (%u remain).\n", | ||
| 5114 | nfailure_points_pushed, nfailure_points_popped, | ||
| 5115 | nfailure_points_pushed - nfailure_points_popped); | ||
| 5116 | DEBUG_PRINT2 ("%u registers pushed.\n", num_regs_pushed); | ||
| 5117 | |||
| 5118 | mcnt = POINTER_TO_OFFSET (d) - pos; | ||
| 5119 | |||
| 5120 | DEBUG_PRINT2 ("Returning %d from re_match_2.\n", mcnt); | ||
| 5121 | |||
| 5122 | FREE_VARIABLES (); | ||
| 5123 | return mcnt; | ||
| 5124 | } | ||
| 5125 | |||
| 5126 | /* Otherwise match next pattern command. */ | ||
| 5127 | switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++)) | ||
| 5128 | { | ||
| 5129 | /* Ignore these. Used to ignore the n of succeed_n's which | ||
| 5130 | currently have n == 0. */ | ||
| 5131 | case no_op: | ||
| 5132 | DEBUG_PRINT1 ("EXECUTING no_op.\n"); | ||
| 5133 | break; | ||
| 5134 | |||
| 5135 | case succeed: | ||
| 5136 | DEBUG_PRINT1 ("EXECUTING succeed.\n"); | ||
| 5137 | goto succeed_label; | ||
| 5138 | |||
| 5139 | /* Match the next n pattern characters exactly. The following | ||
| 5140 | byte in the pattern defines n, and the n bytes after that | ||
| 5141 | are the characters to match. */ | ||
| 5142 | case exactn: | ||
| 5143 | mcnt = *p++; | ||
| 5144 | DEBUG_PRINT2 ("EXECUTING exactn %d.\n", mcnt); | ||
| 5145 | |||
| 5146 | /* Remember the start point to rollback upon failure. */ | ||
| 5147 | dfail = d; | ||
| 5148 | |||
| 5149 | /* This is written out as an if-else so we don't waste time | ||
| 5150 | testing `translate' inside the loop. */ | ||
| 5151 | if (RE_TRANSLATE_P (translate)) | ||
| 5152 | { | ||
| 5153 | if (multibyte) | ||
| 5154 | do | ||
| 5155 | { | ||
| 5156 | int pat_charlen, buf_charlen; | ||
| 5157 | unsigned int pat_ch, buf_ch; | ||
| 5158 | |||
| 5159 | PREFETCH (); | ||
| 5160 | pat_ch = STRING_CHAR_AND_LENGTH (p, pend - p, pat_charlen); | ||
| 5161 | buf_ch = STRING_CHAR_AND_LENGTH (d, dend - d, buf_charlen); | ||
| 5162 | |||
| 5163 | if (RE_TRANSLATE (translate, buf_ch) | ||
| 5164 | != pat_ch) | ||
| 5165 | { | ||
| 5166 | d = dfail; | ||
| 5167 | goto fail; | ||
| 5168 | } | ||
| 5169 | |||
| 5170 | p += pat_charlen; | ||
| 5171 | d += buf_charlen; | ||
| 5172 | mcnt -= pat_charlen; | ||
| 5173 | } | ||
| 5174 | while (mcnt > 0); | ||
| 5175 | else | ||
| 5176 | do | ||
| 5177 | { | ||
| 5178 | PREFETCH (); | ||
| 5179 | if (RE_TRANSLATE (translate, *d) != *p++) | ||
| 5180 | { | ||
| 5181 | d = dfail; | ||
| 5182 | goto fail; | ||
| 5183 | } | ||
| 5184 | d++; | ||
| 5185 | } | ||
| 5186 | while (--mcnt); | ||
| 5187 | } | ||
| 5188 | else | ||
| 5189 | { | ||
| 5190 | do | ||
| 5191 | { | ||
| 5192 | PREFETCH (); | ||
| 5193 | if (*d++ != *p++) | ||
| 5194 | { | ||
| 5195 | d = dfail; | ||
| 5196 | goto fail; | ||
| 5197 | } | ||
| 5198 | } | ||
| 5199 | while (--mcnt); | ||
| 5200 | } | ||
| 5201 | break; | ||
| 5202 | |||
| 5203 | |||
| 5204 | /* Match any character except possibly a newline or a null. */ | ||
| 5205 | case anychar: | ||
| 5206 | { | ||
| 5207 | int buf_charlen; | ||
| 5208 | re_wchar_t buf_ch; | ||
| 5209 | |||
| 5210 | DEBUG_PRINT1 ("EXECUTING anychar.\n"); | ||
| 5211 | |||
| 5212 | PREFETCH (); | ||
| 5213 | buf_ch = RE_STRING_CHAR_AND_LENGTH (d, dend - d, buf_charlen); | ||
| 5214 | buf_ch = TRANSLATE (buf_ch); | ||
| 5215 | |||
| 5216 | if ((!(bufp->syntax & RE_DOT_NEWLINE) | ||
| 5217 | && buf_ch == '\n') | ||
| 5218 | || ((bufp->syntax & RE_DOT_NOT_NULL) | ||
| 5219 | && buf_ch == '\000')) | ||
| 5220 | goto fail; | ||
| 5221 | |||
| 5222 | DEBUG_PRINT2 (" Matched `%d'.\n", *d); | ||
| 5223 | d += buf_charlen; | ||
| 5224 | } | ||
| 5225 | break; | ||
| 5226 | |||
| 5227 | |||
| 5228 | case charset: | ||
| 5229 | case charset_not: | ||
| 5230 | { | ||
| 5231 | register unsigned int c; | ||
| 5232 | boolean not = (re_opcode_t) *(p - 1) == charset_not; | ||
| 5233 | int len; | ||
| 5234 | |||
| 5235 | /* Start of actual range_table, or end of bitmap if there is no | ||
| 5236 | range table. */ | ||
| 5237 | re_char *range_table; | ||
| 5238 | |||
| 5239 | /* Nonzero if there is a range table. */ | ||
| 5240 | int range_table_exists; | ||
| 5241 | |||
| 5242 | /* Number of ranges of range table. This is not included | ||
| 5243 | in the initial byte-length of the command. */ | ||
| 5244 | int count = 0; | ||
| 5245 | |||
| 5246 | DEBUG_PRINT2 ("EXECUTING charset%s.\n", not ? "_not" : ""); | ||
| 5247 | |||
| 5248 | range_table_exists = CHARSET_RANGE_TABLE_EXISTS_P (&p[-1]); | ||
| 5249 | |||
| 5250 | if (range_table_exists) | ||
| 5251 | { | ||
| 5252 | range_table = CHARSET_RANGE_TABLE (&p[-1]); /* Past the bitmap. */ | ||
| 5253 | EXTRACT_NUMBER_AND_INCR (count, range_table); | ||
| 5254 | } | ||
| 5255 | |||
| 5256 | PREFETCH (); | ||
| 5257 | c = RE_STRING_CHAR_AND_LENGTH (d, dend - d, len); | ||
| 5258 | c = TRANSLATE (c); /* The character to match. */ | ||
| 5259 | |||
| 5260 | if (SINGLE_BYTE_CHAR_P (c)) | ||
| 5261 | { /* Lookup bitmap. */ | ||
| 5262 | /* Cast to `unsigned' instead of `unsigned char' in | ||
| 5263 | case the bit list is a full 32 bytes long. */ | ||
| 5264 | if (c < (unsigned) (CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH) | ||
| 5265 | && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) | ||
| 5266 | not = !not; | ||
| 5267 | } | ||
| 5268 | #ifdef emacs | ||
| 5269 | else if (range_table_exists) | ||
| 5270 | { | ||
| 5271 | int class_bits = CHARSET_RANGE_TABLE_BITS (&p[-1]); | ||
| 5272 | |||
| 5273 | if ( (class_bits & BIT_LOWER && ISLOWER (c)) | ||
| 5274 | | (class_bits & BIT_MULTIBYTE) | ||
| 5275 | | (class_bits & BIT_PUNCT && ISPUNCT (c)) | ||
| 5276 | | (class_bits & BIT_SPACE && ISSPACE (c)) | ||
| 5277 | | (class_bits & BIT_UPPER && ISUPPER (c)) | ||
| 5278 | | (class_bits & BIT_WORD && ISWORD (c))) | ||
| 5279 | not = !not; | ||
| 5280 | else | ||
| 5281 | CHARSET_LOOKUP_RANGE_TABLE_RAW (not, c, range_table, count); | ||
| 5282 | } | ||
| 5283 | #endif /* emacs */ | ||
| 5284 | |||
| 5285 | if (range_table_exists) | ||
| 5286 | p = CHARSET_RANGE_TABLE_END (range_table, count); | ||
| 5287 | else | ||
| 5288 | p += CHARSET_BITMAP_SIZE (&p[-1]) + 1; | ||
| 5289 | |||
| 5290 | if (!not) goto fail; | ||
| 5291 | |||
| 5292 | d += len; | ||
| 5293 | break; | ||
| 5294 | } | ||
| 5295 | |||
| 5296 | |||
| 5297 | /* The beginning of a group is represented by start_memory. | ||
| 5298 | The argument is the register number. The text | ||
| 5299 | matched within the group is recorded (in the internal | ||
| 5300 | registers data structure) under the register number. */ | ||
| 5301 | case start_memory: | ||
| 5302 | DEBUG_PRINT2 ("EXECUTING start_memory %d:\n", *p); | ||
| 5303 | |||
| 5304 | /* In case we need to undo this operation (via backtracking). */ | ||
| 5305 | PUSH_FAILURE_REG ((unsigned int)*p); | ||
| 5306 | |||
| 5307 | regstart[*p] = d; | ||
| 5308 | regend[*p] = NULL; /* probably unnecessary. -sm */ | ||
| 5309 | DEBUG_PRINT2 (" regstart: %d\n", POINTER_TO_OFFSET (regstart[*p])); | ||
| 5310 | |||
| 5311 | /* Move past the register number and inner group count. */ | ||
| 5312 | p += 1; | ||
| 5313 | break; | ||
| 5314 | |||
| 5315 | |||
| 5316 | /* The stop_memory opcode represents the end of a group. Its | ||
| 5317 | argument is the same as start_memory's: the register number. */ | ||
| 5318 | case stop_memory: | ||
| 5319 | DEBUG_PRINT2 ("EXECUTING stop_memory %d:\n", *p); | ||
| 5320 | |||
| 5321 | assert (!REG_UNSET (regstart[*p])); | ||
| 5322 | /* Strictly speaking, there should be code such as: | ||
| 5323 | |||
| 5324 | assert (REG_UNSET (regend[*p])); | ||
| 5325 | PUSH_FAILURE_REGSTOP ((unsigned int)*p); | ||
| 5326 | |||
| 5327 | But the only info to be pushed is regend[*p] and it is known to | ||
| 5328 | be UNSET, so there really isn't anything to push. | ||
| 5329 | Not pushing anything, on the other hand deprives us from the | ||
| 5330 | guarantee that regend[*p] is UNSET since undoing this operation | ||
| 5331 | will not reset its value properly. This is not important since | ||
| 5332 | the value will only be read on the next start_memory or at | ||
| 5333 | the very end and both events can only happen if this stop_memory | ||
| 5334 | is *not* undone. */ | ||
| 5335 | |||
| 5336 | regend[*p] = d; | ||
| 5337 | DEBUG_PRINT2 (" regend: %d\n", POINTER_TO_OFFSET (regend[*p])); | ||
| 5338 | |||
| 5339 | /* Move past the register number and the inner group count. */ | ||
| 5340 | p += 1; | ||
| 5341 | break; | ||
| 5342 | |||
| 5343 | |||
| 5344 | /* \<digit> has been turned into a `duplicate' command which is | ||
| 5345 | followed by the numeric value of <digit> as the register number. */ | ||
| 5346 | case duplicate: | ||
| 5347 | { | ||
| 5348 | register re_char *d2, *dend2; | ||
| 5349 | int regno = *p++; /* Get which register to match against. */ | ||
| 5350 | DEBUG_PRINT2 ("EXECUTING duplicate %d.\n", regno); | ||
| 5351 | |||
| 5352 | /* Can't back reference a group which we've never matched. */ | ||
| 5353 | if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno])) | ||
| 5354 | goto fail; | ||
| 5355 | |||
| 5356 | /* Where in input to try to start matching. */ | ||
| 5357 | d2 = regstart[regno]; | ||
| 5358 | |||
| 5359 | /* Remember the start point to rollback upon failure. */ | ||
| 5360 | dfail = d; | ||
| 5361 | |||
| 5362 | /* Where to stop matching; if both the place to start and | ||
| 5363 | the place to stop matching are in the same string, then | ||
| 5364 | set to the place to stop, otherwise, for now have to use | ||
| 5365 | the end of the first string. */ | ||
| 5366 | |||
| 5367 | dend2 = ((FIRST_STRING_P (regstart[regno]) | ||
| 5368 | == FIRST_STRING_P (regend[regno])) | ||
| 5369 | ? regend[regno] : end_match_1); | ||
| 5370 | for (;;) | ||
| 5371 | { | ||
| 5372 | /* If necessary, advance to next segment in register | ||
| 5373 | contents. */ | ||
| 5374 | while (d2 == dend2) | ||
| 5375 | { | ||
| 5376 | if (dend2 == end_match_2) break; | ||
| 5377 | if (dend2 == regend[regno]) break; | ||
