from the build log: 17452733-29:43.81 /var/tmp/portage/www-client/firefox-106.0.4/temp/ccErViv6.s: Assembler messages: 17452823:29:43.81 /var/tmp/portage/www-client/firefox-106.0.4/temp/ccErViv6.s: Error: .size expression for ff_fft_calc_neon does not evaluate to a constant 17452970:29:43.81 /var/tmp/portage/www-client/firefox-106.0.4/temp/ccErViv6.s: Error: .size expression for ff_fft_permute_neon does not evaluate to a constant 17453120-29:43.81 gmake[4]: *** [/var/tmp/portage/www-client/firefox-106.0.4/work/firefox-106.0.4/config/rules.mk:664: fft_neon.o] Error 1 17453250-29:43.81 gmake[4]: Leaving directory '/var/tmp/portage/www-client/firefox-106.0.4/work/firefox_build/media/ffvpx/libavcodec/arm' emerge -pv firefox [ebuild R *] www-client/firefox-106.0.4:rapid::gentoo to /usr/armv7a-unknown-linux-gnueabihf/ USE="X hwaccel system-harfbuzz system-icu system-jpeg system-libevent system-png system-webp -clang -dbus -debug (-eme-free) -geckodriver -gmp-autoupdate -hardened -jack -libproxy -lto -openh264 (-pgo) -pulseaudio -screencast (-selinux) -sndio -system-av1 -system-libvpx -wayland -wifi" CPU_FLAGS_ARM="neon" 0 KiB
Created attachment 827751 [details] output from emerge --info (armv7)
current fix is to set USE="+system-libvpx" I thought this was the fix https://github.com/FFmpeg/FFmpeg/commit/f99d15cca0f6ceedb4f2f38d67a52dacc9556fc7 but sadly it didn't fix it when backported to the firefox-106.0.4 source
Can you report this upstream please (to firefox)?
Any idea if 102 works / doesn't work? I think -system-libvpx is also broken in x86 so the +system-libvpx use flag is always forced on x86. We can do similar for arm.
I found the fix: https://bugzilla.mozilla.org/show_bug.cgi?id=1793289 https://hg.mozilla.org/mozilla-central/rev/81715fef2264
(In reply to Joonas Niilola from comment #4) > Any idea if 102 works / doesn't work? I think -system-libvpx is also broken > in x86 so the +system-libvpx use flag is always forced on x86. We can do > similar for arm. the corresponding bug states that 102esr is not affected by it
I just finished compiling firefox-106.0.5 with the patch, can confirm it fixing and working. Care to add it the 106-patchset? Its still a week for the release of 107.0, and probaly 10-14 days untill the ebuild hits the tree.
Maybe if I get another chance to compile firefox before that... got too much to do anyway. But I do ask myself: why do we carry arm patches when firefox isn't keyworded for arm in gentoo? Maybe you want some overlay that includes all arm-specific patches on top of our ebuild?
(In reply to Joonas Niilola from comment #8) > Maybe if I get another chance to compile firefox before that... got too much > to do anyway. take all the time that you need, I'm all in if you want to focus on supporting llvm-15 > But I do ask myself: why do we carry arm patches when firefox > isn't keyworded for arm in gentoo? keywording is relativ on arm, many many packages actually lack them. I also do my best to catch any arm regression in nightly, therefore its only one arm specific upstream patch at the moment in the patchset, which is from a third_party source (libyuv) - once that third_party patch is fixed it will be down to none. There is one gentoo specific patch about libaom, which is a minimal invasive one liner, and I'd rather get rid of libaom as its painfull to deal with all of its neon breakages. This day will come, but its not today and not tomorrow (see mozbg 1789986)
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=0c755e10851c075aff8dbcdbdc81c014f7df8e50 commit 0c755e10851c075aff8dbcdbdc81c014f7df8e50 Author: Joonas Niilola <juippis@gentoo.org> AuthorDate: 2022-11-08 17:19:39 +0000 Commit: Joonas Niilola <juippis@gentoo.org> CommitDate: 2022-11-08 17:19:39 +0000 www-client/firefox: add a patch to fix building on arm on 106.0.5 Closes: https://bugs.gentoo.org/879855 Signed-off-by: Joonas Niilola <juippis@gentoo.org> www-client/firefox/Manifest | 2 +- www-client/firefox/firefox-106.0.5.ebuild | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
