Paul Eggert [Sat, 13 Feb 2016 00:23:25 +0000 (16:23 -0800)]
Suppress GNUstep hardening
Fedora 23 normally hardens GNUstep applications, which causes
‘./configure --with-ns’ to break Emacs’s funky way of undumping.
Fix this by eliding the hardening options (Bug#22518).
* src/Makefile.in (LIBS_GNUSTEP): Omit options like
‘-specs=/usr/lib/rpm/redhat/redhat-hardened-ld’.
(GNU_OBJC_CFLAGS): Omit options like
‘-specs=/usr/lib/rpm/redhat/redhat-hardened-cc1’.
Paul Eggert [Wed, 10 Feb 2016 19:40:09 +0000 (11:40 -0800)]
make-docfile cleanup for I/O, etc.
* lib-src/make-docfile.c (progname, generate_globals, num_globals)
(num_globals_allocated, globals): Now static.
(generate_globals, struct rcsoc_state, read_c_string_or_comment):
(write_c_args, scan_c_stream, search_lisp_doc_at_eol, scan_lisp_file):
Use bool for boolean.
(verror): New function.
(fatal, error): Use it. API is now like printf. All callers changed.
(main): Remove err_count local that was always 0.
(main, scan_c_stream, scan_lisp_file): Check for I/O error.
(scan_file, scan_c_file, scan_c_stream, scan_lisp_file):
Return void, not 0.
(put_char, scan_keyword_or_put_char, scan_c_file): Use char for byte.
(scan_keyword_or_put_char): Check for missing ( and unexpected EOF.
(close_emacs_globals): Use ptrdiff_t for index, not int.
(scan_c_file, scan_lisp_file): Exit with failure if file cannot be
opened, rather than diagnosing but exiting with status 0.
(search_lisp_doc_at_eol): Don't worry about ungetc of EOF; it's
portable now.
Paul Eggert [Wed, 10 Feb 2016 18:14:50 +0000 (10:14 -0800)]
Memory-management cleanup in make-docfile
I compiled it with -fsanitize=address and fixed the leaks it detected.
Also, I changed it to prefer signed to unsigned integer types,
and to check for integer overflow.
* lib-src/make-docfile.c:
Include <stddef.h>, <stdint.h>, <intprops.h>, <min-max.h>.
(memory_exhausted): New function.
(xmalloc, xrealloc): Use it.
(xmalloc, xrealloc, scan_file, struct rcsoc_state, write_c_args)
(uncompiled, scan_lisp_file):
Prefer signed integer types to unsigned.
(xstrdup): Remove. All uses removed.
(num_globals, num_globals_allocated, write_globals, scan_c_stream):
Use ptrdiff_t, not int, for indexes that in theory could exceed INT_MAX.
(add_global): Use const to pacify --enable-gcc-warnings.
Make a copy here, rather than relying on strdup calls later.
(add_global, write_globals, scan_c_stream):
Avoid integer overflow when calculating sizes.
(write_globals, scan_c_stream, scan_lisp_file): Avoid memory leak.
(scan_c_stream): Check for add_global failure.
Eli Zaretskii [Wed, 10 Feb 2016 18:30:12 +0000 (20:30 +0200)]
Improve doc strings of 'forward/backward-word-strictly'
* lisp/simple.el (backward-word): Refer to 'backward-word-strictly'
in the doc string. Suggested by Glenn Morris <rgm@gnu.org>.
* lisp/subr.el (forward-word-strictly, backward-word-strictly):
Mention 'subword-mode' in the doc strings.
* src/syntax.c (Fforward_word): Refer to 'forward-word-strictly'
in the doc string. (Bug#22560)
Dima Kogan [Wed, 10 Feb 2016 04:40:03 +0000 (15:40 +1100)]
Support integer image rotation and respect EXIF rotations
* src/image.c (imagemagick_load_image): Allow integer rotations in
addition to floating point rotations (bug#22591).