| 5378 | |||
| 5379 | /* End of string1 => advance to string2. */ | ||
| 5380 | d2 = string2; | ||
| 5381 | dend2 = regend[regno]; | ||
| 5382 | } | ||
| 5383 | /* At end of register contents => success */ | ||
| 5384 | if (d2 == dend2) break; | ||
| 5385 | |||
| 5386 | /* If necessary, advance to next segment in data. */ | ||
| 5387 | PREFETCH (); | ||
| 5388 | |||
| 5389 | /* How many characters left in this segment to match. */ | ||
| 5390 | mcnt = dend - d; | ||
| 5391 | |||
| 5392 | /* Want how many consecutive characters we can match in | ||
| 5393 | one shot, so, if necessary, adjust the count. */ | ||
| 5394 | if (mcnt > dend2 - d2) | ||
| 5395 | mcnt = dend2 - d2; | ||
| 5396 | |||
| 5397 | /* Compare that many; failure if mismatch, else move | ||
| 5398 | past them. */ | ||
| 5399 | if (RE_TRANSLATE_P (translate) | ||
| 5400 | ? bcmp_translate (d, d2, mcnt, translate, multibyte) | ||
| 5401 | : memcmp (d, d2, mcnt)) | ||
| 5402 | { | ||
| 5403 | d = dfail; | ||
| 5404 | goto fail; | ||
| 5405 | } | ||
| 5406 | d += mcnt, d2 += mcnt; | ||
| 5407 | } | ||
| 5408 | } | ||
| 5409 | break; | ||
| 5410 | |||
| 5411 | |||
| 5412 | /* begline matches the empty string at the beginning of the string | ||
| 5413 | (unless `not_bol' is set in `bufp'), and after newlines. */ | ||
| 5414 | case begline: | ||
| 5415 | DEBUG_PRINT1 ("EXECUTING begline.\n"); | ||
| 5416 | |||
| 5417 | if (AT_STRINGS_BEG (d)) | ||
| 5418 | { | ||
| 5419 | if (!bufp->not_bol) break; | ||
| 5420 | } | ||
| 5421 | else | ||
| 5422 | { | ||
| 5423 | unsigned char c; | ||
| 5424 | GET_CHAR_BEFORE_2 (c, d, string1, end1, string2, end2); | ||
| 5425 | if (c == '\n') | ||
| 5426 | break; | ||
| 5427 | } | ||
| 5428 | /* In all other cases, we fail. */ | ||
| 5429 | goto fail; | ||
| 5430 | |||
| 5431 | |||
| 5432 | /* endline is the dual of begline. */ | ||
| 5433 | case endline: | ||
| 5434 | DEBUG_PRINT1 ("EXECUTING endline.\n"); | ||
| 5435 | |||
| 5436 | if (AT_STRINGS_END (d)) | ||
| 5437 | { | ||
| 5438 | if (!bufp->not_eol) break; | ||
| 5439 | } | ||
| 5440 | else | ||
| 5441 | { | ||
| 5442 | PREFETCH_NOLIMIT (); | ||
| 5443 | if (*d == '\n') | ||
| 5444 | break; | ||
| 5445 | } | ||
| 5446 | goto fail; | ||
| 5447 | |||
| 5448 | |||
| 5449 | /* Match at the very beginning of the data. */ | ||
| 5450 | case begbuf: | ||
| 5451 | DEBUG_PRINT1 ("EXECUTING begbuf.\n"); | ||
| 5452 | if (AT_STRINGS_BEG (d)) | ||
| 5453 | break; | ||
| 5454 | goto fail; | ||
| 5455 | |||
| 5456 | |||
| 5457 | /* Match at the very end of the data. */ | ||
| 5458 | case endbuf: | ||
| 5459 | DEBUG_PRINT1 ("EXECUTING endbuf.\n"); | ||
| 5460 | if (AT_STRINGS_END (d)) | ||
| 5461 | break; | ||
| 5462 | goto fail; | ||
| 5463 | |||
| 5464 | |||
| 5465 | /* on_failure_keep_string_jump is used to optimize `.*\n'. It | ||
| 5466 | pushes NULL as the value for the string on the stack. Then | ||
| 5467 | `POP_FAILURE_POINT' will keep the current value for the | ||
| 5468 | string, instead of restoring it. To see why, consider | ||
| 5469 | matching `foo\nbar' against `.*\n'. The .* matches the foo; | ||
| 5470 | then the . fails against the \n. But the next thing we want | ||
| 5471 | to do is match the \n against the \n; if we restored the | ||
| 5472 | string value, we would be back at the foo. | ||
| 5473 | |||
| 5474 | Because this is used only in specific cases, we don't need to | ||
| 5475 | check all the things that `on_failure_jump' does, to make | ||
| 5476 | sure the right things get saved on the stack. Hence we don't | ||
| 5477 | share its code. The only reason to push anything on the | ||
| 5478 | stack at all is that otherwise we would have to change | ||
| 5479 | `anychar's code to do something besides goto fail in this | ||
| 5480 | case; that seems worse than this. */ | ||
| 5481 | case on_failure_keep_string_jump: | ||
| 5482 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | ||
| 5483 | DEBUG_PRINT3 ("EXECUTING on_failure_keep_string_jump %d (to %p):\n", | ||
| 5484 | mcnt, p + mcnt); | ||
| 5485 | |||
| 5486 | PUSH_FAILURE_POINT (p - 3, NULL); | ||
| 5487 | break; | ||
| 5488 | |||
| 5489 | /* A nasty loop is introduced by the non-greedy *? and +?. | ||
| 5490 | With such loops, the stack only ever contains one failure point | ||
| 5491 | at a time, so that a plain on_failure_jump_loop kind of | ||
| 5492 | cycle detection cannot work. Worse yet, such a detection | ||
| 5493 | can not only fail to detect a cycle, but it can also wrongly | ||
| 5494 | detect a cycle (between different instantiations of the same | ||
| 5495 | loop). | ||
| 5496 | So the method used for those nasty loops is a little different: | ||
| 5497 | We use a special cycle-detection-stack-frame which is pushed | ||
| 5498 | when the on_failure_jump_nastyloop failure-point is *popped*. | ||
| 5499 | This special frame thus marks the beginning of one iteration | ||
| 5500 | through the loop and we can hence easily check right here | ||
| 5501 | whether something matched between the beginning and the end of | ||
| 5502 | the loop. */ | ||
| 5503 | case on_failure_jump_nastyloop: | ||
| 5504 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | ||
| 5505 | DEBUG_PRINT3 ("EXECUTING on_failure_jump_nastyloop %d (to %p):\n", | ||
| 5506 | mcnt, p + mcnt); | ||
| 5507 | |||
| 5508 | assert ((re_opcode_t)p[-4] == no_op); | ||
| 5509 | { | ||
| 5510 | int cycle = 0; | ||
| 5511 | CHECK_INFINITE_LOOP (p - 4, d); | ||
| 5512 | if (!cycle) | ||
| 5513 | /* If there's a cycle, just continue without pushing | ||
| 5514 | this failure point. The failure point is the "try again" | ||
| 5515 | option, which shouldn't be tried. | ||
| 5516 | We want (x?)*?y\1z to match both xxyz and xxyxz. */ | ||
| 5517 | PUSH_FAILURE_POINT (p - 3, d); | ||
| 5518 | } | ||
| 5519 | break; | ||
| 5520 | |||
| 5521 | /* Simple loop detecting on_failure_jump: just check on the | ||
| 5522 | failure stack if the same spot was already hit earlier. */ | ||
| 5523 | case on_failure_jump_loop: | ||
| 5524 | on_failure: | ||
| 5525 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | ||
| 5526 | DEBUG_PRINT3 ("EXECUTING on_failure_jump_loop %d (to %p):\n", | ||
| 5527 | mcnt, p + mcnt); | ||
| 5528 | { | ||
| 5529 | int cycle = 0; | ||
| 5530 | CHECK_INFINITE_LOOP (p - 3, d); | ||
| 5531 | if (cycle) | ||
| 5532 | /* If there's a cycle, get out of the loop, as if the matching | ||
| 5533 | had failed. We used to just `goto fail' here, but that was | ||
| 5534 | aborting the search a bit too early: we want to keep the | ||
| 5535 | empty-loop-match and keep matching after the loop. | ||
| 5536 | We want (x?)*y\1z to match both xxyz and xxyxz. */ | ||
| 5537 | p += mcnt; | ||
| 5538 | else | ||
| 5539 | PUSH_FAILURE_POINT (p - 3, d); | ||
| 5540 | } | ||
| 5541 | break; | ||
| 5542 | |||
| 5543 | |||
| 5544 | /* Uses of on_failure_jump: | ||
| 5545 | |||
| 5546 | Each alternative starts with an on_failure_jump that points | ||
| 5547 | to the beginning of the next alternative. Each alternative | ||
| 5548 | except the last ends with a jump that in effect jumps past | ||
| 5549 | the rest of the alternatives. (They really jump to the | ||
| 5550 | ending jump of the following alternative, because tensioning | ||
| 5551 | these jumps is a hassle.) | ||
| 5552 | |||
| 5553 | Repeats start with an on_failure_jump that points past both | ||
| 5554 | the repetition text and either the following jump or | ||
| 5555 | pop_failure_jump back to this on_failure_jump. */ | ||
| 5556 | case on_failure_jump: | ||
| 5557 | IMMEDIATE_QUIT_CHECK; | ||
| 5558 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | ||
| 5559 | DEBUG_PRINT3 ("EXECUTING on_failure_jump %d (to %p):\n", | ||
| 5560 | mcnt, p + mcnt); | ||
| 5561 | |||
| 5562 | PUSH_FAILURE_POINT (p -3, d); | ||
| 5563 | break; | ||
| 5564 | |||
| 5565 | /* This operation is used for greedy *. | ||
| 5566 | Compare the beginning of the repeat with what in the | ||
| 5567 | pattern follows its end. If we can establish that there | ||
| 5568 | is nothing that they would both match, i.e., that we | ||
| 5569 | would have to backtrack because of (as in, e.g., `a*a') | ||
| 5570 | then we can use a non-backtracking loop based on | ||
| 5571 | on_failure_keep_string_jump instead of on_failure_jump. */ | ||
| 5572 | case on_failure_jump_smart: | ||
| 5573 | IMMEDIATE_QUIT_CHECK; | ||
| 5574 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | ||
| 5575 | DEBUG_PRINT3 ("EXECUTING on_failure_jump_smart %d (to %p).\n", | ||
| 5576 | mcnt, p + mcnt); | ||
| 5577 | { | ||
| 5578 | re_char *p1 = p; /* Next operation. */ | ||
| 5579 | /* Here, we discard `const', making re_match non-reentrant. */ | ||
| 5580 | unsigned char *p2 = (unsigned char*) p + mcnt; /* Jump dest. */ | ||
| 5581 | unsigned char *p3 = (unsigned char*) p - 3; /* opcode location. */ | ||
| 5582 | |||
| 5583 | p -= 3; /* Reset so that we will re-execute the | ||
| 5584 | instruction once it's been changed. */ | ||
| 5585 | |||
| 5586 | EXTRACT_NUMBER (mcnt, p2 - 2); | ||
| 5587 | |||
| 5588 | /* Ensure this is a indeed the trivial kind of loop | ||
| 5589 | we are expecting. */ | ||
| 5590 | assert (skip_one_char (p1) == p2 - 3); | ||
| 5591 | assert ((re_opcode_t) p2[-3] == jump && p2 + mcnt == p); | ||
| 5592 | DEBUG_STATEMENT (debug += 2); | ||
| 5593 | if (mutually_exclusive_p (bufp, p1, p2)) | ||
| 5594 | { | ||
| 5595 | /* Use a fast `on_failure_keep_string_jump' loop. */ | ||
| 5596 | DEBUG_PRINT1 (" smart exclusive => fast loop.\n"); | ||
| 5597 | *p3 = (unsigned char) on_failure_keep_string_jump; | ||
| 5598 | STORE_NUMBER (p2 - 2, mcnt + 3); | ||
| 5599 | } | ||
| 5600 | else | ||
| 5601 | { | ||
| 5602 | /* Default to a safe `on_failure_jump' loop. */ | ||
| 5603 | DEBUG_PRINT1 (" smart default => slow loop.\n"); | ||
| 5604 | *p3 = (unsigned char) on_failure_jump; | ||
| 5605 | } | ||
| 5606 | DEBUG_STATEMENT (debug -= 2); | ||
| 5607 | } | ||
| 5608 | break; | ||
| 5609 | |||
