Under GCC 10, this package seems to break a bunch of addons if you use either '-O2 -flto' or '-O3' to build. This affects the USE=lto and USE=pgo USE flags, as the latter implicitly enables '-O3'. The issue is most notably manifested as violentmonkey not being able to run any userscripts, and seemingly clearing all userscripts upon restart. Checking about:debugging yields the following error, which doesn't appear when the browser's built with just '-O2': https://files.catbox.moe/vodnfn.png Is this a supported configuration by upstream, or do they only support Clang? If so, should this be reported upstream?
Relevant GentooLTO issue: https://github.com/InBetweenNames/gentooLTO/issues/650
According to your linked Github issue this is not related to LTO, it's caused by -O3 (which will also get enabled when using LTO).
Please show us `emerge --info www-client/firefox | tail -n 11`
`emerge --info sys-devel/gcc | tail -n 11` so we know which GCC patch set you are using.
I could've sworn this was reproducible with USE=lto on -O2, but I guess not. These are the minimum settings I could reproduce the issue with: $ emerge --info www-client/firefox | tail -n 11 ================================================================= Package Settings ================================================================= www-client/firefox-78.5.0::gentoo was built with the following: USE="dbus eme-free openh264 pulseaudio system-av1 system-harfbuzz system-icu system-jpeg system-libevent system-libvpx system-webp -clang -debug -geckodriver -gmp-autoupdate -hardened -hwaccel -jack -lto -pgo (-screencast) (-selinux) -wayland -wifi" ABI_X86="(64)" L10N="ca es-ES -ach -af -an -ar -ast -az -be -bg -bn -br -bs -ca-valencia -cak -cs -cy -da -de -dsb -el -en-CA -en-GB -eo -es-AR -es-CL -es-MX -et -eu -fa -ff -fi -fr -fy -ga -gd -gl -gn -gu -he -hi -hr -hsb -hu -hy -ia -id -is -it -ja -ka -kab -kk -km -kn -ko -lij -lt -lv -mk -mr -ms -my -nb -ne -nl -nn -oc -pa -pl -pt-BR -pt-PT -rm -ro -ru -si -sk -sl -son -sq -sr -sv -ta -te -th -tl -tr -trs -uk -ur -uz -vi -xh -zh-CN -zh-TW" CFLAGS="-march=znver1 -O3 -pipe -w" CXXFLAGS="-march=znver1 -O3 -pipe -w" LDFLAGS="-Wl,-O1 -Wl,--as-needed -march=znver1 -fgraphite-identity -floop-nest-optimize -fdevirtualize-at-ltrans -fno-semantic-interposition -fuse-linker-plugin -pipe -w --param=lto-max-streaming-parallelism=2 -Wl,--compress-debug-sections=zlib -Wl,-rpath=/usr/lib64/firefox,--enable-new-dtags" $ emerge --info sys-devel/gcc | tail -n 11 ================================================================= Package Settings ================================================================= sys-devel/gcc-10.2.0-r3::gentoo was built with the following: USE="(cxx) fortran graphite lto (multilib) nls nptl openmp pch pgo (pie) sanitize vtv (-ada) -d -debug -doc (-fixed-point) -go (-hardened) -jit (-libssp) -objc -objc++ -objc-gc -ssp -systemtap -test -vanilla -zstd" ABI_X86="(64)" CFLAGS="-march=znver1 -pipe -w -Wl,-O1 -Wl,--as-needed -O2" CXXFLAGS="-march=znver1 -pipe -w -Wl,-O1 -Wl,--as-needed -O2" FEATURES="strict preserve-libs multilib-strict unknown-features-warn usersandbox ebuild-locks network-sandbox merge-sync userpriv assume-digests binpkg-dostrip config-protect-if-modified ipc-sandbox sfperms news distlocks userfetch xattr binpkg-docompress protect-owned qa-unresolved-soname-deps unmerge-logs buildpkg sandbox unmerge-orphans binpkg-logs fixlafiles parallel-fetch binpkg-multi-instance pid-sandbox usersync" LDFLAGS="-Wl,-O1 -Wl,--as-needed -march=znver1 -pipe -w"
It's not just ESR and not just add-ons affected.
