Michael Albinus [Mon, 9 Nov 2015 09:00:56 +0000 (10:00 +0100)]
Add kqueue support
* configure.ac (--with-file-notification): Add kqueue.
(top): Remove special test for "${HAVE_NS}" and
${with_file_notification}, this is handled inside gfilenotify
tests. Add kqueue tests. Use NOTIFY_CFLAGS and NOTIFY_LIBS
instead of library specific variables.
* src/Makefile.in: Use NOTIFY_CFLAGS and NOTIFY_LIBS.
* src/emacs.c (main): Call globals_of_kqueue and syms_of_kqueue.
* src/kqueue.c: New file.
* src/lisp.h: Declare extern globals_of_kqueue and syms_of_kqueue.
João Távora [Tue, 17 Nov 2015 22:23:59 +0000 (22:23 +0000)]
linum-mode plays more nicely with other margin-setting extensions
linum.el will only modify the left margin if it needs to, and will
only reset the it back to 0 if it guesses that no-one has touched that
margin in the meantime.
As such, this is a more of a workaround than an actual fix, but fixes
the problems described in bug#20674 regarding the interaction with
modes such as darkroom-mode and olivetti-mode.
Artur Malabarba [Sun, 15 Nov 2015 21:28:37 +0000 (21:28 +0000)]
* lisp/emacs-lisp/package.el: Fix a decoding issue
(package--with-response-buffer): Use `url-insert-buffer-contents'.
The previous code had some issues with decoding. Refactoring that
function allows us to use the decoding from url-handlers while still
treating both sync and async requests the same.
* lisp/url/url-handlers.el (url-insert-file-contents): Move some code to
`url-insert-buffer-contents'.
(url-insert-buffer-contents): New function
(package--with-work-buffer-async): Reimplement as
`package--with-response-buffer'.
(package--with-work-buffer): Mark obsolete.
(package--with-response-buffer): New macro. This is a more self
contained and less contrived version of
`package--with-work-buffer-async'. It uses keyword arguments,
doesn't have async on the name, doesn't fallback on
`package--with-work-buffer', and has _much_ simpler error
handling.
(package--check-signature, package--download-one-archive)
(package-install-from-archive, describe-package-1): Use it.
(package--download-and-read-archives): Let
`package--download-one-archive' take care of calling
`package--update-downloads-in-progress'.
Stefan Monnier [Sun, 15 Nov 2015 19:04:30 +0000 (14:04 -0500)]
* lisp/progmodes/verilog-mode.el: Use with-silent-modifications
(verilog-save-buffer-state): Use with-silent-modifications when available.
(verilog-save-font-no-change-functions): Don't bind
before/after-change-functions if it's not needed.
David Engster [Sat, 14 Nov 2015 22:14:36 +0000 (23:14 +0100)]
gitmerge: Fix git log command
* admin/gitmerge.el (gitmerge-missing): Use '--left-only' since we
only want commits from the branch that is to be merged.
(gitmerge-setup-log-buffer): Use the same symmetric range as in
`gitmerge-missing'.
David Engster [Sat, 14 Nov 2015 16:53:10 +0000 (17:53 +0100)]
gitmerge: Try to detect cherry-picks
* admin/gitmerge.el (gitmerge-default-branch): Change to
origin/emacs-25.
(gitmerge-missing): Use symmetric difference ('...') between
branch and master so that cherry-picks can be detected.
Xue Fuqiao [Sat, 14 Nov 2015 04:49:17 +0000 (12:49 +0800)]
Mention CONTRIBUTE in README
Mention CONTRIBUTE in README, since it was moved from etc/ to root.
* etc/TODO: Remove the reference to `etc/CONTRIBUTE'.
* README: Mention CONTRIBUTE.
* verilog-mode.el (verilog-auto, verilog-delete-auto)
(verilog-modi-cache-results, verilog-save-buffer-state)
(verilog-save-font-no-change-functions): When internally suppressing change
functions, use `inhibit-modification-hooks' and call
`after-change-funtions' to more nicely work with user hooks. Reported by
Stefan Monnier.
(verilog-auto, verilog-delete-auto, verilog-delete-auto-buffer):
Create `verilog-delete-auto-buffer' to avoid double-calling
fontification hooks.