| 5610 | /* Unconditionally jump (without popping any failure points). */ | ||
| 5611 | case jump: | ||
| 5612 | unconditional_jump: | ||
| 5613 | IMMEDIATE_QUIT_CHECK; | ||
| 5614 | EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */ | ||
| 5615 | DEBUG_PRINT2 ("EXECUTING jump %d ", mcnt); | ||
| 5616 | p += mcnt; /* Do the jump. */ | ||
| 5617 | DEBUG_PRINT2 ("(to %p).\n", p); | ||
| 5618 | break; | ||
| 5619 | |||
| 5620 | |||
| 5621 | /* Have to succeed matching what follows at least n times. | ||
| 5622 | After that, handle like `on_failure_jump'. */ | ||
| 5623 | case succeed_n: | ||
| 5624 | /* Signedness doesn't matter since we only compare MCNT to 0. */ | ||
| 5625 | EXTRACT_NUMBER (mcnt, p + 2); | ||
| 5626 | DEBUG_PRINT2 ("EXECUTING succeed_n %d.\n", mcnt); | ||
| 5627 | |||
| 5628 | /* Originally, mcnt is how many times we HAVE to succeed. */ | ||
| 5629 | if (mcnt != 0) | ||
| 5630 | { | ||
| 5631 | /* Here, we discard `const', making re_match non-reentrant. */ | ||
| 5632 | unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */ | ||
| 5633 | mcnt--; | ||
| 5634 | p += 4; | ||
| 5635 | PUSH_NUMBER (p2, mcnt); | ||
| 5636 | } | ||
| 5637 | else | ||
| 5638 | /* The two bytes encoding mcnt == 0 are two no_op opcodes. */ | ||
| 5639 | goto on_failure; | ||
| 5640 | break; | ||
| 5641 | |||
| 5642 | case jump_n: | ||
| 5643 | /* Signedness doesn't matter since we only compare MCNT to 0. */ | ||
| 5644 | EXTRACT_NUMBER (mcnt, p + 2); | ||
| 5645 | DEBUG_PRINT2 ("EXECUTING jump_n %d.\n", mcnt); | ||
| 5646 | |||
| 5647 | /* Originally, this is how many times we CAN jump. */ | ||
| 5648 | if (mcnt != 0) | ||
| 5649 | { | ||
| 5650 | /* Here, we discard `const', making re_match non-reentrant. */ | ||
| 5651 | unsigned char *p2 = (unsigned char*) p + 2; /* counter loc. */ | ||
| 5652 | mcnt--; | ||
| 5653 | PUSH_NUMBER (p2, mcnt); | ||
| 5654 | goto unconditional_jump; | ||
| 5655 | } | ||
| 5656 | /* If don't have to jump any more, skip over the rest of command. */ | ||
| 5657 | else | ||
| 5658 | p += 4; | ||
| 5659 | break; | ||
| 5660 | |||
| 5661 | case set_number_at: | ||
| 5662 | { | ||
| 5663 | unsigned char *p2; /* Location of the counter. */ | ||
| 5664 | DEBUG_PRINT1 ("EXECUTING set_number_at.\n"); | ||
| 5665 | |||
| 5666 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | ||
| 5667 | /* Here, we discard `const', making re_match non-reentrant. */ | ||
| 5668 | p2 = (unsigned char*) p + mcnt; | ||
| 5669 | /* Signedness doesn't matter since we only copy MCNT's bits . */ | ||
| 5670 | EXTRACT_NUMBER_AND_INCR (mcnt, p); | ||
| 5671 | DEBUG_PRINT3 (" Setting %p to %d.\n", p2, mcnt); | ||
| 5672 | PUSH_NUMBER (p2, mcnt); | ||
| 5673 | break; | ||
| 5674 | } | ||
| 5675 | |||
| 5676 | case wordbound: | ||
| 5677 | case notwordbound: | ||
| 5678 | not = (re_opcode_t) *(p - 1) == notwordbound; | ||
| 5679 | DEBUG_PRINT2 ("EXECUTING %swordbound.\n", not?"not":""); | ||
| 5680 | |||
| 5681 | /* We SUCCEED (or FAIL) in one of the following cases: */ | ||
| 5682 | |||
| 5683 | /* Case 1: D is at the beginning or the end of string. */ | ||
| 5684 | if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d)) | ||
| 5685 | not = !not; | ||
| 5686 | else | ||
| 5687 | { | ||
| 5688 | /* C1 is the character before D, S1 is the syntax of C1, C2 | ||
| 5689 | is the character at D, and S2 is the syntax of C2. */ | ||
| 5690 | re_wchar_t c1, c2; | ||
| 5691 | int s1, s2; | ||
| 5692 | #ifdef emacs | ||
| 5693 | int offset = PTR_TO_OFFSET (d - 1); | ||
| 5694 | int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | ||
| 5695 | UPDATE_SYNTAX_TABLE (charpos); | ||
| 5696 | #endif | ||
| 5697 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | ||
| 5698 | s1 = SYNTAX (c1); | ||
| 5699 | #ifdef emacs | ||
| 5700 | UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1); | ||
| 5701 | #endif | ||
| 5702 | PREFETCH_NOLIMIT (); | ||
| 5703 | c2 = RE_STRING_CHAR (d, dend - d); | ||
| 5704 | s2 = SYNTAX (c2); | ||
| 5705 | |||
| 5706 | if (/* Case 2: Only one of S1 and S2 is Sword. */ | ||
| 5707 | ((s1 == Sword) != (s2 == Sword)) | ||
| 5708 | /* Case 3: Both of S1 and S2 are Sword, and macro | ||
| 5709 | WORD_BOUNDARY_P (C1, C2) returns nonzero. */ | ||
| 5710 | || ((s1 == Sword) && WORD_BOUNDARY_P (c1, c2))) | ||
| 5711 | not = !not; | ||
| 5712 | } | ||
| 5713 | if (not) | ||
| 5714 | break; | ||
| 5715 | else | ||
| 5716 | goto fail; | ||
| 5717 | |||
| 5718 | case wordbeg: | ||
| 5719 | DEBUG_PRINT1 ("EXECUTING wordbeg.\n"); | ||
| 5720 | |||
| 5721 | /* We FAIL in one of the following cases: */ | ||
| 5722 | |||
| 5723 | /* Case 1: D is at the end of string. */ | ||
| 5724 | if (AT_STRINGS_END (d)) | ||
| 5725 | goto fail; | ||
| 5726 | else | ||
| 5727 | { | ||
| 5728 | /* C1 is the character before D, S1 is the syntax of C1, C2 | ||
| 5729 | is the character at D, and S2 is the syntax of C2. */ | ||
| 5730 | re_wchar_t c1, c2; | ||
| 5731 | int s1, s2; | ||
| 5732 | #ifdef emacs | ||
| 5733 | int offset = PTR_TO_OFFSET (d); | ||
| 5734 | int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | ||
| 5735 | UPDATE_SYNTAX_TABLE (charpos); | ||
| 5736 | #endif | ||
| 5737 | PREFETCH (); | ||
| 5738 | c2 = RE_STRING_CHAR (d, dend - d); | ||
| 5739 | s2 = SYNTAX (c2); | ||
| 5740 | |||
| 5741 | /* Case 2: S2 is not Sword. */ | ||
| 5742 | if (s2 != Sword) | ||
| 5743 | goto fail; | ||
| 5744 | |||
| 5745 | /* Case 3: D is not at the beginning of string ... */ | ||
| 5746 | if (!AT_STRINGS_BEG (d)) | ||
| 5747 | { | ||
| 5748 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | ||
| 5749 | #ifdef emacs | ||
| 5750 | UPDATE_SYNTAX_TABLE_BACKWARD (charpos - 1); | ||
| 5751 | #endif | ||
| 5752 | s1 = SYNTAX (c1); | ||
| 5753 | |||
| 5754 | /* ... and S1 is Sword, and WORD_BOUNDARY_P (C1, C2) | ||
| 5755 | returns 0. */ | ||
| 5756 | if ((s1 == Sword) && !WORD_BOUNDARY_P (c1, c2)) | ||
| 5757 | goto fail; | ||
| 5758 | } | ||
| 5759 | } | ||
| 5760 | break; | ||
| 5761 | |||
| 5762 | case wordend: | ||
| 5763 | DEBUG_PRINT1 ("EXECUTING wordend.\n"); | ||
| 5764 | |||
| 5765 | /* We FAIL in one of the following cases: */ | ||
| 5766 | |||
| 5767 | /* Case 1: D is at the beginning of string. */ | ||
| 5768 | if (AT_STRINGS_BEG (d)) | ||
| 5769 | goto fail; | ||
| 5770 | else | ||
| 5771 | { | ||
| 5772 | /* C1 is the character before D, S1 is the syntax of C1, C2 | ||
| 5773 | is the character at D, and S2 is the syntax of C2. */ | ||
| 5774 | re_wchar_t c1, c2; | ||
| 5775 | int s1, s2; | ||
| 5776 | #ifdef emacs | ||
| 5777 | int offset = PTR_TO_OFFSET (d) - 1; | ||
| 5778 | int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | ||
| 5779 | UPDATE_SYNTAX_TABLE (charpos); | ||
| 5780 | #endif | ||
| 5781 | GET_CHAR_BEFORE_2 (c1, d, string1, end1, string2, end2); | ||
| 5782 | s1 = SYNTAX (c1); | ||
| 5783 | |||
| 5784 | /* Case 2: S1 is not Sword. */ | ||
| 5785 | if (s1 != Sword) | ||
| 5786 | goto fail; | ||
| 5787 | |||
| 5788 | /* Case 3: D is not at the end of string ... */ | ||
| 5789 | if (!AT_STRINGS_END (d)) | ||
| 5790 | { | ||
| 5791 | PREFETCH_NOLIMIT (); | ||
| 5792 | c2 = RE_STRING_CHAR (d, dend - d); | ||
| 5793 | #ifdef emacs | ||
| 5794 | UPDATE_SYNTAX_TABLE_FORWARD (charpos); | ||
| 5795 | #endif | ||
| 5796 | s2 = SYNTAX (c2); | ||
| 5797 | |||
| 5798 | /* ... and S2 is Sword, and WORD_BOUNDARY_P (C1, C2) | ||
| 5799 | returns 0. */ | ||
| 5800 | if ((s2 == Sword) && !WORD_BOUNDARY_P (c1, c2)) | ||
| 5801 | goto fail; | ||
| 5802 | } | ||
| 5803 | } | ||
| 5804 | break; | ||
| 5805 | |||
| 5806 | case syntaxspec: | ||
| 5807 | case notsyntaxspec: | ||
| 5808 | not = (re_opcode_t) *(p - 1) == notsyntaxspec; | ||
| 5809 | mcnt = *p++; | ||
| 5810 | DEBUG_PRINT3 ("EXECUTING %ssyntaxspec %d.\n", not?"not":"", mcnt); | ||
| 5811 | PREFETCH (); | ||
| 5812 | #ifdef emacs | ||
| 5813 | { | ||
| 5814 | int offset = PTR_TO_OFFSET (d); | ||
| 5815 | int pos1 = SYNTAX_TABLE_BYTE_TO_CHAR (offset); | ||
| 5816 | UPDATE_SYNTAX_TABLE (pos1); | ||
| 5817 | } | ||
| 5818 | #endif | ||
| 5819 | { | ||
| 5820 | int len; | ||
| 5821 | re_wchar_t c; | ||
| 5822 | |||
| 5823 | c = RE_STRING_CHAR_AND_LENGTH (d, dend - d, len); | ||
| 5824 | |||
| 5825 | if ((SYNTAX (c) != (enum syntaxcode) mcnt) ^ not) | ||
| 5826 | goto fail; | ||
| 5827 | d += len; | ||
| 5828 | } | ||
| 5829 | break; | ||
| 5830 | |||
| 5831 | #ifdef emacs | ||
| 5832 | case before_dot: | ||
| 5833 | DEBUG_PRINT1 ("EXECUTING before_dot.\n"); | ||
| 5834 | if (PTR_BYTE_POS (d) >= PT_BYTE) | ||
| 5835 | goto fail; | ||
| 5836 | break; | ||
| 5837 | |||
| 5838 | case at_dot: | ||
| 5839 | DEBUG_PRINT1 ("EXECUTING at_dot.\n"); | ||
| 5840 | if (PTR_BYTE_POS (d) != PT_BYTE) | ||
| 5841 | goto fail; | ||
| 5842 | break; | ||
| 5843 | |||
| 5844 | case after_dot: | ||
| 5845 | DEBUG_PRINT1 ("EXECUTING after_dot.\n"); | ||
| 5846 | if (PTR_BYTE_POS (d) <= PT_BYTE) | ||
| 5847 | goto fail; | ||
| 5848 | break; | ||
| 5849 | |||
| 5850 | case categoryspec: | ||
| 5851 | case notcategoryspec: | ||
| 5852 | not = (re_opcode_t) *(p - 1) == notcategoryspec; | ||
| 5853 | mcnt = *p++; | ||
| 5854 | DEBUG_PRINT3 ("EXECUTING %scategoryspec %d.\n", not?"not":"", mcnt); | ||
| 5855 | PREFETCH (); | ||
| 5856 | { | ||
| 5857 | int len; | ||
| 5858 | re_wchar_t c; | ||
| 5859 | |||
| 5860 | c = RE_STRING_CHAR_AND_LENGTH (d, dend - d, len); | ||
| 5861 | |||
| 5862 | if ((!CHAR_HAS_CATEGORY (c, mcnt)) ^ not) | ||
| 5863 | goto fail; | ||
| 5864 | d += len; | ||
| 5865 | } | ||
| 5866 | break; | ||
| 5867 | |||
| 5868 | #endif /* emacs */ | ||
| 5869 | |||
| 5870 | default: | ||
| 5871 | abort (); | ||
| 5872 | } | ||
| 5873 | continue; /* Successfully executed one pattern command; keep going. */ | ||
| 5874 | |||
| 5875 | |||
| 5876 | /* We goto here if a matching operation fails. */ | ||
| 5877 | fail: | ||
| 5878 | IMMEDIATE_QUIT_CHECK; | ||
| 5879 | if (!FAIL_STACK_EMPTY ()) | ||
| 5880 | { | ||
| 5881 | re_char *str, *pat; | ||
| 5882 | /* A restart point is known. Restore to that state. */ | ||
