bug#37095: [PATCH] Save match data in ucs-normalize-region

[Akinori MUSHA <knu@iDaemons.org>]

[-- Attachment #1.1.1: Type: text/plain, Size: 1581 bytes --]

A patch generated by git format-patch is attached below, which simply
wraps `ucs-normalize-region` with `save-match-data`.

I'm a user of the Emacs Mac port by mituharu was investigating a bug
where dired fails to open a certain local directory on macOS.  The
error was raised at `replace-match` in the `insert-directory`
function:

```
		(when (re-search-forward "^ *\\(total\\)" nil t)
		  (let ((available (get-free-disk-space ".")))
		    (when available
		      ;; Replace "total" with "used", to avoid confusion.
		      (replace-match "total used in directory" nil nil nil 1)
```

And it turned out the match data changed after returning from
`get-free-disk-space` and that was why `replace-match` failed.

Inside of `get-free-disk-space` most platforms uses a generic method
to get the free space, and that part is fine because it is surrounded
by `save-match-data`.  However, the Mac port is one of the few
platforms that implements a native 'file-system-info` function, which
is called if it exists.  Then, the `file-system-info` in `src/mac.c`
calls ENCODE_FILE() on a given directory name, which in the end calls
`ucs-normalize-region` to normalize the filename, where the match data
is clobbered.

https://bitbucket.org/mituharu/emacs-mac/src/df827786d7a7fb0a0e2f27577af67e32d9a888a9/src/mac.c#lines-2337

ENCODE_FILE() is transparently called by many C functions, which means
`ucs-normalize-region` can be called at unpredictable timings, so I
think it should keep match data unchanged.

-- 
Akinori MUSHA / https://akinori.org/


[-- Attachment #1.1.2: 0001-Save-match-data-in-ucs-normalize-region.patch --]
[-- Type: text/plain, Size: 2416 bytes --]

From 11d49adcbadcfcbe844f873e63ba1d596f72e4c5 Mon Sep 17 00:00:00 2001
From: Akinori MUSHA <knu@idaemons.org>
Date: Mon, 19 Aug 2019 23:53:50 +0900
Subject: [PATCH] Save match data in ucs-normalize-region

* lisp/international/ucs-normalize.el (ucs-normalize-region): Save
match data.  This function can be transparently called via
ENCODE_FILE() when the underlying filesytem uses normalized Unicode
filenames, so it requires much care not to cause any side effects.
---
 lisp/international/ucs-normalize.el | 27 ++++++++++++++-------------
 1 file changed, 14 insertions(+), 13 deletions(-)

diff --git a/lisp/international/ucs-normalize.el b/lisp/international/ucs-normalize.el
index 6f1e770c09..d02d4b8e3e 100644
--- a/lisp/international/ucs-normalize.el
+++ b/lisp/international/ucs-normalize.el
@@ -514,19 +514,20 @@ ucs-normalize-region
       (narrow-to-region from to)
       (goto-char (point-min))
       (let (start-pos starter)
-        (while (re-search-forward quick-check-regexp nil t)
-          (setq starter (string-to-char (match-string 0)))
-          (setq start-pos (match-beginning 0))
-          (ucs-normalize-block
-           ;; from
-           (if (or (= start-pos (point-min))
-                   (and (= 0 (ucs-normalize-ccc starter))
-                        (not (memq starter ucs-normalize-combining-chars))))
-               start-pos (1- start-pos))
-           ;; to
-           (if (looking-at ucs-normalize-combining-chars-regexp)
-               (match-end 0) (1+ start-pos))
-           translation-table composition-predicate))))))
+        (save-match-data
+          (while (re-search-forward quick-check-regexp nil t)
+            (setq starter (string-to-char (match-string 0)))
+            (setq start-pos (match-beginning 0))
+            (ucs-normalize-block
+             ;; from
+             (if (or (= start-pos (point-min))
+                     (and (= 0 (ucs-normalize-ccc starter))
+                          (not (memq starter ucs-normalize-combining-chars))))
+                 start-pos (1- start-pos))
+             ;; to
+             (if (looking-at ucs-normalize-combining-chars-regexp)
+                 (match-end 0) (1+ start-pos))
+             translation-table composition-predicate)))))))
 
 ;; --------------------------------------------------------------------------------
 
-- 
2.22.0


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 195 bytes --]