* src/process.c (wait_reading_process_output): Allow
child_signal_read_fd < 0.

If we receive and handle a SIGCHLD signal for a process while waiting
for that process, 'pselect' might never return.  Instead, we have to
explicitly 'pselect' that the process status has changed.  We do this
by writing to a pipe in the SIGCHLD handler and having
'wait_reading_process_output' select on it.

* src/process.c (child_signal_init): New helper function to create a
pipe for SIGCHLD notifications.
(child_signal_read, child_signal_notify): New helper functions to
read from/write to the child signal pipe.
(create_process): Initialize the child signal pipe on first use.
(handle_child_signal): Notify waiters that a process status has
changed.
(wait_reading_process_output): Make sure that we also catch
SIGCHLD/process status changes.

* test/src/process-tests.el
(process-tests/fd-setsize-no-crash/make-process): Remove workaround,
which is no longer needed.

* src/frame.c (Fset_mouse_position, Fset_mouse_pixel_position):
Don't compile the FRAME_MSDOS_P case on platforms other than
MSDOS, as that will never happen there.

* src/frame.c (Fset_mouse_position, Fset_mouse_pixel_position): Call
Fselect_frame and appropriate mouse_moveto function on all non-GUI
frame types, independent of #ifdef's.
* src/term.c (init_tty): Initialize mouse_face_window for all
non-GUI frame types.
(term_mouse_moveto) [HAVE_GPM]: Make available even if
HAVE_WINDOW_SYSTEM is defined.
* src/xdisp.c (try_window_id): Call gui_clear_window_mouse_face
in all cases.

Before adbb4eacc2a984c0fc0b65ec761368fd9067d6c5,
`read_and_dispose_of_process_output` called
`record_asynch_buffer_change` which added "artificial" input events
(in the form of BUFFER_SWITCH_EVENTs), causing sit_for to return
Qnil when interrupted by process output.  Without those BUFFER_SWITCH_EVENTs,
sit_for now tends to return Qt when interrupted by process output
making `read_char` believe that we've waited the whole timeout,
As consequence incoming process output tended to cause premature
auto-saving of files (sometimes right after almost every key press).

This patch recovers the previous behavior, which is not ideal
(incoming process output can delay auto-save indefinitely), but has
been good enough for many years.

First, scan to find the first non-whitespace character and then
backtrack to find the beginning of the line.  The previous
algorithm always started on the non-whitespace character during
the backtrack, causing it to stop immediately and not actually
find the beginning of the line.  The same applies to the end of
line calculation.
* src/xdisp.c: (Fwindow_text_pixel_size): Fix off by one error.
(Bug#45748)

* test/src/xdisp-tests.el (xdisp-tests--window-text-pixel-size)
(xdisp-tests--window-text-pixel-size-leading-space)
(xdisp-tests--window-text-pixel-size-trailing-space): New tests.

Change calls from 'read-event' to 'read-key' in libraries expecting
mouse events.  Do this only when 'xterm-mouse-mode' is enabled.  That
way those libraries read decoded mouse events instead of the
underlying escape sequence.  Add a parameter to 'read-key' that avoids
running any of the unbound fallbacks in 'read-key-sequence' so the
libraries can read mouse button-down events.

For backward compatibility purposes, the above logic is contained in a
new internal-only function: 'read--potential-mouse-event'.

* doc/lispref/commands.texi (Reading One Event): Document new
parameter to 'read-key'.  Mention that non-character events on
terminals need 'read-key'.
* lisp/subr.el (read-key-full-map): Add new keymap used by 'read-key'.
(read-key): Add new parameter 'fallbacks-disabled' to prevent running
any of the unbound fallbacks normally run by 'read-key-sequence'.
(read--potential-mouse-event): Add new function that calls 'read-key'
or 'read-event' depending on if 'xterm-mouse-mode' is set.
* lisp/foldout.el (foldout-mouse-swallow-events):
* lisp/isearch.el (isearch-pre-command-hook):
* lisp/mouse-drag.el (mouse-drag-throw, mouse-drag-drag):
* lisp/mouse.el (mouse-drag-secondary):
* lisp/ruler-mode.el (ruler-mode-mouse-grab-any-column)
(ruler-mode-mouse-drag-any-column-iteration):
* lisp/strokes.el (strokes-read-stroke, strokes-read-complex-stroke):
* lisp/textmodes/artist.el (artist-mouse-draw-continously)
(artist-mouse-draw-poly, artist-mouse-draw-2points):
* lisp/vc/ediff-wind.el (ediff-get-window-by-clicking):
* lisp/wid-edit.el (widget-button--check-and-call-button)
(widget-button-click): Call 'read--potential-mouse-event' instead of
'read-event'.
* lisp/wid-edit.el (widget-key-sequence-read-event): Call 'read-key'
with 'fallbacks-disabled' set instead of 'read-event'.  Unlike above
changes, this is unconditionally applied so it works for function
keys too.  Apply 'local-function-key-map' instead of
'function-key-map' as that contains the full terminal translations.
* lisp/vc/ediff.el (ediff-windows): Use 'display-mouse-p' to check if
a mouse is available.
* src/lread.c (Fread_event): Recommend 'read-key' in docstring for
'read-event' for non-character events.

* lisp/subr.el (local-key-binding, global-key-binding): New defuns.
* src/keymap.c (Flocal_key_binding, Fglobal_key_binding): Remove DEFUNs.
(syms_of_keymap): Remove defsubrs for above DEFUNs.
* test/lisp/subr-tests.el (subr-test-local-key-binding)
(subr-test-global-key-binding): New tests.

* src/xdisp.c (move_it_in_display_line_to): Don't reset
next_may_wrap after saving a potential wrap point.  This fixes the
case where several characters in a row can serve as a wrap point.
(Bug#45837)

* src/fns.c (Fbuffer_line_statistics): New function.