| 5883 | DEBUG_PRINT1 ("\nFAIL:\n"); | ||
| 5884 | POP_FAILURE_POINT (str, pat); | ||
| 5885 | switch (SWITCH_ENUM_CAST ((re_opcode_t) *pat++)) | ||
| 5886 | { | ||
| 5887 | case on_failure_keep_string_jump: | ||
| 5888 | assert (str == NULL); | ||
| 5889 | goto continue_failure_jump; | ||
| 5890 | |||
| 5891 | case on_failure_jump_nastyloop: | ||
| 5892 | assert ((re_opcode_t)pat[-2] == no_op); | ||
| 5893 | PUSH_FAILURE_POINT (pat - 2, str); | ||
| 5894 | /* Fallthrough */ | ||
| 5895 | |||
| 5896 | case on_failure_jump_loop: | ||
| 5897 | case on_failure_jump: | ||
| 5898 | case succeed_n: | ||
| 5899 | d = str; | ||
| 5900 | continue_failure_jump: | ||
| 5901 | EXTRACT_NUMBER_AND_INCR (mcnt, pat); | ||
| 5902 | p = pat + mcnt; | ||
| 5903 | break; | ||
| 5904 | |||
| 5905 | case no_op: | ||
| 5906 | /* A special frame used for nastyloops. */ | ||
| 5907 | goto fail; | ||
| 5908 | |||
| 5909 | default: | ||
| 5910 | abort(); | ||
| 5911 | } | ||
| 5912 | |||
| 5913 | assert (p >= bufp->buffer && p <= pend); | ||
| 5914 | |||
| 5915 | if (d >= string1 && d <= end1) | ||
| 5916 | dend = end_match_1; | ||
| 5917 | } | ||
| 5918 | else | ||
| 5919 | break; /* Matching at this starting point really fails. */ | ||
| 5920 | } /* for (;;) */ | ||
| 5921 | |||
| 5922 | if (best_regs_set) | ||
| 5923 | goto restore_best_regs; | ||
| 5924 | |||
| 5925 | FREE_VARIABLES (); | ||
| 5926 | |||
| 5927 | return -1; /* Failure to match. */ | ||
| 5928 | } /* re_match_2 */ | ||
| 5929 | |||
| 5930 | /* Subroutine definitions for re_match_2. */ | ||
| 5931 | |||
| 5932 | /* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN | ||
| 5933 | bytes; nonzero otherwise. */ | ||
| 5934 | |||
| 5935 | static int | ||
| 5936 | bcmp_translate (s1, s2, len, translate, multibyte) | ||
| 5937 | re_char *s1, *s2; | ||
| 5938 | register int len; | ||
| 5939 | RE_TRANSLATE_TYPE translate; | ||
| 5940 | const int multibyte; | ||
| 5941 | { | ||
| 5942 | register re_char *p1 = s1, *p2 = s2; | ||
| 5943 | re_char *p1_end = s1 + len; | ||
| 5944 | re_char *p2_end = s2 + len; | ||
| 5945 | |||
| 5946 | /* FIXME: Checking both p1 and p2 presumes that the two strings might have | ||
| 5947 | different lengths, but relying on a single `len' would break this. -sm */ | ||
| 5948 | while (p1 < p1_end && p2 < p2_end) | ||
| 5949 | { | ||
| 5950 | int p1_charlen, p2_charlen; | ||
| 5951 | re_wchar_t p1_ch, p2_ch; | ||
| 5952 | |||
| 5953 | p1_ch = RE_STRING_CHAR_AND_LENGTH (p1, p1_end - p1, p1_charlen); | ||
| 5954 | p2_ch = RE_STRING_CHAR_AND_LENGTH (p2, p2_end - p2, p2_charlen); | ||
| 5955 | |||
| 5956 | if (RE_TRANSLATE (translate, p1_ch) | ||
| 5957 | != RE_TRANSLATE (translate, p2_ch)) | ||
| 5958 | return 1; | ||
| 5959 | |||
| 5960 | p1 += p1_charlen, p2 += p2_charlen; | ||
| 5961 | } | ||
| 5962 | |||
| 5963 | if (p1 != p1_end || p2 != p2_end) | ||
| 5964 | return 1; | ||
| 5965 | |||
| 5966 | return 0; | ||
| 5967 | } | ||
| 5968 | |||
| 5969 | /* Entry points for GNU code. */ | ||
| 5970 | |||
| 5971 | /* re_compile_pattern is the GNU regular expression compiler: it | ||
| 5972 | compiles PATTERN (of length SIZE) and puts the result in BUFP. | ||
| 5973 | Returns 0 if the pattern was valid, otherwise an error string. | ||
| 5974 | |||
| 5975 | Assumes the `allocated' (and perhaps `buffer') and `translate' fields | ||
| 5976 | are set in BUFP on entry. | ||
| 5977 | |||
| 5978 | We call regex_compile to do the actual compilation. */ | ||
| 5979 | |||
| 5980 | const char * | ||
| 5981 | re_compile_pattern (pattern, length, bufp) | ||
| 5982 | const char *pattern; | ||
| 5983 | size_t length; | ||
| 5984 | struct re_pattern_buffer *bufp; | ||
| 5985 | { | ||
| 5986 | reg_errcode_t ret; | ||
| 5987 | |||
| 5988 | /* GNU code is written to assume at least RE_NREGS registers will be set | ||
| 5989 | (and at least one extra will be -1). */ | ||
| 5990 | bufp->regs_allocated = REGS_UNALLOCATED; | ||
| 5991 | |||
| 5992 | /* And GNU code determines whether or not to get register information | ||
| 5993 | by passing null for the REGS argument to re_match, etc., not by | ||
| 5994 | setting no_sub. */ | ||
| 5995 | bufp->no_sub = 0; | ||
| 5996 | |||
| 5997 | ret = regex_compile ((re_char*) pattern, length, re_syntax_options, bufp); | ||
| 5998 | |||
| 5999 | if (!ret) | ||
| 6000 | return NULL; | ||
| 6001 | return gettext (re_error_msgid[(int) ret]); | ||
| 6002 | } | ||
| 6003 | WEAK_ALIAS (__re_compile_pattern, re_compile_pattern) | ||
| 6004 | |||
| 6005 | /* Entry points compatible with 4.2 BSD regex library. We don't define | ||
| 6006 | them unless specifically requested. */ | ||
| 6007 | |||
| 6008 | #if defined _REGEX_RE_COMP || defined _LIBC | ||
| 6009 | |||
| 6010 | /* BSD has one and only one pattern buffer. */ | ||
| 6011 | static struct re_pattern_buffer re_comp_buf; | ||
| 6012 | |||
| 6013 | char * | ||
| 6014 | # ifdef _LIBC | ||
| 6015 | /* Make these definitions weak in libc, so POSIX programs can redefine | ||
| 6016 | these names if they don't use our functions, and still use | ||
| 6017 | regcomp/regexec below without link errors. */ | ||
| 6018 | weak_function | ||
| 6019 | # endif | ||
| 6020 | re_comp (s) | ||
| 6021 | const char *s; | ||
| 6022 | { | ||
| 6023 | reg_errcode_t ret; | ||
| 6024 | |||
| 6025 | if (!s) | ||
| 6026 | { | ||
| 6027 | if (!re_comp_buf.buffer) | ||
| 6028 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ | ||
| 6029 | return (char *) gettext ("No previous regular expression"); | ||
| 6030 | return 0; | ||
| 6031 | } | ||
| 6032 | |||
| 6033 | if (!re_comp_buf.buffer) | ||
| 6034 | { | ||
| 6035 | re_comp_buf.buffer = (unsigned char *) malloc (200); | ||
| 6036 | if (re_comp_buf.buffer == NULL) | ||
| 6037 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ | ||
| 6038 | return (char *) gettext (re_error_msgid[(int) REG_ESPACE]); | ||
| 6039 | re_comp_buf.allocated = 200; | ||
| 6040 | |||
| 6041 | re_comp_buf.fastmap = (char *) malloc (1 << BYTEWIDTH); | ||
| 6042 | if (re_comp_buf.fastmap == NULL) | ||
| 6043 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ | ||
| 6044 | return (char *) gettext (re_error_msgid[(int) REG_ESPACE]); | ||
| 6045 | } | ||
| 6046 | |||
| 6047 | /* Since `re_exec' always passes NULL for the `regs' argument, we | ||
| 6048 | don't need to initialize the pattern buffer fields which affect it. */ | ||
| 6049 | |||
| 6050 | ret = regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf); | ||
| 6051 | |||
| 6052 | if (!ret) | ||
| 6053 | return NULL; | ||
| 6054 | |||
| 6055 | /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ | ||
| 6056 | return (char *) gettext (re_error_msgid[(int) ret]); | ||
| 6057 | } | ||
| 6058 | |||
| 6059 | |||
| 6060 | int | ||
| 6061 | # ifdef _LIBC | ||
| 6062 | weak_function | ||
| 6063 | # endif | ||
| 6064 | re_exec (s) | ||
| 6065 | const char *s; | ||
| 6066 | { | ||
| 6067 | const int len = strlen (s); | ||
| 6068 | return | ||
| 6069 | 0 <= re_search (&re_comp_buf, s, len, 0, len, (struct re_registers *) 0); | ||
| 6070 | } | ||
| 6071 | #endif /* _REGEX_RE_COMP */ | ||
| 6072 | |||
| 6073 | /* POSIX.2 functions. Don't define these for Emacs. */ | ||
| 6074 | |||
| 6075 | #ifndef emacs | ||
| 6076 | |||
| 6077 | /* regcomp takes a regular expression as a string and compiles it. | ||
| 6078 | |||
| 6079 | PREG is a regex_t *. We do not expect any fields to be initialized, | ||
| 6080 | since POSIX says we shouldn't. Thus, we set | ||
| 6081 | |||
| 6082 | `buffer' to the compiled pattern; | ||
| 6083 | `used' to the length of the compiled pattern; | ||
| 6084 | `syntax' to RE_SYNTAX_POSIX_EXTENDED if the | ||
| 6085 | REG_EXTENDED bit in CFLAGS is set; otherwise, to | ||
| 6086 | RE_SYNTAX_POSIX_BASIC; | ||
| 6087 | `fastmap' to an allocated space for the fastmap; | ||
| 6088 | `fastmap_accurate' to zero; | ||
| 6089 | `re_nsub' to the number of subexpressions in PATTERN. | ||
| 6090 | |||
| 6091 | PATTERN is the address of the pattern string. | ||
| 6092 | |||
| 6093 | CFLAGS is a series of bits which affect compilation. | ||
| 6094 | |||
| 6095 | If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we | ||
| 6096 | use POSIX basic syntax. | ||
| 6097 | |||
| 6098 | If REG_NEWLINE is set, then . and [^...] don't match newline. | ||
| 6099 | Also, regexec will try a match beginning after every newline. | ||
| 6100 | |||
| 6101 | If REG_ICASE is set, then we considers upper- and lowercase | ||
| 6102 | versions of letters to be equivalent when matching. | ||
| 6103 | |||
| 6104 | If REG_NOSUB is set, then when PREG is passed to regexec, that | ||
| 6105 | routine will report only success or failure, and nothing about the | ||
| 6106 | registers. | ||
| 6107 | |||
| 6108 | It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for | ||
| 6109 | the return codes and their meanings.) */ | ||
| 6110 | |||
| 6111 | int | ||
| 6112 | regcomp (preg, pattern, cflags) | ||
| 6113 | regex_t *__restrict preg; | ||
| 6114 | const char *__restrict pattern; | ||
| 6115 | int cflags; | ||
| 6116 | { | ||
| 6117 | reg_errcode_t ret; | ||
| 6118 | reg_syntax_t syntax | ||
| 6119 | = (cflags & REG_EXTENDED) ? | ||
| 6120 | RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC; | ||
| 6121 | |||
| 6122 | /* regex_compile will allocate the space for the compiled pattern. */ | ||
| 6123 | preg->buffer = 0; | ||
| 6124 | preg->allocated = 0; | ||
| 6125 | preg->used = 0; | ||
| 6126 | |||
| 6127 | /* Try to allocate space for the fastmap. */ | ||
| 6128 | preg->fastmap = (char *) malloc (1 << BYTEWIDTH); | ||
| 6129 | |||
| 6130 | if (cflags & REG_ICASE) | ||
| 6131 | { | ||
| 6132 | unsigned i; | ||
| 6133 | |||
| 6134 | preg->translate | ||
| 6135 | = (RE_TRANSLATE_TYPE) malloc (CHAR_SET_SIZE | ||
| 6136 | * sizeof (*(RE_TRANSLATE_TYPE)0)); | ||
| 6137 | if (preg->translate == NULL) | ||
| 6138 | return (int) REG_ESPACE; | ||
| 6139 | |||
| 6140 | /* Map uppercase characters to corresponding lowercase ones. */ | ||
