In addition to the automatic changes, also do the following,
needed due to recent Gnulib changes.
* admin/merge-gnulib (AVOIDED_MODULES): Add btoc32,
c32_apply_type_test, c32_get_type_test, c32isalnum, c32rtomb,
c32tolower, c32toupper, localeinfo, mbrtoc32-regular.
Remove btowc, iswctype, mbrtowc, wcrtomb, wctype, wctype-h.
Also remove iswblank, iswdigit, iswxdigit, locale-h, raise,
stdarg-h, some of which perhaps could have been removed earlier.
* configure.ac (_REGEX_AVOID_UCHAR_H): New macro.
* configure.ac: When configured with --enable-gcc-warnings,
use the -Wno-analyzer-allocation-size option if available.
This works around a false positive bug in
GCC 16.0.1 20260321 (Red Hat 16.0.1-0) x86-64; see
<https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125116>.
This enables features unrelated to native image APIs, such as
'w32-export-frame' and 'yank-media', which need to call GDI+
functions.
* configure.ac (HAVE_W32) [!CYGWIN]: Always include w32image.o in
W32_OBJ.
* etc/NEWS: Drop 'yank-media' '--without-native-image-api' caveat.
* src/emacs.c (main):
* src/w32select.c (convert_dibv5_to_png):
* src/w32term.c (w32_draw_image_foreground): Replace
HAVE_NATIVE_IMAGE_API with WINDOWSNT.
* src/w32.c (term_ntproc): Remove redundant HAVE_NATIVE_IMAGE_API guard.
* src/w32fns.c (Fw32_export_frame) [!WINDOWSNT]: Return nil on Cygwin.
By disabling hw-acceleration we can re-enable xwidgets with newer
webkit2gtk versions (Bug#80728).
Newer versions were disallowed (Bug#66068) since they abort when a
webkit view is created; however, they work fine as long as we disable
hw-acceleration, since the off-screen windows that xwidgets uses do not
support that.
* configure.ac: Remove upper-bound for webkit2gtk version check.
* src/xwidget.c (Fmake_xwidget): Turn off hardware-acceleration.
Copyright-paperwork-exempt: yes
Now we use treesit_traverse_sibling_helper when it exists, and
use the old code otherwise. So we still support older
tree-sitter versions.
* configure.ac (LIBSYSTEMD_CFLAGS): Add back old config checking
for the malloc function. Add another check for
ts_tree_cursor_goto_previous_sibling.
* src/treesit.c: (treesit_traverse_sibling_helper): Add back the
old code that doesn't require
ts_tree_cursor_goto_previous_sibling.
Without this fix, a 64-bit build on Ubuntu 25.10 x86-64 failed
when the libselinux1-dev:i386 package (providing <selinux/selinux.h>)
was installed but the libselinux1-dev package (providing
/usr/lib/x86_64-linux-gnu/libselinux.so) was not.
The first symptom was when compiling emacsclient:
‘/usr/bin/ld: ../lib/libgnu.a(file-has-acl.o): in function
`get_aclinfo': .../lib/file-has-acl.c:247:(.text+0x4d8): undefined
reference to `getfilecon'’.
* configure.ac: Default with_selinux to maybe, not to yes.
Defer our SELinux configuration to after Gnulib’s, so that
we handle any SELinux problems that Gnulib discovers.
* configure.ac (LIBSYSTEMD_CFLAGS): Increase minimal required
tree-sitter version to 0.20.10.
* src/treesit.c (treesit_traverse_sibling_helper): When
traversing forward, use the new function
ts_tree_cursor_goto_previous_sibling.
* configure.ac [HAVE_WEBP]: Explicitly check for both libwebpdemux
and libwebp (where the former depends on the latter), since we use
both (bug#57420, bug#61988, bug#66221). Check for WebPDecodeRGBA as
a representative of libwebp and libwebpdecoder now that we no longer
use WebPGetInfo elsewhere. Prepend WEBP_LIBS to LIBS as per
AC_CHECK_LIB. Clarify commentary.
* configure.ac: If --enable-gcc-warnings, don’t use
-Wzero-as-null-pointer-constant, which is enabled by default with
bleeding-edge Gnulib (as there seems to be movement in this
direction in C2y). Although -Wzero-as-null-pointer-constant is
useful, Emacs isn’t clean for it yet.
