Prevent over-eager rx character range condensation

`rx' incorrectly considers character ranges between ASCII and raw bytes to
cover all codes in-between, which includes all non-ASCII Unicode chars.
This causes (any "\000-\377" ?Å) to be simplified to (any "\000-\377"),
which is not at all the same thing: [\000-\377] really means
[\000-\177\200-\377] (Bug#34492).

* lisp/emacs-lisp/rx.el (rx-any-condense-range): Split ranges going
from ASCII to raw bytes.
* test/lisp/emacs-lisp/rx-tests.el (rx-char-any-raw-byte): Add test case.
* etc/NEWS: Mention the overall change (Bug#33205).

1 files changed, 5 insertions, 1 deletions

diff --git a/test/lisp/emacs-lisp/rx-tests.el b/test/lisp/emacs-lisp/rx-tests.el
index f15e1016f7c..e14feda347f 100644
--- a/test/lisp/emacs-lisp/rx-tests.el
+++ b/test/lisp/emacs-lisp/rx-tests.el
@@ -53,7 +53,11 @@
   ;; Range of raw characters, multibyte.
   (should (equal (string-match-p (rx (any "Å\211\326-\377\177"))
                                  "XY\355\177\327")
-                 2)))
+                 2))
+  ;; Split range; \177-\377ÿ should not be optimised to \177-\377.
+  (should (equal (string-match-p (rx (any "\177-\377" ?ÿ))
+                                 "ÿA\310B")
+                 0)))
 
 (ert-deftest rx-pcase ()
   (should (equal (pcase "a 1 2 3 1 1 b"