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=2fed8b79dc9c6284243ef81f7d32e8e528728a95 commit 2fed8b79dc9c6284243ef81f7d32e8e528728a95 Author: Thomas Deutschmann <whissi@gentoo.org> AuthorDate: 2020-12-14 00:15:24 +0000 Commit: Thomas Deutschmann <whissi@gentoo.org> CommitDate: 2020-12-14 00:17:05 +0000 profiles/base/package.use.force: force USE=clang for firefox/thunderbird Bug: https://bugs.gentoo.org/758446 Signed-off-by: Thomas Deutschmann <whissi@gentoo.org> profiles/base/package.use.force | 8 ++++++++ 1 file changed, 8 insertions(+) https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=4f3afb07a1550e4133182c736491f72a3d06c09f commit 4f3afb07a1550e4133182c736491f72a3d06c09f Author: Thomas Deutschmann <whissi@gentoo.org> AuthorDate: 2020-12-14 00:10:00 +0000 Commit: Thomas Deutschmann <whissi@gentoo.org> CommitDate: 2020-12-14 00:17:05 +0000 mail-client/thunderbird: use Clang by default Bug: https://bugs.gentoo.org/758446 Package-Manager: Portage-3.0.12, Repoman-3.0.2 Signed-off-by: Thomas Deutschmann <whissi@gentoo.org> mail-client/thunderbird/thunderbird-78.5.1.ebuild | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=5b92571ae9ad649ad61e9886500b4feedb8de915 commit 5b92571ae9ad649ad61e9886500b4feedb8de915 Author: Thomas Deutschmann <whissi@gentoo.org> AuthorDate: 2020-12-14 00:06:31 +0000 Commit: Thomas Deutschmann <whissi@gentoo.org> CommitDate: 2020-12-14 00:17:04 +0000 www-client/firefox: use Clang by default Bug: https://bugs.gentoo.org/758446 Package-Manager: Portage-3.0.12, Repoman-3.0.2 Signed-off-by: Thomas Deutschmann <whissi@gentoo.org> www-client/firefox/firefox-78.5.0.ebuild | 2 +- www-client/firefox/firefox-83.0.ebuild | 2 +- www-client/firefox/firefox-84.0_rc2.ebuild | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-)
Is forcing
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=10b8a24ed8ea22f75979d312ccc4f6bc8dd986e6 commit 10b8a24ed8ea22f75979d312ccc4f6bc8dd986e6 Author: Thomas Deutschmann <whissi@gentoo.org> AuthorDate: 2020-12-21 21:36:46 +0000 Commit: Thomas Deutschmann <whissi@gentoo.org> CommitDate: 2020-12-21 23:40:54 +0000 profiles/base/package.use.force: no longer force USE=clang for Mozilla Bug: https://bugs.gentoo.org/758446 Signed-off-by: Thomas Deutschmann <whissi@gentoo.org> profiles/base/package.use.force | 9 --------- 1 file changed, 9 deletions(-) https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=f373b422ad975add8b59468f781d10b3b2106b3f commit f373b422ad975add8b59468f781d10b3b2106b3f Author: Thomas Deutschmann <whissi@gentoo.org> AuthorDate: 2020-12-21 21:36:02 +0000 Commit: Thomas Deutschmann <whissi@gentoo.org> CommitDate: 2020-12-21 23:40:53 +0000 dev-lang/spidermonkey: synchronize with www-client/firefox Bug: https://bugs.gentoo.org/758446 Package-Manager: Portage-3.0.12, Repoman-3.0.2 Signed-off-by: Thomas Deutschmann <whissi@gentoo.org> dev-lang/spidermonkey/spidermonkey-78.6.0.ebuild | 9 +++++++++ 1 file changed, 9 insertions(+) https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=7b374e16a74185d720c52c887d746bc2cc1206cd commit 7b374e16a74185d720c52c887d746bc2cc1206cd Author: Thomas Deutschmann <whissi@gentoo.org> AuthorDate: 2020-12-21 21:23:09 +0000 Commit: Thomas Deutschmann <whissi@gentoo.org> CommitDate: 2020-12-21 23:40:53 +0000 mail-client/thunderbird: synchronize with www-client/firefox Bug: https://bugs.gentoo.org/758446 Package-Manager: Portage-3.0.12, Repoman-3.0.2 Signed-off-by: Thomas Deutschmann <whissi@gentoo.org> mail-client/thunderbird/thunderbird-78.6.0.ebuild | 7 +++++++ 1 file changed, 7 insertions(+) https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=09886ee677ea97b80f1469dc7534ff6aea6247cb commit 09886ee677ea97b80f1469dc7534ff6aea6247cb Author: Thomas Deutschmann <whissi@gentoo.org> AuthorDate: 2020-12-21 21:18:14 +0000 Commit: Thomas Deutschmann <whissi@gentoo.org> CommitDate: 2020-12-21 23:40:52 +0000 www-client/firefox: disable loop vectorization Loop vectorization, -ftree-loop-vectorize, which will be auto-enabled when using -O3, is broken in GCC 10+. Many thanks to Dave Hughes (xaero) for running countless tests to find the issue. Bug: https://bugs.gentoo.org/758446 Package-Manager: Portage-3.0.12, Repoman-3.0.2 Signed-off-by: Thomas Deutschmann <whissi@gentoo.org> www-client/firefox/firefox-78.6.0.ebuild | 7 +++++++ www-client/firefox/firefox-84.0.ebuild | 7 +++++++ 2 files changed, 14 insertions(+)
so, is this a gcc-10 bug? maybe you want to cc toolchain then?
(In reply to tt_1 from comment #10) > so, is this a gcc-10 bug? maybe you want to cc toolchain then? It was debugged with slyfox, so they are aware. But we can do it anyway.
I was not able to reproduce the failure locally on a clean profile. Generally if -ftree-loop-vectorize helps it usually hints at aliasing bugs in the program (and probably not in a complier). Having a smll deterministic reproducer of a bug would help to nail down what is at fault.
Since the workaround commits from comment #9 make the build work with gcc-10, I'm removing this from the gcc-10 stable blockers.
I'm pretty sure it works now with versions we have in tree.