* src/pdumper.c (pdumper_set_emacs_execdir): Add workaround for
!NS_SELF_CONTAINED so that launching Emacs from the app bundle
works flawlessly when built with native compilation.
* configure.ac: Define BINDIR. (Bug#80094)
* configure.ac: Check for 'NSImageNameCaution' declaration when
building on GNUstep.
* src/nsterm.m ([EmacsBell init]): Use fallback implementation on
GNUstep only if 'NSImageNameCaution' is not declared. (Bug#80107)
* configure.ac: Stop suppresssing -Wunused-const-variable=2 when
configured --with-gcc-warnings. Suppressing no longer seems to be
needed with current GCC and src/lisp.h.
Problem reported by Ulrich Müller (bug#79974).
* configure.ac (gl_ALIGNASOF): Arrange for it to be defined
even if aclocal is invoked other than via admin/merge-gnulib.
Also, change comments from # to dnl as they don’t make
much sense in the shell script.
Remove Gnulib sources no longer needed or used.
admin/merge-gnulib is typically not smart enough to do that.
* configure.ac: Do not call gl_EEMALLOC, as it is no longer
supplied or needed.
* lib/explicit_bzero.c, lib/malloc/scratch_buffer_dupfree.c:
* lib/xalloc-oversized.h, m4/eealloc.m4, m4/explicit_bzero.m4:
* m4/st_dm_mode.m4, m4/year2038.m4:
Remove.
* configure.ac: Move the gl_INIT call back to where it was,
as Mattias Engdegård reported the following problem:
checking whether pthread_sigmask is a macro... no
./configure: line 32767: test: =: unary operator expected
because moving gl_INIT up made the check for pthread_sigmask’s
availability appear later than Gnulib’s check whether it’s a macro.
Instead of moving gl_INIT up, call gl_ALIGNASOF earlier,
as that’s the only part of gl_INIT that is needed earlier.
This is mostly a cleanup patch to fix growing discrepancies
between src/lisp.h and configure.ac in terms of how they deduce
how EMACS_INT aligns. The patch includes a static check that the
two methods now agree. It also speeds up ‘configure’ a bit.
* configure.ac (ALIGNOF_INT, ALIGNOF_LONG, ALIGNOF_LONG_LONG):
Remove; no longer used.
(ALIGNOF_EMACS_INT): New macro.
(system_malloc): Set to 'no' more consistently with what’s
in src/lisp.h.
* src/lisp.h (ALIGNOF_EMACS_INT): Do not define here, as config.h
defines it now. Check that config.h’s definition equals the
actual alignof (EMACS_INT).
(USE_USB_TAG): Set more consistently with what’s in configure.ac.
(alignas): Don’t second-guess Gnulib.
These have not been needed for some time in configure.ac.
* configure.ac (_AC_PROG_CC_C89): Don’t bother redefining to empty.
(gl_PROG_CC_C99, AC_PROG_CC_C_O): Remove; no longer needed.
* configure.ac: When the WebKit header files are not available
in the hard-coded location, query for the location of the
currently active SDK, and resort to the header files there.
(Bug#79847)
On this platform, you cannot reliably malloc objects containing
values of type pthread_mutex_t or pthread_cond_t, since malloc
guarantees only 8-byte alignment but these two types require
16-byte alignment. See GCC bug 115750
<https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115750> reported by
Sam James and Emacs bug 79936 <https://bugs.gnu.org/79936>
reported by John Paul Adrian Glaubitz.
* configure.ac (ALIGNOF_MAX_ALIGN_T, ALIGNOF_PTHREAD_COND_T)
(ALIGNOF_PTHREAD_MUTEX_T) [HAVE_PTHREAD]: New C macros.
* src/systhread.c (sys_mutex_init, sys_mutex_lock)
(sys_mutex_unlock, sys_cond_init, sys_cond_wait)
(sys_cond_signal, sys_cond_broadcast, sys_cond_destroy):
Use SYTHREAD_ALIGN_PTR to convert sys_mutex_t and sys_cond_t to
pthread_mutex_t and pthread_cond_t, since runtime conversion is
needed on GNU/Linux HPPA.
