Add bmo-2080242-include-WidgetUtilsGtk.h-for-wayland-or-nodbus-builds.patch to
fix building with '-dbus'.
Add llvm-23 compatibility now that rust-1.99 has it.
Always build jpegxl support as it's now default-enabled by upstream. I see no
benefit in keeping the use flag for disabling it (and the option altogether
might be removed from upstream code at some point).
Bump nss requirement to 3.130.
Rebase bmo-847568-support-system-harfbuzz-graphite2.patch (rebase provided by:
tt_1). '+system-harfbuzz' doesn't compile with Firefox-158.0 though. It could
be due to Firefox now bundling 14.5.1, while the gentoo repo has 12.3.2. It
could also be due to the heavy general changes in build system for this
release, or because upstream patches their bundled copy. Needs more
investigation. 'system-harfbuzz' will be disabled via profiles for now.
Rebase of nsTerminator.cpp parts in gcc-lto-pgo-gentoo.patch provided by tt_1.
Remove LTO-Only-enable-LTO-for-Rust-when-complete-build-use.patch and the
equivalent ebuild condition for it. Upstream now enables LTO for Rust via build
system, when clang compiler is detected. Update and append our
gcc-lto-pgo-gentoo.patch to match this previous condition, to only allow
enabling LTO for Rust in the build system if user specifies lto via CFLAGS that
triggers --enable-lto.
Update bmo-1657849-musl-sandbox-sched_setscheduler.patch by tt_1.
Update bmo-2071536-audio_thread_priority-musl-fix.patch by adding the upstream
patch that was merged into Firefox-159.0.
Signed-off-by: Joonas Niilola <juippis@gentoo.org>
Upstream Highway fixed the issue, but V8 has not picked up a Highway
version that includes it, so Node still hits the old headers.
this workaround will stop Highway from compiling RVV, and stop the host
compiler from emitting RVV in non-V8 code. V8 can still emit RVV at runtime
on machines that have it.
-DHWY_DISABLED_TARGETS=HWY_RVV"
Additionally, force the EMU128 (128-bit scalar emulation) for boards
without RVV (VisionFive V2 / JH7110).
This will ensure V8's FixedTag<..., 16> in the JSON stringifier is legal.
Closes: https://bugs.gentoo.org/983938
Signed-off-by: Mike Pagano <mpagano@gentoo.org>
7.1 allows for py3.14 support, which helps solve the releng build
issues recently.
By pure chance, I also found a 1/30 juncture that a race condition
could happen in external/opal-prd. Adding a -j1 to this line only
fixes the chance of opal turning into starburst again.
Closes: https://bugs.gentoo.org/974136
Signed-off-by: Ian Jordan <immoloism@gmail.com>
Signed-off-by: Andreas K. Hüttel <dilfridge@gentoo.org>
No revbump here -- while the file is technically installed, it is only
used for testing, and the timeout affects corner case only.
Closes: https://bugs.gentoo.org/984159
Signed-off-by: Michał Górny <mgorny@gentoo.org>
Rejoice (or grieve) as jemalloc has risen from the ashes!
Its development revived sometime around Q1/Q2 of 2026 and, apparently,
it was so fruitful that two releases have been delivered since then:
5.3.1 and 5.4.0. The latter does not carry the fix for bug 967868, so I
decided to skip it and go right ahead to 5.4.0.
ABI considerations below. Three patches were dropped, one was retained, one updated:
- PR 2338 was merged upstream and is present in 5.4.0 and the
corresponding backport was dropped.
- PR 2312 patch (HPA support) was dropped because it would not apply
on 5.4.0 and we no longer support IA64 (and the issue in original
bug report was prsent on IA64). See upstream issue 2305.
- The fix in 'dont-call-libstcxx-internals' was merged upstream and is
present in 5.4.0 so the patch is dropped.
- AArch64 page size fix has not been upstreamed, so it stays.
- Gentoo fixups is renamed 'dont-override-cflags', the patch is
amended with further removal of unconditional adding of -g flags.
Also, instead of patching the build system to not install HTML doc
to unversioned /usr/share/doc/${PN} (should be ${PF}/html), we just
remove the offender in src_install().
Multilib wrapped header is removed (since its use is discouraged) and
per-ABI headers now live in corresponding libdirs. jemalloc uses
autoconf so we just pass --includedir.
IUSE="+stats" is default enabled now because it has been default
enabled upstream since 2012.
ABI considerations
==================
Quick glance at abidiff (5.3.0-r3 vs 5.4.0) on amd64, on both
ABI_X86="64 32", shows the addition of three new functions and (get
ready!) changing of types of arguments of function pointers used for
tests:
```
Functions changes summary: 0 Removed, 0 Changed, 3 Added functions
Variables changes summary: 0 Removed, 2 Changed, 2 Added variables
3 Added functions:
[A] 'function void free_aligned_sized(void*, size_t, size_t)'
{free_aligned_sized}
[A] 'function void free_sized(void*, size_t)' {free_sized}
[A] 'function void* pvalloc(size_t)' {pvalloc}
2 variables with incompatible sub-type changes:
[C] 'void (* je_test_hooks_arena_new_hook)(...)' was changed to
'void (* je_test_hooks_arena_new_hook)(void)' at test_hooks.c:9:1:
type of variable changed:
in pointed to type 'function type void (...)':
parameter 1 of type '...' was removed
[C] 'void (* je_test_hooks_libc_hook)(...)' was changed to
'void (* je_test_hooks_libc_hook)(void)' at test_hooks.c:12:1:
type of variable changed:
in pointed to type 'function type void (...)':
parameter 1 of type '...' was removed
```
The reason why these test function pointers are even exported is:
The hooks are a little bit screwy -- they're not genuinely exported
in the sense that we want them available to end-users, but we do
want them visible from outside the generated library, so that we can
use them in test code.
-- src/test-hooks.c
So, nothing serious.
See-also: https://github.com/jemalloc/jemalloc/issues/2305
Bug: https://bugs.gentoo.org/967868
Bug: https://bugs.gentoo.org/969223
Signed-off-by: Zurab Kvachadze <zurabid2016@gmail.com>
Part-of: https://codeberg.org/gentoo/gentoo/pulls/2060
Merges: https://codeberg.org/gentoo/gentoo/pulls/2060
Signed-off-by: Sam James <sam@gentoo.org>