| 6141 | for (i = 0; i < CHAR_SET_SIZE; i++) | ||
| 6142 | preg->translate[i] = ISUPPER (i) ? TOLOWER (i) : i; | ||
| 6143 | } | ||
| 6144 | else | ||
| 6145 | preg->translate = NULL; | ||
| 6146 | |||
| 6147 | /* If REG_NEWLINE is set, newlines are treated differently. */ | ||
| 6148 | if (cflags & REG_NEWLINE) | ||
| 6149 | { /* REG_NEWLINE implies neither . nor [^...] match newline. */ | ||
| 6150 | syntax &= ~RE_DOT_NEWLINE; | ||
| 6151 | syntax |= RE_HAT_LISTS_NOT_NEWLINE; | ||
| 6152 | } | ||
| 6153 | else | ||
| 6154 | syntax |= RE_NO_NEWLINE_ANCHOR; | ||
| 6155 | |||
| 6156 | preg->no_sub = !!(cflags & REG_NOSUB); | ||
| 6157 | |||
| 6158 | /* POSIX says a null character in the pattern terminates it, so we | ||
| 6159 | can use strlen here in compiling the pattern. */ | ||
| 6160 | ret = regex_compile ((re_char*) pattern, strlen (pattern), syntax, preg); | ||
| 6161 | |||
| 6162 | /* POSIX doesn't distinguish between an unmatched open-group and an | ||
| 6163 | unmatched close-group: both are REG_EPAREN. */ | ||
| 6164 | if (ret == REG_ERPAREN) | ||
| 6165 | ret = REG_EPAREN; | ||
| 6166 | |||
| 6167 | if (ret == REG_NOERROR && preg->fastmap) | ||
| 6168 | { /* Compute the fastmap now, since regexec cannot modify the pattern | ||
| 6169 | buffer. */ | ||
| 6170 | re_compile_fastmap (preg); | ||
| 6171 | if (preg->can_be_null) | ||
| 6172 | { /* The fastmap can't be used anyway. */ | ||
| 6173 | free (preg->fastmap); | ||
| 6174 | preg->fastmap = NULL; | ||
| 6175 | } | ||
| 6176 | } | ||
| 6177 | return (int) ret; | ||
| 6178 | } | ||
| 6179 | WEAK_ALIAS (__regcomp, regcomp) | ||
| 6180 | |||
| 6181 | |||
| 6182 | /* regexec searches for a given pattern, specified by PREG, in the | ||
| 6183 | string STRING. | ||
| 6184 | |||
| 6185 | If NMATCH is zero or REG_NOSUB was set in the cflags argument to | ||
| 6186 | `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at | ||
| 6187 | least NMATCH elements, and we set them to the offsets of the | ||
| 6188 | corresponding matched substrings. | ||
| 6189 | |||
| 6190 | EFLAGS specifies `execution flags' which affect matching: if | ||
| 6191 | REG_NOTBOL is set, then ^ does not match at the beginning of the | ||
| 6192 | string; if REG_NOTEOL is set, then $ does not match at the end. | ||
| 6193 | |||
| 6194 | We return 0 if we find a match and REG_NOMATCH if not. */ | ||
| 6195 | |||
| 6196 | int | ||
| 6197 | regexec (preg, string, nmatch, pmatch, eflags) | ||
| 6198 | const regex_t *__restrict preg; | ||
| 6199 | const char *__restrict string; | ||
| 6200 | size_t nmatch; | ||
| 6201 | regmatch_t pmatch[__restrict_arr]; | ||
| 6202 | int eflags; | ||
| 6203 | { | ||
| 6204 | int ret; | ||
| 6205 | struct re_registers regs; | ||
| 6206 | regex_t private_preg; | ||
| 6207 | int len = strlen (string); | ||
| 6208 | boolean want_reg_info = !preg->no_sub && nmatch > 0 && pmatch; | ||
| 6209 | |||
| 6210 | private_preg = *preg; | ||
| 6211 | |||
| 6212 | private_preg.not_bol = !!(eflags & REG_NOTBOL); | ||
| 6213 | private_preg.not_eol = !!(eflags & REG_NOTEOL); | ||
| 6214 | |||
| 6215 | /* The user has told us exactly how many registers to return | ||
| 6216 | information about, via `nmatch'. We have to pass that on to the | ||
| 6217 | matching routines. */ | ||
| 6218 | private_preg.regs_allocated = REGS_FIXED; | ||
| 6219 | |||
| 6220 | if (want_reg_info) | ||
| 6221 | { | ||
| 6222 | regs.num_regs = nmatch; | ||
| 6223 | regs.start = TALLOC (nmatch * 2, regoff_t); | ||
| 6224 | if (regs.start == NULL) | ||
| 6225 | return (int) REG_NOMATCH; | ||
| 6226 | regs.end = regs.start + nmatch; | ||
| 6227 | } | ||
| 6228 | |||
| 6229 | /* Instead of using not_eol to implement REG_NOTEOL, we could simply | ||
| 6230 | pass (&private_preg, string, len + 1, 0, len, ...) pretending the string | ||
| 6231 | was a little bit longer but still only matching the real part. | ||
| 6232 | This works because the `endline' will check for a '\n' and will find a | ||
| 6233 | '\0', correctly deciding that this is not the end of a line. | ||
| 6234 | But it doesn't work out so nicely for REG_NOTBOL, since we don't have | ||
| 6235 | a convenient '\0' there. For all we know, the string could be preceded | ||
| 6236 | by '\n' which would throw things off. */ | ||
| 6237 | |||
| 6238 | /* Perform the searching operation. */ | ||
| 6239 | ret = re_search (&private_preg, string, len, | ||
| 6240 | /* start: */ 0, /* range: */ len, | ||
| 6241 | want_reg_info ? ®s : (struct re_registers *) 0); | ||
| 6242 | |||
| 6243 | /* Copy the register information to the POSIX structure. */ | ||
| 6244 | if (want_reg_info) | ||
| 6245 | { | ||
| 6246 | if (ret >= 0) | ||
| 6247 | { | ||
| 6248 | unsigned r; | ||
| 6249 | |||
| 6250 | for (r = 0; r < nmatch; r++) | ||
| 6251 | { | ||
| 6252 | pmatch[r].rm_so = regs.start[r]; | ||
| 6253 | pmatch[r].rm_eo = regs.end[r]; | ||
| 6254 | } | ||
| 6255 | } | ||
| 6256 | |||
| 6257 | /* If we needed the temporary register info, free the space now. */ | ||
| 6258 | free (regs.start); | ||
| 6259 | } | ||
| 6260 | |||
| 6261 | /* We want zero return to mean success, unlike `re_search'. */ | ||
| 6262 | return ret >= 0 ? (int) REG_NOERROR : (int) REG_NOMATCH; | ||
| 6263 | } | ||
| 6264 | WEAK_ALIAS (__regexec, regexec) | ||
| 6265 | |||
| 6266 | |||
| 6267 | /* Returns a message corresponding to an error code, ERRCODE, returned | ||
| 6268 | from either regcomp or regexec. We don't use PREG here. */ | ||
| 6269 | |||
| 6270 | size_t | ||
| 6271 | regerror (errcode, preg, errbuf, errbuf_size) | ||
| 6272 | int errcode; | ||
| 6273 | const regex_t *preg; | ||
| 6274 | char *errbuf; | ||
| 6275 | size_t errbuf_size; | ||
| 6276 | { | ||
| 6277 | const char *msg; | ||
| 6278 | size_t msg_size; | ||
| 6279 | |||
| 6280 | if (errcode < 0 | ||
| 6281 | || errcode >= (sizeof (re_error_msgid) / sizeof (re_error_msgid[0]))) | ||
| 6282 | /* Only error codes returned by the rest of the code should be passed | ||
| 6283 | to this routine. If we are given anything else, or if other regex | ||
| 6284 | code generates an invalid error code, then the program has a bug. | ||
| 6285 | Dump core so we can fix it. */ | ||
| 6286 | abort (); | ||
| 6287 | |||
| 6288 | msg = gettext (re_error_msgid[errcode]); | ||
| 6289 | |||
| 6290 | msg_size = strlen (msg) + 1; /* Includes the null. */ | ||
| 6291 | |||
| 6292 | if (errbuf_size != 0) | ||
| 6293 | { | ||
| 6294 | if (msg_size > errbuf_size) | ||
| 6295 | { | ||
| 6296 | strncpy (errbuf, msg, errbuf_size - 1); | ||
| 6297 | errbuf[errbuf_size - 1] = 0; | ||
| 6298 | } | ||
| 6299 | else | ||
| 6300 | strcpy (errbuf, msg); | ||
| 6301 | } | ||
| 6302 | |||
| 6303 | return msg_size; | ||
| 6304 | } | ||
| 6305 | WEAK_ALIAS (__regerror, regerror) | ||
| 6306 | |||
| 6307 | |||
| 6308 | /* Free dynamically allocated space used by PREG. */ | ||
| 6309 | |||
| 6310 | void | ||
| 6311 | regfree (preg) | ||
| 6312 | regex_t *preg; | ||
| 6313 | { | ||
| 6314 | if (preg->buffer != NULL) | ||
| 6315 | free (preg->buffer); | ||
| 6316 | preg->buffer = NULL; | ||
| 6317 | |||
| 6318 | preg->allocated = 0; | ||
| 6319 | preg->used = 0; | ||
| 6320 | |||
| 6321 | if (preg->fastmap != NULL) | ||
| 6322 | free (preg->fastmap); | ||
| 6323 | preg->fastmap = NULL; | ||
| 6324 | preg->fastmap_accurate = 0; | ||
| 6325 | |||
| 6326 | if (preg->translate != NULL) | ||
| 6327 | free (preg->translate); | ||
| 6328 | preg->translate = NULL; | ||
| 6329 | } | ||
| 6330 | WEAK_ALIAS (__regfree, regfree) | ||
| 6331 | |||
| 6332 | #endif /* not emacs */ | ||
| 6333 | |||
| 6334 | /* arch-tag: 4ffd68ba-2a9e-435b-a21a-018990f9eeb2 | ||
| 6335 | (do not change this comment) */ | ||
diff --git a/src/regex.h b/src/regex.h deleted file mode 100644 index 1818d5f9681..00000000000 --- a/src/regex.h +++ /dev/null | |||
| @@ -1,576 +0,0 @@ | |||
| 1 | /* Definitions for data structures and routines for the regular | ||
| 2 | expression library, version 0.12. | ||
| 3 | |||
| 4 | Copyright (C) 1985,89,90,91,92,93,95,2000 Free Software Foundation, Inc. | ||
| 5 | |||
| 6 | This program is free software; you can redistribute it and/or modify | ||
| 7 | it under the terms of the GNU General Public License as published by | ||
| 8 | the Free Software Foundation; either version 2, or (at your option) | ||
| 9 | any later version. | ||
| 10 | |||
| 11 | This program is distributed in the hope that it will be useful, | ||
| 12 | but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | ||
| 14 | GNU General Public License for more details. | ||
| 15 | |||
| 16 | You should have received a copy of the GNU General Public License | ||
| 17 | along with this program; if not, write to the Free Software | ||
| 18 | Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, | ||
| 19 | USA. */ | ||
| 20 | |||
| 21 | #ifndef _REGEX_H | ||
| 22 | #define _REGEX_H 1 | ||
| 23 | |||
| 24 | /* Allow the use in C++ code. */ | ||
| 25 | #ifdef __cplusplus | ||
| 26 | extern "C" { | ||
| 27 | #endif | ||
| 28 | |||
| 29 | /* POSIX says that <sys/types.h> must be included (by the caller) before | ||
| 30 | <regex.h>. */ | ||
| 31 | |||
| 32 | #if !defined _POSIX_C_SOURCE && !defined _POSIX_SOURCE && defined VMS | ||
| 33 | /* VMS doesn't have `size_t' in <sys/types.h>, even though POSIX says it | ||
| 34 | should be there. */ | ||
| 35 | # include <stddef.h> | ||
| 36 | #endif | ||
| 37 | |||
| 38 | /* The following bits are used to determine the regexp syntax we | ||
| 39 | recognize. The set/not-set meanings where historically chosen so | ||
| 40 | that Emacs syntax had the value 0. | ||
| 41 | The bits are given in alphabetical order, and | ||
| 42 | the definitions shifted by one from the previous bit; thus, when we | ||
| 43 | add or remove a bit, only one other definition need change. */ | ||
| 44 | typedef unsigned long int reg_syntax_t; | ||
| 45 | |||
| 46 | /* If this bit is not set, then \ inside a bracket expression is literal. | ||