* src/image.c (imagemagick_load_image): Images that have an
orientation given in EXIF and have no explicit :rotation tag are now
pre-rotated. All information such as width/height is reported for the
rotated image.
Paul Eggert [Tue, 9 Feb 2016 23:01:44 +0000 (15:01 -0800)]
Suppress ACL ops if configured with --disable-acl
Without this patch, some ACL operations were suppressed, but not all.
* src/fileio.c [!USE_ACL]: Do not include sys/acl.h.
(Ffile_acl, Fset_file_acl) [!USE_ACL]: Return nil in this case.
Paul Eggert [Tue, 9 Feb 2016 17:06:40 +0000 (09:06 -0800)]
Sync with gnulib
This incorporates:
2016-02-09 stdalign: port to clang 3.7.0
2016-02-06 misc: port better to gcc -fsanitize=address
* doc/misc/texinfo.tex, lib/stdalign.in.h, m4/acl.m4, m4/getgroups.m4:
Copy from gnulib.
Paul Eggert [Tue, 9 Feb 2016 04:24:55 +0000 (20:24 -0800)]
Increase success rate of fallback lmalloc
* src/alloc.c (lmalloc, lrealloc): Reallocate with (typically)
larger and larger sizes, to increase the probability that
the allocator will return a Lisp-aligned pointer.
David Edmondson [Mon, 8 Feb 2016 23:26:52 +0000 (10:26 +1100)]
Compare recipient and keys case-insensitively
* lisp/gnus/mml2015.el: (mml-secure-check-user-id): When comparing a
recipient address with that from a key, do so in a case insensitive
manner (bug#22603).
Katsumi Yamaoka [Mon, 8 Feb 2016 22:41:25 +0000 (22:41 +0000)]
Refactor HTML images handling of Gnus and mm-* (a part of bug#21650)
* doc/misc/emacs-mime.texi (Display Customization):
Remove mm-inline-text-html-with-images; add documentations for
mm-html-inhibit-images and mm-html-blocked-images.
* lisp/gnus/gnus-art.el (gnus-article-show-images):
No need to bind mm-inline-text-html-with-images.
(gnus-bind-safe-url-regexp): Rename to gnus-bind-mm-vars.
(gnus-bind-mm-vars): Rename from gnus-bind-safe-url-regexp;
bind mm-html-inhibit-images and mm-html-blocked-images.
(gnus-mime-view-all-parts, gnus-mime-view-part-internally)
(gnus-mm-display-part, gnus-mime-display-single)
(gnus-mime-display-alternative): Use gnus-bind-mm-vars.
* lisp/gnus/mm-decode.el (mm-inline-text-html-with-images): Remove.
(mm-html-inhibit-images, mm-html-blocked-images): New user options.
(mm-shr): Bind shr-inhibit-images and shr-blocked-images with
mm-html-inhibit-images and mm-html-blocked-images respectively
instead of gnus-inhibit-images and gnus-blocked-images.
* lisp/gnus/mm-view.el (mm-setup-w3m): Use mm-html-inhibit-images
instead of mm-inline-text-html-with-images.
Paul Eggert [Mon, 8 Feb 2016 19:38:35 +0000 (11:38 -0800)]
Port to FreeBSD 11-CURRENT i386
Problem reported by Herbert J. Skuhra in:
http://lists.gnu.org/archive/html/emacs-devel/2016-02/msg00354.html
Instead of trying
* src/alloc.c (lmalloc, lrealloc, laligned): New functions.
(xmalloc, xzalloc, xrealloc, lisp_malloc): Use them.
(__alignof__) [!__GNUC__ && !__alignof__]: New macro.
(MALLOC_IS_GC_ALIGNED): New macro.
* src/lisp.h (NONPOINTER_BITS): Remove. All uses removed.
No longer needed now that alloc.c uses lmalloc and lrealloc.