(verilog-restore-buffer-modified-p, verilog-auto)
(verilog-save-buffer-state): Prefer restore-buffer-modified-p over
set-buffer-modified-p. Reported by Stefan Monnier.
Dmitry Gutov [Sat, 14 Nov 2015 00:37:01 +0000 (02:37 +0200)]
Use generic dispatch for xref backends
* lisp/progmodes/xref.el (xref-backend-functions):
New variable.
(xref-find-function): Remove.
(xref-find-backend)
(xref--etags-backend): New functions.
(xref-identifier-at-point-function)
(xref-identifier-completion-table-function): Remove.
(xref-backend-definitions, xref-backend-references)
(xref-backend-apropos, xref-backend-identifier-at-point)
(xref-backend-identifier-completion-table):
New generic functions.
* lisp/progmodes/elisp-mode.el (emacs-lisp-mode): Add
`elisp--xref-backend' to the beginning of
`xref-backend-functions', locally. Delete references to
removed functions and vars.
(elisp-xref-find): Remove.
(elisp--xref-backend): New function.
(elisp--xref-find-references, elisp--xref-find-apropos)
(elisp--xref-identifier-completion-table):
Turn into appropriately named generic methods.
Juri Linkov [Fri, 13 Nov 2015 23:28:03 +0000 (01:28 +0200)]
Support rectangular regions for more commands
* lisp/simple.el (region-extract-function): Handle the arg value ‘bounds’.
(region-insert-function): New function.
(shell-command-on-region): Add arg ‘region-noncontiguous-p’.
If non-nil, operate on multiple chunks.
(region-noncontiguous-p): New function.
* lisp/rect.el: Add function rectangle--insert-region
around region-insert-function.
(extract-rectangle-bounds): New function.
(rectangle--extract-region): Handle the arg value ‘bounds’.
(rectangle--insert-region): New function.
* lisp/emulation/cua-rect.el: Add function cua--insert-rectangle
around region-insert-function.
(cua--extract-rectangle-bounds): New function.
(cua--rectangle-region-extract): Handle the arg value ‘bounds’.
* lisp/replace.el (query-replace, query-replace-regexp): Add arg
‘region-noncontiguous-p’. Use ‘use-region-p’.
(query-replace-regexp-eval, map-query-replace-regexp)
(replace-string, replace-regexp): Use ‘use-region-p’.
(keep-lines, flush-lines, how-many): Use ‘use-region-p’.
(perform-replace): Add arg ‘region-noncontiguous-p’.
If non-nil, operate on multiple chunks.
* src/casefiddle.c (Fdowncase_region): Add arg ‘region-noncontiguous-p’.
If non-nil, operate on multiple chunks. (Bug#19829)
Dmitry Gutov [Mon, 9 Nov 2015 03:24:23 +0000 (05:24 +0200)]
Handle multiple matches on the same line; add highlighting
* lisp/progmodes/xref.el (xref-location-marker): Interpret the
column value in characters.
(xref--collect-matches): Rename from `xref--collect-match'.
Search for all matches in the hit line. Add `highlight' face to
the matched region in the summary. Update both callers.
Dmitry Gutov [Sun, 8 Nov 2015 03:01:05 +0000 (05:01 +0200)]
Replace xref-match-bounds with xref-match-length
Relying on xref-location-marker to point to the beginning of the match
* lisp/progmodes/xref.el (xref-match-bounds): Remove.
(xref-match-length): Add.
(xref-make-match): Change the arguments.
(xref--match-buffer-bounds): Remove.
(xref-match-item): Store length, instead of end-column.
(xref-pulse-momentarily)
(xref--collect-match)
(xref--query-replace-1): Update accordingly.
(xref-query-replace): Ditto. And check that the search results
are up-to-date.
Paul Eggert [Fri, 13 Nov 2015 20:02:21 +0000 (12:02 -0800)]
Spruce up ftfont.c memory allocation
* src/ftfont.c (setup_otf_gstring):
Avoid O(N**2) behavior when reallocating.
(ftfont_shape_by_flt): Prefer xpalloc to xrealloc when
reallocating buffers; this simplifies the code. Do not trust
mflt_run to leave the output areas unchanged on failure, as
this isn’t part of its interface spec.