| 47 | If set, then such a \ quotes the following character. */ | ||
| 48 | #define RE_BACKSLASH_ESCAPE_IN_LISTS ((unsigned long int) 1) | ||
| 49 | |||
| 50 | /* If this bit is not set, then + and ? are operators, and \+ and \? are | ||
| 51 | literals. | ||
| 52 | If set, then \+ and \? are operators and + and ? are literals. */ | ||
| 53 | #define RE_BK_PLUS_QM (RE_BACKSLASH_ESCAPE_IN_LISTS << 1) | ||
| 54 | |||
| 55 | /* If this bit is set, then character classes are supported. They are: | ||
| 56 | [:alpha:], [:upper:], [:lower:], [:digit:], [:alnum:], [:xdigit:], | ||
| 57 | [:space:], [:print:], [:punct:], [:graph:], and [:cntrl:]. | ||
| 58 | If not set, then character classes are not supported. */ | ||
| 59 | #define RE_CHAR_CLASSES (RE_BK_PLUS_QM << 1) | ||
| 60 | |||
| 61 | /* If this bit is set, then ^ and $ are always anchors (outside bracket | ||
| 62 | expressions, of course). | ||
| 63 | If this bit is not set, then it depends: | ||
| 64 | ^ is an anchor if it is at the beginning of a regular | ||
| 65 | expression or after an open-group or an alternation operator; | ||
| 66 | $ is an anchor if it is at the end of a regular expression, or | ||
| 67 | before a close-group or an alternation operator. | ||
| 68 | |||
| 69 | This bit could be (re)combined with RE_CONTEXT_INDEP_OPS, because | ||
| 70 | POSIX draft 11.2 says that * etc. in leading positions is undefined. | ||
| 71 | We already implemented a previous draft which made those constructs | ||
| 72 | invalid, though, so we haven't changed the code back. */ | ||
| 73 | #define RE_CONTEXT_INDEP_ANCHORS (RE_CHAR_CLASSES << 1) | ||
| 74 | |||
| 75 | /* If this bit is set, then special characters are always special | ||
| 76 | regardless of where they are in the pattern. | ||
| 77 | If this bit is not set, then special characters are special only in | ||
| 78 | some contexts; otherwise they are ordinary. Specifically, | ||
| 79 | * + ? and intervals are only special when not after the beginning, | ||
| 80 | open-group, or alternation operator. */ | ||
| 81 | #define RE_CONTEXT_INDEP_OPS (RE_CONTEXT_INDEP_ANCHORS << 1) | ||
| 82 | |||
| 83 | /* If this bit is set, then *, +, ?, and { cannot be first in an re or | ||
| 84 | immediately after an alternation or begin-group operator. */ | ||
| 85 | #define RE_CONTEXT_INVALID_OPS (RE_CONTEXT_INDEP_OPS << 1) | ||
| 86 | |||
| 87 | /* If this bit is set, then . matches newline. | ||
| 88 | If not set, then it doesn't. */ | ||
| 89 | #define RE_DOT_NEWLINE (RE_CONTEXT_INVALID_OPS << 1) | ||
| 90 | |||
| 91 | /* If this bit is set, then . doesn't match NUL. | ||
| 92 | If not set, then it does. */ | ||
| 93 | #define RE_DOT_NOT_NULL (RE_DOT_NEWLINE << 1) | ||
| 94 | |||
| 95 | /* If this bit is set, nonmatching lists [^...] do not match newline. | ||
| 96 | If not set, they do. */ | ||
| 97 | #define RE_HAT_LISTS_NOT_NEWLINE (RE_DOT_NOT_NULL << 1) | ||
| 98 | |||
| 99 | /* If this bit is set, either \{...\} or {...} defines an | ||
| 100 | interval, depending on RE_NO_BK_BRACES. | ||
| 101 | If not set, \{, \}, {, and } are literals. */ | ||
| 102 | #define RE_INTERVALS (RE_HAT_LISTS_NOT_NEWLINE << 1) | ||
| 103 | |||
| 104 | /* If this bit is set, +, ? and | aren't recognized as operators. | ||
| 105 | If not set, they are. */ | ||
| 106 | #define RE_LIMITED_OPS (RE_INTERVALS << 1) | ||
| 107 | |||
| 108 | /* If this bit is set, newline is an alternation operator. | ||
| 109 | If not set, newline is literal. */ | ||
| 110 | #define RE_NEWLINE_ALT (RE_LIMITED_OPS << 1) | ||
| 111 | |||
| 112 | /* If this bit is set, then `{...}' defines an interval, and \{ and \} | ||
| 113 | are literals. | ||
| 114 | If not set, then `\{...\}' defines an interval. */ | ||
| 115 | #define RE_NO_BK_BRACES (RE_NEWLINE_ALT << 1) | ||
| 116 | |||
| 117 | /* If this bit is set, (...) defines a group, and \( and \) are literals. | ||
| 118 | If not set, \(...\) defines a group, and ( and ) are literals. */ | ||
| 119 | #define RE_NO_BK_PARENS (RE_NO_BK_BRACES << 1) | ||
| 120 | |||
| 121 | /* If this bit is set, then \<digit> matches <digit>. | ||
| 122 | If not set, then \<digit> is a back-reference. */ | ||
| 123 | #define RE_NO_BK_REFS (RE_NO_BK_PARENS << 1) | ||
| 124 | |||
| 125 | /* If this bit is set, then | is an alternation operator, and \| is literal. | ||
| 126 | If not set, then \| is an alternation operator, and | is literal. */ | ||
| 127 | #define RE_NO_BK_VBAR (RE_NO_BK_REFS << 1) | ||
| 128 | |||
| 129 | /* If this bit is set, then an ending range point collating higher | ||
| 130 | than the starting range point, as in [z-a], is invalid. | ||
| 131 | If not set, then when ending range point collates higher than the | ||
| 132 | starting range point, the range is ignored. */ | ||
| 133 | #define RE_NO_EMPTY_RANGES (RE_NO_BK_VBAR << 1) | ||
| 134 | |||
| 135 | /* If this bit is set, then an unmatched ) is ordinary. | ||
| 136 | If not set, then an unmatched ) is invalid. */ | ||
| 137 | #define RE_UNMATCHED_RIGHT_PAREN_ORD (RE_NO_EMPTY_RANGES << 1) | ||
| 138 | |||
| 139 | /* If this bit is set, succeed as soon as we match the whole pattern, | ||
| 140 | without further backtracking. */ | ||
| 141 | #define RE_NO_POSIX_BACKTRACKING (RE_UNMATCHED_RIGHT_PAREN_ORD << 1) | ||
| 142 | |||
| 143 | /* If this bit is set, do not process the GNU regex operators. | ||
| 144 | If not set, then the GNU regex operators are recognized. */ | ||
| 145 | #define RE_NO_GNU_OPS (RE_NO_POSIX_BACKTRACKING << 1) | ||
| 146 | |||
| 147 | /* If this bit is set, then *?, +? and ?? match non greedily. */ | ||
| 148 | #define RE_FRUGAL (RE_NO_GNU_OPS << 1) | ||
| 149 | |||
| 150 | /* If this bit is set, then (?:...) is treated as a shy group. */ | ||
| 151 | #define RE_SHY_GROUPS (RE_FRUGAL << 1) | ||
| 152 | |||
| 153 | /* If this bit is set, ^ and $ only match at beg/end of buffer. */ | ||
| 154 | #define RE_NO_NEWLINE_ANCHOR (RE_SHY_GROUPS << 1) | ||
| 155 | |||
| 156 | /* If this bit is set, turn on internal regex debugging. | ||
| 157 | If not set, and debugging was on, turn it off. | ||
| 158 | This only works if regex.c is compiled -DDEBUG. | ||
| 159 | We define this bit always, so that all that's needed to turn on | ||
| 160 | debugging is to recompile regex.c; the calling code can always have | ||
| 161 | this bit set, and it won't affect anything in the normal case. */ | ||
| 162 | #define RE_DEBUG (RE_NO_NEWLINE_ANCHOR << 1) | ||
| 163 | |||
| 164 | /* This global variable defines the particular regexp syntax to use (for | ||
| 165 | some interfaces). When a regexp is compiled, the syntax used is | ||
| 166 | stored in the pattern buffer, so changing this does not affect | ||
| 167 | already-compiled regexps. */ | ||
| 168 | extern reg_syntax_t re_syntax_options; | ||
| 169 | |||
| 170 | #ifdef emacs | ||
| 171 | /* In Emacs, this is the string or buffer in which we | ||
| 172 | are matching. It is used for looking up syntax properties. */ | ||
| 173 | extern Lisp_Object re_match_object; | ||
| 174 | #endif | ||
| 175 | |||
| 176 | |||
| 177 | /* Define combinations of the above bits for the standard possibilities. | ||
| 178 | (The [[[ comments delimit what gets put into the Texinfo file, so | ||
| 179 | don't delete them!) */ | ||
| 180 | /* [[[begin syntaxes]]] */ | ||
| 181 | #define RE_SYNTAX_EMACS \ | ||
| 182 | (RE_CHAR_CLASSES | RE_INTERVALS | RE_SHY_GROUPS | RE_FRUGAL) | ||
| 183 | |||
| 184 | #define RE_SYNTAX_AWK \ | ||
| 185 | (RE_BACKSLASH_ESCAPE_IN_LISTS | RE_DOT_NOT_NULL \ | ||
| 186 | | RE_NO_BK_PARENS | RE_NO_BK_REFS \ | ||
| 187 | | RE_NO_BK_VBAR | RE_NO_EMPTY_RANGES \ | ||
| 188 | | RE_DOT_NEWLINE | RE_CONTEXT_INDEP_ANCHORS \ | ||
| 189 | | RE_UNMATCHED_RIGHT_PAREN_ORD | RE_NO_GNU_OPS) | ||
| 190 | |||
| 191 | #define RE_SYNTAX_GNU_AWK \ | ||
| 192 | ((RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS | RE_DEBUG) \ | ||
| 193 | & ~(RE_DOT_NOT_NULL | RE_INTERVALS | RE_CONTEXT_INDEP_OPS)) | ||
| 194 | |||
| 195 | #define RE_SYNTAX_POSIX_AWK \ | ||
| 196 | (RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS \ | ||
| 197 | | RE_INTERVALS | RE_NO_GNU_OPS) | ||
| 198 | |||
| 199 | #define RE_SYNTAX_GREP \ | ||
| 200 | (RE_BK_PLUS_QM | RE_CHAR_CLASSES \ | ||
| 201 | | RE_HAT_LISTS_NOT_NEWLINE | RE_INTERVALS \ | ||
| 202 | | RE_NEWLINE_ALT) | ||
| 203 | |||
| 204 | #define RE_SYNTAX_EGREP \ | ||
| 205 | (RE_CHAR_CLASSES | RE_CONTEXT_INDEP_ANCHORS \ | ||
| 206 | | RE_CONTEXT_INDEP_OPS | RE_HAT_LISTS_NOT_NEWLINE \ | ||
| 207 | | RE_NEWLINE_ALT | RE_NO_BK_PARENS \ | ||
| 208 | | RE_NO_BK_VBAR) | ||
| 209 | |||
| 210 | #define RE_SYNTAX_POSIX_EGREP \ | ||
| 211 | (RE_SYNTAX_EGREP | RE_INTERVALS | RE_NO_BK_BRACES) | ||
| 212 | |||
| 213 | /* P1003.2/D11.2, section 4.20.7.1, lines 5078ff. */ | ||
| 214 | #define RE_SYNTAX_ED RE_SYNTAX_POSIX_BASIC | ||
| 215 | |||
| 216 | #define RE_SYNTAX_SED RE_SYNTAX_POSIX_BASIC | ||
| 217 | |||
| 218 | /* Syntax bits common to both basic and extended POSIX regex syntax. */ | ||
| 219 | #define _RE_SYNTAX_POSIX_COMMON \ | ||
| 220 | (RE_CHAR_CLASSES | RE_DOT_NEWLINE | RE_DOT_NOT_NULL \ | ||
| 221 | | RE_INTERVALS | RE_NO_EMPTY_RANGES) | ||
| 222 | |||
| 223 | #define RE_SYNTAX_POSIX_BASIC \ | ||
| 224 | (_RE_SYNTAX_POSIX_COMMON | RE_BK_PLUS_QM) | ||
| 225 | |||
| 226 | /* Differs from ..._POSIX_BASIC only in that RE_BK_PLUS_QM becomes | ||
| 227 | RE_LIMITED_OPS, i.e., \? \+ \| are not recognized. Actually, this | ||
| 228 | isn't minimal, since other operators, such as \`, aren't disabled. */ | ||
| 229 | #define RE_SYNTAX_POSIX_MINIMAL_BASIC \ | ||
| 230 | (_RE_SYNTAX_POSIX_COMMON | RE_LIMITED_OPS) | ||
| 231 | |||
| 232 | #define RE_SYNTAX_POSIX_EXTENDED \ | ||
| 233 | (_RE_SYNTAX_POSIX_COMMON | RE_CONTEXT_INDEP_ANCHORS \ | ||
| 234 | | RE_CONTEXT_INDEP_OPS | RE_NO_BK_BRACES \ | ||
| 235 | | RE_NO_BK_PARENS | RE_NO_BK_VBAR \ | ||
| 236 | | RE_CONTEXT_INVALID_OPS | RE_UNMATCHED_RIGHT_PAREN_ORD) | ||
| 237 | |||
| 238 | /* Differs from ..._POSIX_EXTENDED in that RE_CONTEXT_INDEP_OPS is | ||