Michael Albinus [Mon, 8 Feb 2016 09:54:43 +0000 (10:54 +0100)]
Some fixes in file-notify-tests.el
* test/automated/file-notify-tests.el
(file-notify--test-with-events-check)
(file-notify--test-with-events-explainer): New defuns.
(file-notify--test-with-events): Use it.
(file-notify-test07-backup): Fix docstring. Some of the
backends fire two `changed' events. Backup by rename doesn't
work for kqueue.
Michael Albinus [Sun, 7 Feb 2016 18:30:01 +0000 (19:30 +0100)]
Fix Bug#22557
* lisp/filenotify.el (file-notify-callback): Do not send a
`stopped' event in case of backup by renaming. (Bug#22557)
* test/automated/Makefile.in: Use $(SELECTOR_EXPENSIVE) for
all targets but check and check-maybe.
* test/automated/file-notify-tests.el
(file-notify--test-read-event-timeout): New defconst.
(file-notify--deftest-remote, file-notify--wait-for-events)
(file-notify-test02-events)
(file-notify-test04-file-validity)
(file-notify-test06-many-events): Use it.
(file-notify--test-cleanup): Make it more robust. Delete also
backup file.
(file-notify-test07-backup): New test.
Paul Eggert [Sun, 7 Feb 2016 15:46:59 +0000 (07:46 -0800)]
Fix test for dladdr
Problem reported by Andreas Schwab in:
http://lists.gnu.org/archive/html/emacs-devel/2016-02/msg00327.html
* configure.ac (dladdr): Link with LIBMODULES when checking for
this function.
Artur Malabarba [Sun, 7 Feb 2016 08:35:49 +0000 (08:35 +0000)]
* lisp/comint.el (comint-prompt-read-only): Clean tabs in docstring
While tabs in code are mostly fine because the Emacs sources have a
.dir-locals file specifying tab-width, the same is not true of tabs in
code examples inside docstrings. The docstring is printed on a *Help*
buffer, which can be created on any directory and won't necessarily have
the same tab-width set.
* src/process.c (conv_lisp_to_sockaddr): AF_INET6 addresses are
converted to a list of 16 bit quantities by
conv_sockaddr_to_lisp(). conv_lisp_to_sockaddr() should follow the
same scheme rather than expecting a (longer) list of 8 bit
quantities.
Oscar Fuentes [Sat, 6 Feb 2016 21:12:53 +0000 (22:12 +0100)]
Use monitor's resolution for positioning tooltips
* src/xfns.c (compute_tip_xy): Use the resolution of the monitor where
the mouse pointer is to avoid placing the tooltip over the border of
the monitor on multi-head displays. Fixes bug#22549.
Eli Zaretskii [Sat, 6 Feb 2016 19:53:38 +0000 (21:53 +0200)]
Fix file-notify-test on MS-Windows
* test/automated/file-notify-tests.el (file-notify--test-timeout):
Reduce w32notify timeout to 10 sec.
(file-notify-test06-many-events): Call read-event after each
rename, to keep the w32notify backend happy in batch mode.
(Bug#22534)
Eli Zaretskii [Sat, 6 Feb 2016 16:30:35 +0000 (18:30 +0200)]
Fix issues found by auditing w32notify code
* src/w32inevt.c (handle_file_notifications): Count the number of
events to be returned.
* src/w32notify.c (send_notifications): Don't copy to the file
notification buffer more than it can hold. (Bug#22534)
Alan Mackenzie [Thu, 4 Feb 2016 19:01:50 +0000 (19:01 +0000)]
Correctly fontify C++ initializations which "look like" functions.
Fixes bug#7579.
lisp/progmodes/cc-engine.el (c-forward-declarator): Add extra optional
parameter to enable handling of "anonymous" declarators in declarations.
lisp/progmodes/cc-fonts.el (c-font-lock-declarators): Check more rigorously
whether a "(" opens a parameter list of a function, or an initialization of a
variable.