Paul Eggert [Fri, 13 Nov 2015 17:28:53 +0000 (09:28 -0800)]
Port recent XCB changes to 64-bit ‘long int’
For historical reasons, libX11 represents 32-bit values like Atoms as
‘long int’ even on platforms where ‘long int’ is 64 bits. XCB doesn’t
do that, so adapt the recent XCB code to behave properly on 64-bit
platforms. Also, fix what appears to be a bug in the interpretation
of xcb_get_property_value_length, at least on my Fedora platform
which is running libxcb-1.11-5.fc21.
* src/xfns.c (x_real_pos_and_offsets):
* src/xterm.c (get_current_wm_state):
xcb_get_property_value_length returns a byte count, not a word count.
For 32-bit quantities, xcb_get_property_value returns a vector
of 32-bit words, not of (possibly 64-bit) long int.
Eli Zaretskii [Fri, 13 Nov 2015 13:33:29 +0000 (15:33 +0200)]
Remove support for ':timeout' from w32 tray notifications
* src/w32fns.c (Fw32_notification_notify): Delete the code that
supports ':timeout'.
(syms_of_w32fns): Don't DEFSYM ':timeout'. This avoids clashes
with dbusbind.c when D-Bus is compiled in.
Eli Barzilay [Fri, 13 Nov 2015 09:46:20 +0000 (11:46 +0200)]
Fix point positioning after transposing with negative arg
* lisp/simple.el (transpose-subr): When invoked with a negative
argument, move point to after the transposed text, like we do
when invoked with a positive argument. (Bug#21885)
Eli Zaretskii [Fri, 13 Nov 2015 09:02:04 +0000 (11:02 +0200)]
Another fix for MinGW64 and Cygwin builds due to notifications
* src/w32fns.c: Ifdef away tray notification code if D-Bus is
being compiled into Emacs.
(syms_of_w32fns) [WINDOWSNT && !HAVE_DBUS]: Don't defsubr
Sw32_notification_notify and Sw32_notification_close if the code
is not compiled. Reported by Andy Moreton <andrewjmoreton@gmail.com>.
Remove intern calls and XXX comments from Fx_export_frames
* src/xfns.c (Fx_export_frames): Use Qpdf, Qpng, Qpostscript, and
Qsvg instead of intern calls. Use "postscript" instead of "ps"
for consistency with image types. Remove XXX comments.
(syms_of_xfns) <Qpdf>: DEFSYM it.
* test/automated/keymaps-test.el: Fix test to make it repeatable
(keymap-store_in_keymap-FASTINT-on-nonchars): Reset Buffer-menu-mode-map
entry to its initial value to make the test repeatable in interactive
sessions (assuming it doesn't fail and crashes Emacs, of course).
Phillip Lord [Thu, 6 Aug 2015 20:33:58 +0000 (21:33 +0100)]
The heuristic that Emacs uses to add an `undo-boundary' has been
reworked, as it interacts poorly with functions on `post-command-hook'
or `after-change-functions'.
* lisp/simple.el: New section added.
* src/cmds.c (remove_excessive_undo_boundaries): Now in lisp.
(self_insert_command): Calls simple.el to amalgamate.
(delete_char): Calls simple.el to amalgamate.
* src/keyboard.c (last_undo_boundary): Removed.
* src/undo.c (run_undoable_change): New function.
Eli Zaretskii [Thu, 12 Nov 2015 20:40:31 +0000 (22:40 +0200)]
Fix the MinGW64 and Cygwin-w32 builds
* src/w32fns.c (MYNOTIFYICONDATAW_V1_SIZE)
(MYNOTIFYICONDATAW_V2_SIZE, MYNOTIFYICONDATAW_V3_SIZE): Define and
use instead of the corresponding NOTIFYICONDATAW_Vn_SIZE macros,
which cause trouble with MinGW42 headers. Ifdef away tray
notifications code for Cygwin. Reported by Andy Moreton
<andrewjmoreton@gmail.com>.
Simen Heggestøyl [Thu, 12 Nov 2015 17:30:37 +0000 (18:30 +0100)]
Enable sorting of JSON object keys when encoding
* lisp/json.el (json-encoding-object-sort-predicate): New variable for
specifying a sorting predicate for JSON objects during encoding.
(json--plist-to-alist): New utility function.
(json-encode-hash-table): Re-use `json-encode-alist' when object keys
are to be sorted.
(json-encode-alist): Sort output by
`json-encoding-object-sort-predicate, when set.