| 239 | removed and RE_NO_BK_REFS is added. */ | ||
| 240 | #define RE_SYNTAX_POSIX_MINIMAL_EXTENDED \ | ||
| 241 | (_RE_SYNTAX_POSIX_COMMON | RE_CONTEXT_INDEP_ANCHORS \ | ||
| 242 | | RE_CONTEXT_INVALID_OPS | RE_NO_BK_BRACES \ | ||
| 243 | | RE_NO_BK_PARENS | RE_NO_BK_REFS \ | ||
| 244 | | RE_NO_BK_VBAR | RE_UNMATCHED_RIGHT_PAREN_ORD) | ||
| 245 | /* [[[end syntaxes]]] */ | ||
| 246 | |||
| 247 | /* Maximum number of duplicates an interval can allow. Some systems | ||
| 248 | (erroneously) define this in other header files, but we want our | ||
| 249 | value, so remove any previous define. */ | ||
| 250 | #ifdef RE_DUP_MAX | ||
| 251 | # undef RE_DUP_MAX | ||
| 252 | #endif | ||
| 253 | /* If sizeof(int) == 2, then ((1 << 15) - 1) overflows. */ | ||
| 254 | #define RE_DUP_MAX (0x7fff) | ||
| 255 | |||
| 256 | |||
| 257 | /* POSIX `cflags' bits (i.e., information for `regcomp'). */ | ||
| 258 | |||
| 259 | /* If this bit is set, then use extended regular expression syntax. | ||
| 260 | If not set, then use basic regular expression syntax. */ | ||
| 261 | #define REG_EXTENDED 1 | ||
| 262 | |||
| 263 | /* If this bit is set, then ignore case when matching. | ||
| 264 | If not set, then case is significant. */ | ||
| 265 | #define REG_ICASE (REG_EXTENDED << 1) | ||
| 266 | |||
| 267 | /* If this bit is set, then anchors do not match at newline | ||
| 268 | characters in the string. | ||
| 269 | If not set, then anchors do match at newlines. */ | ||
| 270 | #define REG_NEWLINE (REG_ICASE << 1) | ||
| 271 | |||
| 272 | /* If this bit is set, then report only success or fail in regexec. | ||
| 273 | If not set, then returns differ between not matching and errors. */ | ||
| 274 | #define REG_NOSUB (REG_NEWLINE << 1) | ||
| 275 | |||
| 276 | |||
| 277 | /* POSIX `eflags' bits (i.e., information for regexec). */ | ||
| 278 | |||
| 279 | /* If this bit is set, then the beginning-of-line operator doesn't match | ||
| 280 | the beginning of the string (presumably because it's not the | ||
| 281 | beginning of a line). | ||
| 282 | If not set, then the beginning-of-line operator does match the | ||
| 283 | beginning of the string. */ | ||
| 284 | #define REG_NOTBOL 1 | ||
| 285 | |||
| 286 | /* Like REG_NOTBOL, except for the end-of-line. */ | ||
| 287 | #define REG_NOTEOL (1 << 1) | ||
| 288 | |||
| 289 | |||
| 290 | /* If any error codes are removed, changed, or added, update the | ||
| 291 | `re_error_msg' table in regex.c. */ | ||
| 292 | typedef enum | ||
| 293 | { | ||
| 294 | #ifdef _XOPEN_SOURCE | ||
| 295 | REG_ENOSYS = -1, /* This will never happen for this implementation. */ | ||
| 296 | #endif | ||
| 297 | |||
| 298 | REG_NOERROR = 0, /* Success. */ | ||
| 299 | REG_NOMATCH, /* Didn't find a match (for regexec). */ | ||
| 300 | |||
| 301 | /* POSIX regcomp return error codes. (In the order listed in the | ||
| 302 | standard.) */ | ||
| 303 | REG_BADPAT, /* Invalid pattern. */ | ||
| 304 | REG_ECOLLATE, /* Not implemented. */ | ||
| 305 | REG_ECTYPE, /* Invalid character class name. */ | ||
| 306 | REG_EESCAPE, /* Trailing backslash. */ | ||
| 307 | REG_ESUBREG, /* Invalid back reference. */ | ||
| 308 | REG_EBRACK, /* Unmatched left bracket. */ | ||
| 309 | REG_EPAREN, /* Parenthesis imbalance. */ | ||
| 310 | REG_EBRACE, /* Unmatched \{. */ | ||
| 311 | REG_BADBR, /* Invalid contents of \{\}. */ | ||
| 312 | REG_ERANGE, /* Invalid range end. */ | ||
| 313 | REG_ESPACE, /* Ran out of memory. */ | ||
| 314 | REG_BADRPT, /* No preceding re for repetition op. */ | ||
| 315 | |||
| 316 | /* Error codes we've added. */ | ||
| 317 | REG_EEND, /* Premature end. */ | ||
| 318 | REG_ESIZE, /* Compiled pattern bigger than 2^16 bytes. */ | ||
| 319 | REG_ERPAREN /* Unmatched ) or \); not returned from regcomp. */ | ||
| 320 | } reg_errcode_t; | ||
| 321 | |||
| 322 | /* This data structure represents a compiled pattern. Before calling | ||
| 323 | the pattern compiler, the fields `buffer', `allocated', `fastmap', | ||
| 324 | `translate', and `no_sub' can be set. After the pattern has been | ||
| 325 | compiled, the `re_nsub' field is available. All other fields are | ||
| 326 | private to the regex routines. */ | ||
| 327 | |||
| 328 | #ifndef RE_TRANSLATE_TYPE | ||
| 329 | # define RE_TRANSLATE_TYPE char * | ||
| 330 | #endif | ||
| 331 | |||
| 332 | struct re_pattern_buffer | ||
| 333 | { | ||
| 334 | /* [[[begin pattern_buffer]]] */ | ||
| 335 | /* Space that holds the compiled pattern. It is declared as | ||
| 336 | `unsigned char *' because its elements are | ||
| 337 | sometimes used as array indexes. */ | ||
| 338 | unsigned char *buffer; | ||
| 339 | |||
| 340 | /* Number of bytes to which `buffer' points. */ | ||
| 341 | size_t allocated; | ||
| 342 | |||
| 343 | /* Number of bytes actually used in `buffer'. */ | ||
| 344 | size_t used; | ||
| 345 | |||
| 346 | /* Syntax setting with which the pattern was compiled. */ | ||
| 347 | reg_syntax_t syntax; | ||
| 348 | |||
| 349 | /* Pointer to a fastmap, if any, otherwise zero. re_search uses | ||
| 350 | the fastmap, if there is one, to skip over impossible | ||
| 351 | starting points for matches. */ | ||
| 352 | char *fastmap; | ||
| 353 | |||
| 354 | /* Either a translate table to apply to all characters before | ||
| 355 | comparing them, or zero for no translation. The translation | ||
| 356 | is applied to a pattern when it is compiled and to a string | ||
| 357 | when it is matched. */ | ||
| 358 | RE_TRANSLATE_TYPE translate; | ||
| 359 | |||
| 360 | /* Number of subexpressions found by the compiler. */ | ||
| 361 | size_t re_nsub; | ||
| 362 | |||
| 363 | /* Zero if this pattern cannot match the empty string, one else. | ||
| 364 | Well, in truth it's used only in `re_search_2', to see | ||
| 365 | whether or not we should use the fastmap, so we don't set | ||
| 366 | this absolutely perfectly; see `re_compile_fastmap'. */ | ||
| 367 | unsigned can_be_null : 1; | ||
| 368 | |||
| 369 | /* If REGS_UNALLOCATED, allocate space in the `regs' structure | ||
| 370 | for `max (RE_NREGS, re_nsub + 1)' groups. | ||
| 371 | If REGS_REALLOCATE, reallocate space if necessary. | ||
| 372 | If REGS_FIXED, use what's there. */ | ||
| 373 | #define REGS_UNALLOCATED 0 | ||
| 374 | #define REGS_REALLOCATE 1 | ||
| 375 | #define REGS_FIXED 2 | ||
| 376 | unsigned regs_allocated : 2; | ||
| 377 | |||
| 378 | /* Set to zero when `regex_compile' compiles a pattern; set to one | ||
| 379 | by `re_compile_fastmap' if it updates the fastmap. */ | ||
| 380 | unsigned fastmap_accurate : 1; | ||
| 381 | |||
| 382 | /* If set, `re_match_2' does not return information about | ||
| 383 | subexpressions. */ | ||
| 384 | unsigned no_sub : 1; | ||
| 385 | |||
| 386 | /* If set, a beginning-of-line anchor doesn't match at the | ||
| 387 | beginning of the string. */ | ||
| 388 | unsigned not_bol : 1; | ||
| 389 | |||
| 390 | /* Similarly for an end-of-line anchor. */ | ||
| 391 | unsigned not_eol : 1; | ||
| 392 | |||
| 393 | #ifdef emacs | ||
| 394 | /* If true, multi-byte form in the `buffer' should be recognized as a | ||
| 395 | multibyte character. */ | ||
| 396 | unsigned multibyte : 1; | ||
| 397 | #endif | ||
| 398 | |||
| 399 | /* [[[end pattern_buffer]]] */ | ||
| 400 | }; | ||
| 401 | |||
| 402 | typedef struct re_pattern_buffer regex_t; | ||
| 403 | |||
| 404 | /* Type for byte offsets within the string. POSIX mandates this. */ | ||
| 405 | typedef int regoff_t; | ||
| 406 | |||
| 407 | |||
| 408 | /* This is the structure we store register match data in. See | ||
| 409 | regex.texinfo for a full description of what registers match. */ | ||
| 410 | struct re_registers | ||
| 411 | { | ||
| 412 | unsigned num_regs; | ||
| 413 | regoff_t *start; | ||
| 414 | regoff_t *end; | ||
| 415 | }; | ||
| 416 | |||
| 417 | |||
| 418 | /* If `regs_allocated' is REGS_UNALLOCATED in the pattern buffer, | ||
| 419 | `re_match_2' returns information about at least this many registers | ||
| 420 | the first time a `regs' structure is passed. */ | ||
| 421 | #ifndef RE_NREGS | ||
| 422 | # define RE_NREGS 30 | ||
| 423 | #endif | ||
| 424 | |||
| 425 | |||
| 426 | /* POSIX specification for registers. Aside from the different names than | ||
| 427 | `re_registers', POSIX uses an array of structures, instead of a | ||
| 428 | structure of arrays. */ | ||
| 429 | typedef struct | ||
| 430 | { | ||
| 431 | regoff_t rm_so; /* Byte offset from string's start to substring's start. */ | ||
| 432 | regoff_t rm_eo; /* Byte offset from string's start to substring's end. */ | ||
| 433 | } regmatch_t; | ||
| 434 | |||
| 435 | /* Declarations for routines. */ | ||
| 436 | |||
| 437 | /* To avoid duplicating every routine declaration -- once with a | ||
| 438 | prototype (if we are ANSI), and once without (if we aren't) -- we | ||
| 439 | use the following macro to declare argument types. This | ||
| 440 | unfortunately clutters up the declarations a bit, but I think it's | ||
| 441 | worth it. */ | ||
| 442 | |||
| 443 | #if defined __STDC__ || defined PROTOTYPES | ||
| 444 | |||
| 445 | # define _RE_ARGS(args) args | ||
| 446 | |||
| 447 | #else /* not __STDC__ || PROTOTYPES */ | ||
| 448 | |||
| 449 | # define _RE_ARGS(args) () | ||
| 450 | |||
| 451 | #endif /* not __STDC__ || PROTOTYPES */ | ||
| 452 | |||
| 453 | /* Sets the current default syntax to SYNTAX, and return the old syntax. | ||
| 454 | You can also simply assign to the `re_syntax_options' variable. */ | ||
| 455 | extern reg_syntax_t re_set_syntax _RE_ARGS ((reg_syntax_t syntax)); | ||
| 456 | |||
| 457 | /* Compile the regular expression PATTERN, with length LENGTH | ||
| 458 | and syntax given by the global `re_syntax_options', into the buffer | ||
| 459 | BUFFER. Return NULL if successful, and an error string if not. */ | ||
| 460 | extern const char *re_compile_pattern | ||
| 461 | _RE_ARGS ((const char *pattern, size_t length, | ||
| 462 | struct re_pattern_buffer *buffer)); | ||
| 463 | |||
| 464 | |||