* src/systhread.h (SYSTHREAD_ALIGN_ROOM, SYSTHREAD_ALIGN_PTR):
New macros, which affect the generated code only on
unusual platforms like GNU/Linux HPPA.
(sys_mutex_t, sys_cond_t) [HAVE_PTHREAD]: Use them.
Problem reported by David Caldwell (bug#79879).
* configure.ac: When configuring for GNUstep and the
Objective C compiler does not support even C99 features,
use the first flag of -std=gnu23, -std=gnu17, -std=gnu11, -std=gnu99
that works, instead of insisting on -std=c99 which (a) is too
strict and (b) can override another -std= option in an
undesirable way.
* configure.ac: Use gl_CHECK_FUNCS_ANDROID_MACOS to check for
'posix_spawn_file_actions_addchdir', since it checks which
version of macOS is being used for the build, rather than what
is available in the SDK. (Bug#79489)
Document the change in dependency. As we still need to work
around GCC bug 119085, there is no functional change in ‘configure’.
* configure.ac (emacs_cv_gcc_bug_119085_CFLAGS):
Rename from emacs_cv_gcc_bug_117423_CFLAGS. All uses changed.
Problem reported by Oliver Reiter and Pip Cet (Emacs bug#78473).
It looks like GCC tree-sra is trouble in general with how Emacs
uses unions, so disable tree-sra until the GCC bug is fixed.
* configure.ac (C_SWITCH_MACHINE): Also work around GCC bug 117423.
* configure.ac (gl_cv_onwards_func_tzalloc): Define to "future
OS version" on Android API 35 and later.
Detect posix_spawn* by means of gl_CHECK_FUNCS_ANDROID.
* src/conf_post.h (tzalloc, tzfree): Define to non-conflicting
names on Android 35 and later.
* configure.ac (systemduserunitdir): New variable and option.
* Makefile.in (systemduserunitdir): New, set by configure.
(install-etc, uninstall): Don't install emacs.service when
systemduserunitdir is unspecified.
* etc/NEWS: Announce the new configure option. (Bug#76924)
* configure.ac: Don't claim that Xft is being considered for
deletion or that recent releases are buggy, or imply that
HarfBuzz support is exclusive to Cairo.
* configure.ac (ALIGNOF_INT, ALIGNOF_LONG, ALIGNOF_LONG_LONG):
New variables.
(emacs_cv_alignas_unavailable): Define if alignas and structure
alignment primitives are unavailable. In such an environment,
the MSB tagging scheme must be enabled, as must the GNU malloc.
* msdos/sed2v2.inp: Adjust correspondingly.
* src/alloc.c (union emacs_align_type): Remove types which
contain flexible array members. The address of a field
subsequent to an aggregate with flexible array members cannot
validly be taken.
(mark_memory) [!USE_LSB_TAG && !WIDE_EMACS_INT]: Strip type bits
before scanning memory.
* src/emacs.c (main):
* src/eval.c (Fautoload_do_load):
* src/fns.c (Frequire): Rename a number of illogically named
fields.
* src/lisp.h (ALIGNOF_EMACS_INT): Define to the natural
alignment of EMACS_INT.
(IDEAL_GCALIGNMENT): New macro.
(USE_LSB_TAG): Disable if no alignment specifiers are available,
WIDE_EMACS_INT is undefined, and the natural alignment of
EMACS_INT falls short of LSB tagging's requirements.
(gflags): Rename illogically named fields and don't define them
as bitfields, which runs afoul of certain compiler issues.
(will_dump_p, will_bootstrap_p, will_dump_with_pdumper_p)
(dumped_with_pdumper_p): Adjust accordingly.
* src/pdumper.c (VM_SUPPORTED): Define to 0 when !USE_LSB_TAG.
It is better to read dump files into the heap by hand than to be
supplied with an address that is not representable.
(_dump_object_start_pseudovector): Rename to
dump_object_start_pseudovector, to avoid encroaching on reserved
names.
(START_DUMP_PVEC): Adjust correspondingly.
(dump_mmap_contiguous_vm): Preserve errno around failure
cleanup.
(dump_bitset_bit_set_p): Work around certain compiler issues.
(pdumper_load) [!USE_LSB_TAG]: Reject dump file allocations
that are not representable as Lisp_Objects.
Tested on i386-unknown-solaris2.10, sparc-sun-solaris2.10.