(json-encode-plist): Re-use `json-encode-alist' when object keys are
to be sorted.
(json-pretty-print-buffer-ordered): New command to pretty print the
buffer with object keys sorted alphabetically.
(json-pretty-print-ordered): New command to pretty print the region with
object keys sorted alphabetically.
* test/automated/json-tests.el (test-json-plist-to-alist)
(test-json-encode-plist, test-json-encode-hash-table)
(test-json-encode-alist-with-sort-predicate)
(test-json-encode-plist-with-sort-predicate): New tests.
Ken Raeburn [Sat, 7 Nov 2015 08:06:32 +0000 (03:06 -0500)]
Speed up x_real_pos_and_offsets using XCB
* src/xfns.c (x_real_pos_and_offsets) [USE_XCB]: Add XCB flavors of
all X calls, and pipeline requests when possible, collecting results
later. Eliminate use of x_catch_errors (and thus XSync) in XCB case.
Ken Raeburn [Sat, 7 Nov 2015 08:06:32 +0000 (03:06 -0500)]
Enable use of XCB for checking window manager state
* src/xterm.c (get_current_wm_state) [USE_XCB]: Use XCB calls instead
of XGetWindowProperty plus error-catching, since we can explicitly
check for errors in the XCB version. This eliminates 3 XSync calls on
top of the round-trip actually fetching the information.
Ken Raeburn [Sat, 7 Nov 2015 08:06:32 +0000 (03:06 -0500)]
Reduce some data dependencies between X calls
Gains nothing in the traditional-Xlib code, but more closely aligns
with how the XCB version will work.
* src/xfns.c (x_real_pos_and_offsets): When translating coordinates,
send coordinates (0,0) to the X server and add in the real coordinates
after getting the response. Move XGetGeometry for outer window inside
error-trapping block. Use DPY variable more, since it's available.
Eli Barzilay [Thu, 12 Nov 2015 08:07:38 +0000 (03:07 -0500)]
Add "^" to the interactive specs of `dired-next/previous-line'
* lisp/dired.el (dired-next-line, dired-previous-line): It makes sense
to bind these commands to the arrow keys, and that means that they work
better with a "^" in the `interactive' declaration so selection works
as expected.
Alan Mackenzie [Wed, 11 Nov 2015 22:06:12 +0000 (22:06 +0000)]
CC Mode: Respect users' settings of open-paren-in-column-0-is-defun-start.
lisp/progmodes/cc-engine.el (c-backward-single-comment, c-backward-comments)
(c-invalidate-state-cache-1, c-parse-state-1, c-guess-basic-syntax):
remove bindings of open-paren-in-column-0-is-defun-start to nil.
(c-get-fallback-scan-pos): "New" function (existed several years ago).
(c-parse-state-get-strategy): Reintroduce the 'BOD strategy, using
c-get-fallback-scan-pos.
(c-parse-state-1): Handle 'BOD strategy.
lisp/progmodes/cc-mode.el (c-before-change, c-after-change)
c-font-lock-fontify-region): remove bindings of
open-paren-in-column-0-is-defun-start to nil.
cc-mode.texi (Performance Issues, Limitations and Known Bugs): Fix mix up
between @chapter and @appendix.
Nicolas Petton [Wed, 11 Nov 2015 17:18:32 +0000 (18:18 +0100)]
Rename seq-p and map-p to seqp and mapp
* lisp/emacs-lisp/seq.el (seqp): New name.
* lisp/emacs-lisp/map.el (mapp): New name.
* doc/lispref/sequences.texi: Update the documentation for seqp.
* test/automated/map-tests.el: Update the tests for mapp.
David Reitter [Tue, 10 Nov 2015 20:41:17 +0000 (15:41 -0500)]
Avoid creating notification objects when possible
* src/nsterm.m (windowWillEnterFullScreen, windowWillExitFullScreen:,
windowDidEnterFullScreen, windowDidExitFullScreen): provide
convenience functions that do not require a notification object. When
needed, define NSWindowDidEnterFullScreenNotification to allow for
compilation on OS X 10.6.8.
Paul Eggert [Tue, 10 Nov 2015 20:46:22 +0000 (12:46 -0800)]
Move INTEGER_TO_CONS body out of .h file
* src/data.c (INTBIG_TO_LISP): New macro, with most
of the contents of the old INTEGER_TO_CONS.