| 465 | /* Compile a fastmap for the compiled pattern in BUFFER; used to | ||
| 466 | accelerate searches. Return 0 if successful and -2 if was an | ||
| 467 | internal error. */ | ||
| 468 | extern int re_compile_fastmap _RE_ARGS ((struct re_pattern_buffer *buffer)); | ||
| 469 | |||
| 470 | |||
| 471 | /* Search in the string STRING (with length LENGTH) for the pattern | ||
| 472 | compiled into BUFFER. Start searching at position START, for RANGE | ||
| 473 | characters. Return the starting position of the match, -1 for no | ||
| 474 | match, or -2 for an internal error. Also return register | ||
| 475 | information in REGS (if REGS and BUFFER->no_sub are nonzero). */ | ||
| 476 | extern int re_search | ||
| 477 | _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string, | ||
| 478 | int length, int start, int range, struct re_registers *regs)); | ||
| 479 | |||
| 480 | |||
| 481 | /* Like `re_search', but search in the concatenation of STRING1 and | ||
| 482 | STRING2. Also, stop searching at index START + STOP. */ | ||
| 483 | extern int re_search_2 | ||
| 484 | _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string1, | ||
| 485 | int length1, const char *string2, int length2, | ||
| 486 | int start, int range, struct re_registers *regs, int stop)); | ||
| 487 | |||
| 488 | |||
| 489 | /* Like `re_search', but return how many characters in STRING the regexp | ||
| 490 | in BUFFER matched, starting at position START. */ | ||
| 491 | extern int re_match | ||
| 492 | _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string, | ||
| 493 | int length, int start, struct re_registers *regs)); | ||
| 494 | |||
| 495 | |||
| 496 | /* Relates to `re_match' as `re_search_2' relates to `re_search'. */ | ||
| 497 | extern int re_match_2 | ||
| 498 | _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string1, | ||
| 499 | int length1, const char *string2, int length2, | ||
| 500 | int start, struct re_registers *regs, int stop)); | ||
| 501 | |||
| 502 | |||
| 503 | /* Set REGS to hold NUM_REGS registers, storing them in STARTS and | ||
| 504 | ENDS. Subsequent matches using BUFFER and REGS will use this memory | ||
| 505 | for recording register information. STARTS and ENDS must be | ||
| 506 | allocated with malloc, and must each be at least `NUM_REGS * sizeof | ||
| 507 | (regoff_t)' bytes long. | ||
| 508 | |||
| 509 | If NUM_REGS == 0, then subsequent matches should allocate their own | ||
| 510 | register data. | ||
| 511 | |||
| 512 | Unless this function is called, the first search or match using | ||
| 513 | PATTERN_BUFFER will allocate its own register data, without | ||
| 514 | freeing the old data. */ | ||
| 515 | extern void re_set_registers | ||
| 516 | _RE_ARGS ((struct re_pattern_buffer *buffer, struct re_registers *regs, | ||
| 517 | unsigned num_regs, regoff_t *starts, regoff_t *ends)); | ||
| 518 | |||
| 519 | #if defined _REGEX_RE_COMP || defined _LIBC | ||
| 520 | # ifndef _CRAY | ||
| 521 | /* 4.2 bsd compatibility. */ | ||
| 522 | extern char *re_comp _RE_ARGS ((const char *)); | ||
| 523 | extern int re_exec _RE_ARGS ((const char *)); | ||
| 524 | # endif | ||
| 525 | #endif | ||
| 526 | |||
| 527 | /* GCC 2.95 and later have "__restrict"; C99 compilers have | ||
| 528 | "restrict", and "configure" may have defined "restrict". */ | ||
| 529 | #ifndef __restrict | ||
| 530 | # if ! (2 < __GNUC__ || (2 == __GNUC__ && 95 <= __GNUC_MINOR__)) | ||
| 531 | # if defined restrict || 199901L <= __STDC_VERSION__ | ||
| 532 | # define __restrict restrict | ||
| 533 | # else | ||
| 534 | # define __restrict | ||
| 535 | # endif | ||
| 536 | # endif | ||
| 537 | #endif | ||
| 538 | /* For now conditionally define __restrict_arr to expand to nothing. | ||
| 539 | Ideally we would have a test for the compiler which allows defining | ||
| 540 | it to restrict. */ | ||
| 541 | #ifndef __restrict_arr | ||
| 542 | # define __restrict_arr | ||
| 543 | #endif | ||
| 544 | |||
| 545 | /* POSIX compatibility. */ | ||
| 546 | extern int regcomp _RE_ARGS ((regex_t *__restrict __preg, | ||
| 547 | const char *__restrict __pattern, | ||
| 548 | int __cflags)); | ||
| 549 | |||
| 550 | extern int regexec _RE_ARGS ((const regex_t *__restrict __preg, | ||
| 551 | const char *__restrict __string, size_t __nmatch, | ||
| 552 | regmatch_t __pmatch[__restrict_arr], | ||
| 553 | int __eflags)); | ||
| 554 | |||
| 555 | extern size_t regerror _RE_ARGS ((int __errcode, const regex_t *__preg, | ||
| 556 | char *__errbuf, size_t __errbuf_size)); | ||
| 557 | |||
| 558 | extern void regfree _RE_ARGS ((regex_t *__preg)); | ||
| 559 | |||
| 560 | |||
| 561 | #ifdef __cplusplus | ||
| 562 | } | ||
| 563 | #endif /* C++ */ | ||
| 564 | |||
| 565 | #endif /* regex.h */ | ||
| 566 | |||
| 567 | /* | ||
| 568 | Local variables: | ||
| 569 | make-backup-files: t | ||
| 570 | version-control: t | ||
| 571 | trim-versions-without-asking: nil | ||
| 572 | End: | ||
| 573 | */ | ||
| 574 | |||
| 575 | /* arch-tag: bda6e3ec-3c02-4237-a55a-01ad2e120083 | ||
| 576 | (do not change this comment) */ | ||
diff --git a/src/tparam.c b/src/tparam.c index 7306e0d5c10..ea208692161 100644 --- a/src/tparam.c +++ b/src/tparam.c | |||
| @@ -144,7 +144,9 @@ tparam1 (string, outstring, len, up, left, argp) | |||
| 144 | int outlen = 0; | 144 | int outlen = 0; |
| 145 | 145 | ||
| 146 | register int tem; | 146 | register int tem; |
| 147 | int *old_argp = argp; | 147 | int *old_argp = argp; /* can move */ |
| 148 | int *fixed_argp = argp; /* never moves */ | ||
| 149 | int explicit_param_p = 0; /* set by %p */ | ||
| 148 | int doleft = 0; | 150 | int doleft = 0; |
| 149 | int doup = 0; | 151 | int doup = 0; |
| 150 | 152 | ||
| @@ -180,7 +182,10 @@ tparam1 (string, outstring, len, up, left, argp) | |||
| 180 | if (c == '%') | 182 | if (c == '%') |
| 181 | { | 183 | { |
| 182 | c = *p++; | 184 | c = *p++; |
| 183 | tem = *argp; | 185 | if (explicit_param_p) |
| 186 | explicit_param_p = 0; | ||
| 187 | else | ||
| 188 | tem = *argp; | ||
| 184 | switch (c) | 189 | switch (c) |
| 185 | { | 190 | { |
| 186 | case 'd': /* %d means output in decimal. */ | 191 | case 'd': /* %d means output in decimal. */ |
| @@ -203,7 +208,10 @@ tparam1 (string, outstring, len, up, left, argp) | |||
| 203 | *op++ = tem % 10 + '0'; | 208 | *op++ = tem % 10 + '0'; |
| 204 | argp++; | 209 | argp++; |
| 205 | break; | 210 | break; |
| 206 | 211 | case 'p': /* %pN means use param N for next subst. */ | |
| 212 | tem = fixed_argp[(*p++) - '1']; | ||
| 213 | explicit_param_p = 1; | ||
| 214 | break; | ||
| 207 | case 'C': | 215 | case 'C': |
| 208 | /* For c-100: print quotient of value by 96, if nonzero, | 216 | /* For c-100: print quotient of value by 96, if nonzero, |
| 209 | then do like %+. */ | 217 | then do like %+. */ |
diff --git a/src/xfaces.c b/src/xfaces.c index b3eba748720..fecd8e546cf 100644 --- a/src/xfaces.c +++ b/src/xfaces.c | |||
| @@ -1528,8 +1528,11 @@ If FRAME is nil or omitted, use the selected frame. */) | |||
| 1528 | { | 1528 | { |
| 1529 | struct frame *f; | 1529 | struct frame *f; |
| 1530 | 1530 | ||
| 1531 | CHECK_FRAME (frame); | ||
| 1532 | CHECK_STRING (color); | 1531 | CHECK_STRING (color); |
| 1532 | if (NILP (frame)) | ||
| 1533 | frame = selected_frame; | ||
| 1534 | else | ||
| 1535 | CHECK_FRAME (frame); | ||
| 1533 | f = XFRAME (frame); | 1536 | f = XFRAME (frame); |
| 1534 | return face_color_gray_p (f, SDATA (color)) ? Qt : Qnil; | 1537 | return face_color_gray_p (f, SDATA (color)) ? Qt : Qnil; |
| 1535 | } | 1538 | } |
| @@ -1546,8 +1549,11 @@ COLOR must be a valid color name. */) | |||
| 1546 | { | 1549 | { |
| 1547 | struct frame *f; | 1550 | struct frame *f; |
| 1548 | 1551 | ||
| 1549 | CHECK_FRAME (frame); | ||
| 1550 | CHECK_STRING (color); | 1552 | CHECK_STRING (color); |
| 1553 | if (NILP (frame)) | ||
| 1554 | frame = selected_frame; | ||
| 1555 | else | ||
| 1556 | CHECK_FRAME (frame); | ||
| 1551 | f = XFRAME (frame); | 1557 | f = XFRAME (frame); |
| 1552 | if (face_color_supported_p (f, SDATA (color), !NILP (background_p))) | 1558 | if (face_color_supported_p (f, SDATA (color), !NILP (background_p))) |
| 1553 | return Qt; | 1559 | return Qt; |
| @@ -2254,7 +2260,7 @@ static double | |||
| 2254 | font_rescale_ratio (name) | 2260 | font_rescale_ratio (name) |
| 2255 | char *name; | 2261 | char *name; |
| 2256 | { | 2262 | { |
| 2257 | Lisp_Object tail, elt; | 2263 | Lisp_Object tail, elt; |
| 2258 | 2264 | ||
| 2259 | for (tail = Vface_font_rescale_alist; CONSP (tail); tail = XCDR (tail)) | 2265 | for (tail = Vface_font_rescale_alist; CONSP (tail); tail = XCDR (tail)) |
| 2260 | { | 2266 | { |
| @@ -2467,7 +2473,7 @@ x_face_list_fonts (f, pattern, pfonts, nfonts, try_alternatives_p) | |||
| 2467 | 2473 | ||
| 2468 | if (nfonts < 0 && CONSP (lfonts)) | 2474 | if (nfonts < 0 && CONSP (lfonts)) |
| 2469 | num_fonts = XFASTINT (Flength (lfonts)); | 2475 | num_fonts = XFASTINT (Flength (lfonts)); |
| 2470 | 2476 | ||
| 2471 | /* Make a copy of the font names we got from X, and | 2477 | /* Make a copy of the font names we got from X, and |
| 2472 | split them into fields. */ | 2478 | split them into fields. */ |
| 2473 | n = nignored = 0; | 2479 | n = nignored = 0; |
| @@ -3182,7 +3188,13 @@ lface_fully_specified_p (attrs) | |||
| 3182 | for (i = 1; i < LFACE_VECTOR_SIZE; ++i) | 3188 | for (i = 1; i < LFACE_VECTOR_SIZE; ++i) |
| 3183 | if (i != LFACE_FONT_INDEX && i != LFACE_INHERIT_INDEX | 3189 | if (i != LFACE_FONT_INDEX && i != LFACE_INHERIT_INDEX |
| 3184 | && i != LFACE_AVGWIDTH_INDEX) | 3190 | && i != LFACE_AVGWIDTH_INDEX) |
| 3185 | if (UNSPECIFIEDP (attrs[i])) | 3191 | if (UNSPECIFIEDP (attrs[i]) |
| 3192 | #ifdef MAC_OS | ||
| 3193 | /* MAC_TODO: No stipple support on Mac OS yet, this index is | ||
| 3194 | always unspecified. */ | ||
| 3195 | && i != LFACE_STIPPLE_INDEX | ||
| 3196 | #endif | ||
| 3197 | ) | ||
| 3186 | break; | 3198 | break; |
| 3187 | 3199 | ||
| 3188 | return i == LFACE_VECTOR_SIZE; | 3200 | return i == LFACE_VECTOR_SIZE; |