(intbig_to_lisp, uintbig_to_lisp): New functions.
* src/lisp.h (INTEGER_TO_CONS):
Simplify by using EXPR_SIGNED and the new functions.
This shrinks code size a bit, and makes it easier to
put a breakpoint on handling of large integers.
Artur Malabarba [Tue, 10 Nov 2015 13:26:00 +0000 (13:26 +0000)]
* lisp/files.el: Don't allow customization of dir-locals sorting
In retrospect, this is not a good idea for the same reason that
`dir-locals-file' is a defconst, because it is important that this
behaviour be "uniform across different environments and users".
Sure, the user can still change the sorting with a hack, but we
shouldn't encourage them to change it.
(dir-locals--all-files): Return list in the order returned by
`file-expand-wildcards'.
(file-expand-wildcards): Document the sorting predicate used.
(dir-locals-sort-predicate): Delete variable.
(dir-locals-find-file, dir-locals-collect-variables)
(dir-locals-read-from-file): Update accordingly.
(hack-dir-local-variables): Rename a local variable.
Thanks to Juanma Barranquero for noticing the autoload problems
introduced by my recent commit adding/changing the above functions
(Sun Nov 8 14:16:43 2015 -0500, git commit 3812e17978).
Noah Friedman [Tue, 10 Nov 2015 01:34:40 +0000 (17:34 -0800)]
(ydump-buffer): Handle case where gap is at the start of buffer.
I don't recall if older versions of gdb were less strict but you
cannot dump a 0-length range in gdb 7.9.1.
Dmitry Gutov [Tue, 10 Nov 2015 00:41:06 +0000 (02:41 +0200)]
Fold `project-ask-user' into `project-current'
* lisp/progmodes/project.el (project-find-functions): Remove
`project-ask-user'.
(project-ask-user): Remove function and the corresponding
`project-roots' implementation.
(project-current): Add a new argument, MAYBE-PROMPT. Prompt the
user in case there's no project in the current directory. Update
all callers.
Karl Fogel [Mon, 9 Nov 2015 21:57:23 +0000 (15:57 -0600)]
When VC detects a conflict, specify which file
* lisp/vc/vc.el (vc-message-unresolved-conflicts): New function.
* lisp/vc/vc-svn.el (vc-svn-find-file-hook):
* lisp/vc/vc-hg.el (vc-hg-find-file-hook):
* lisp/vc/vc-bzr.el (vc-bzr-find-file-hook):
* lisp/vc/vc-git.el (vc-git-find-file-hook): Use above new function
to display a standard message that specifies the conflicted file.
Before this change, the message VC used for indicating a conflicted
file was just "There are unresolved conflicts in this file" without
naming the file (and this language was duplicated in several places).
After this change, it's "There are unresolved conflicts in file FOO"
(and this language is now centralized in one function in vc.el).
Justification: It's important for the message to name the conflicted
file because the moment when VC realizes a file is conflicted does not
always come interactively. For example, some people automatically
find a set of Org Mode files on startup, and may keep those .org files
under version control. If any of the files are conflicted, the user
just sees some messages fly by, and might later check the "*Messages*"
buffer to find out what files were conflicted. I'm not saying this
happened to me or anything; it's a purely hypothetical example.
Eli Zaretskii [Mon, 9 Nov 2015 18:31:45 +0000 (20:31 +0200)]
Fix assertion violation in define-key
* src/keymap.c (store_in_keymap): Don't use XFASTINT on non-character
objects. Reported by Drew Adams <drew.adams@oracle.com>
and Juanma Barranquero <lekktu@gmail.com>.
Dima Kogan [Mon, 9 Nov 2015 16:36:05 +0000 (18:36 +0200)]
Fix a memory leak in GC of font cache
* src/alloc.c (compact_font_cache_entry): Don't GC unmarked font
entities if some of the fonts it references are marked. This
plugs a memory leak. (Bug#21556)
Add support for retrieving the path to a JSON element. This can for
instance be useful to retrieve paths in deeply nested JSON
structures.
* lisp/json.el (json-pre-element-read-function)
(json-post-element-read-function): New variables to hold pre- and post
read callback functions for `json-read-array' and `json-read-object'.
(json--path): New variable used internally by `json-path-to-position'.
(json--record-path, json--check-position): New functions used
internally by `json-path-to-position'.
(json-path-to-position): New function for retrieving the path to a
JSON element at a given position.
(json-read-object, json-read-array): Call
`json-pre-element-read-function' and `json-post-element-read-function'
when set.
* test/automated/json-tests.el (test-json-path-to-position-with-objects)
(test-json-path-to-position-with-arrays)
(test-json-path-to-position-no-match): New tests for
`json-path-to-position'.
Karl Fogel [Sun, 8 Nov 2015 19:16:43 +0000 (14:16 -0500)]
Offer non-overwrite bookmark setter (Bug#15746)
* lisp/bookmark.el (bookmark-set-internal): New helper function to do
what `bookmark-set' used to do, but with more choices for overwrite
vs push, and with minor changes to the interactive prompt format.
(bookmark-set): Rewrite as wrapper around above.
If overwriting, inform the user of that in the prompt.
(bookmark-set-no-overwrite): New function, also done as wrapper.
Bind to "M" in `ctl-x-r-map' autoloads.
(bookmark-map): Similarly bind "M" here.
Alan Modra [Sun, 8 Nov 2015 17:29:00 +0000 (09:29 -0800)]
ELF unexec: Don't insert a new section
Reuse the .bss section instead, making it SHT_PROGBITS. This way we
don't need to mess with symbol st_shndx, or section sh_link and
sh_info.
This does lead to eu-elflint complaints about symbols defined in .bss
with a needed version, because normally it is undefined symbols that
have needed versions; Defined symbols have version definitions.
The exception is symbols defined by the linker in .dynbss for
variables copied from a shared library in order to avoid text
relocations, with copy relocs to copy their initial values from the
shared library. These symbols are both defined and have needed
versions, and eu-elflink only expects to see them in SHT_NOBITS
sections. Of course there is no real problem with having such symbols
in SHT_PROGBITS sections. glibc ld.so handles them fine.
* unexelf.c: Delete outdated comments.
(PATCH_INDEX): Delete.
(find_section): Delete.
(unexec): Don't add a new section. Instead reuse the last bss
section, extending it to cover dumped data. Make bss sections
SHT_PROGBITS. Remove all patching of sh_link, sh_info and
st_shndx. Rename bss sections.
Alan Modra [Sun, 8 Nov 2015 17:29:00 +0000 (09:29 -0800)]
ELF unexec: Drive from PT_LOAD header rather than sections
This rewrites bss handling in the ELF unexec code. Finding bss
sections by name results in complicated code that
- does not account for all names of possible bss sections,
- assumes specific ordering of bss sections,
- can wrongly choose a SHT_NOBITS section not in the bss segment,
- incorrectly calculates bss size (no accounting for alignment gaps),
- assumes .data and .bss are in the same segment.
All of these problems and more are solved by finding the bss segment
in PT_LOAD headers, ie. the address range included in p_memsz but not
p_filesz of the last PT_LOAD header, then matching SHT_NOBITS sections
in that address range.
* unexelf.c: Delete old ppc comment.
(OLD_PROGRAM_H): Define.
(round_up): Delete.
(unexec): Don't search for bss style sections by name. Instead,
use the last PT_LOAD header address range covered by p_memsz
but not p_filesz and match any SHT_NOBITS section in that
address range. Simplify initialisation of section header vars.
Don't assume that section headers are above bss segment. Move
copying of bss area out of section loop. Align .data2 section
to 1, since it now covers the entire bss area. For SHT_NOBITS
sections in the bss segment, leave sh_addr and sh_addralign
unchanged, but correct sh_offset. Clear memory corresponding
to SHT_NOBITS .plt section. Delete comment and hacks for
sections partly overlapping bss range now that the full range
is properly calculated. Delete now dead .sbss code.
(Bug#20614)
Alan Modra [Sun, 8 Nov 2015 17:29:00 +0000 (09:29 -0800)]
ELF unexec: _OBJC_ symbols in bss sections
This code assumed that there was only one bss section. Rather than
checking for a particular index, check the section type. Also, handle
the possibility that the section was SHT_NOBITS originally and is
unchanged, in which case no clearing is needed (and sh_offset isn't
necessarily valid, which can lead to a wild memset).
* unexelf.c (unexec): Properly handle _OBJC_ symbols in bss sections.