Trying to build firefox-116, w/ mold linker current fails: 1:10.70 error: failed to run custom build command for `proc-macro2 v1.0.59` 1:10.70 note: To improve backtraces for build dependencies, set the CARGO_PROFILE_RELEASE_BUILD_OVERRIDE_DEBUG=true environment variable to enable debug information generation. 1:10.70 Caused by: 1:10.70 process didn't exit successfully: `/var/tmp/portage/www-client/firefox-116.0/work/firefox_build/instrumented/release/build/proc-macro2-3bec9dfcd1853c93/build-script-build` (exit status: 127) 1:10.70 --- stderr 1:10.70 /var/tmp/portage/www-client/firefox-116.0/work/firefox_build/instrumented/release/build/proc-macro2-3bec9dfcd1853c93/build-script-build: error while loading shared libraries: /var/tmp/portage/www-client/firefox-116.0/work/firefox_build/instrumented/release/build/proc-macro2-3bec9dfcd1853c93/build-script-build: DT_RELR without GLIBC_ABI_DT_RELR dependency Might be due to https://github.com/rui314/mold/issues/653
On further debugging (and manually building proc-macro2 with mold), it seems like the firefox provided cargo linker is the issue passed via -C linker=/var/tmp/portage/www-client/firefox-116.0/work/firefox-116.0/build/cargo-linker ``` CARGO=/usr/lib/rust/1.71.0/bin/cargo-1.71.0 CARGO_CRATE_NAME=build_script_build CARGO_MANIFEST_DIR=/var/tmp/portage/www-client/firefox-116.0/work/firefox-116.0/third_party/rust/proc-macro2 CARGO_PKG_AUTHORS='David Tolnay <dtolnay@gmail.com>:Alex Crichton <alex@alexcrichton.com>' CARGO_PKG_DESCRIPTION='A substitute implementation of the compiler'\''s `proc_macro` API to decouple token-based libraries from the procedural macro use case.' CARGO_PKG_HOMEPAGE='' CARGO_PKG_LICENSE='MIT OR Apache-2.0' CARGO_PKG_LICENSE_FILE='' CARGO_PKG_NAME=proc-macro2 CARGO_PKG_README=README.md CARGO_PKG_REPOSITORY='https://github.com/dtolnay/proc-macro2' CARGO_PKG_RUST_VERSION=1.31 CARGO_PKG_VERSION=1.0.59 CARGO_PKG_VERSION_MAJOR=1 CARGO_PKG_VERSION_MINOR=0 CARGO_PKG_VERSION_PATCH=59 CARGO_PKG_VERSION_PRE='' LD_LIBRARY_PATH='/var/tmp/portage/www-client/firefox-116.0/work/firefox_build/instrumented/release/deps:/usr/lib/rust/1.71.0/lib' /usr/bin/rustc --crate-name build_script_build --edition=2018 /var/tmp/portage/www-client/firefox-116.0/work/firefox-116.0/third_party/rust/proc-macro2/build.rs --error-format=json --json=diagnostic-rendered-ansi,artifacts,future-incompat --crate-type bin --emit=dep-info,link -C opt-level=1 -C embed-bitcode=no --cfg 'feature="default"' --cfg 'feature="proc-macro"' -C metadata=3bec9dfcd1853c93 -C extra-filename=-3bec9dfcd1853c93 --out-dir /var/tmp/portage/www-client/firefox-116.0/work/firefox_build/instrumented/release/build/proc-macro2-3bec9dfcd1853c93 -C linker=/var/tmp/portage/www-client/firefox-116.0/work/firefox-116.0/build/cargo-linker -L dependency=/var/tmp/portage/www-client/firefox-116.0/work/firefox_build/instrumented/release/deps --cap-lints warn` ``` Omitting `-C linker=/var/tmp/portage/www-client/firefox-116.0/work/firefox-116.0/build/cargo-linker` works correctly
Thanks for reporting and your investigations. I just had the same issue for AMD64, and initially had no idea for the cause. Have not suspected mold as the reason, as mold was working fine before with Firefox. sys-devel/mold-2.0.0-r1. Will try with old ldd and will report is that should fail as well, but I think that will work.
Hmm is this with clang+mold or gcc+mold? I think at least elfhack needs to be disabled if gcc+mold is detected.
I have clang use flag enabled in make.conf. ``` $ emerge -pv firefox These are the packages that would be merged, in order: Calculating dependencies... done! Dependency resolution took 3.08 s. [ebuild U ~] www-client/firefox-116.0:rapid::gentoo [115.0.2:rapid::gentoo] USE="X clang dbus gmp-autoupdate hwaccel jumbo-build lto openh264 pgo system-av1 system-harfbuzz system-icu system-jpeg system-libevent system-libvpx system-png system-webp telemetry wayland -debug -eme-free -geckodriver -hardened -jack -libproxy -pulseaudio -screencast (-selinux) -sndio (-system-python-libs) (-valgrind) -wifi" L10N="de -ach -af -an -ar -ast -az -be -bg -bn -br -bs -ca -ca-valencia -cak -cs -cy -da -dsb -el -en-CA -en-GB -eo -es-AR -es-CL -es-ES -es-MX -et -eu -fa -ff -fi -fr -fur -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 -sc -sco -si -sk -sl -son -sq -sr -sv -szl -ta -te -th -tl -tr -trs -uk -ur -uz -vi -xh -zh-CN -zh-TW" 0 KiB Total: 1 package (1 upgrade), Size of downloads: 0 KiB salewski@hx90 ~ $ cat /etc/portage/package.use/firefox www-client/firefox hwaccel openh264 system-av1 system-harfbuzz system-icu system-jpeg system-libevent system-libvpx system-png system-python-libs system-webp ``` Disabling mold fails now also, I think ldd is not available for latest llvm. I was not aware that the linker is coupled to llvm, maybe I would have to reinstall llvm with mold disabled. But I will test without clang use flag tomorrow first. Maybe I should mention that I am using the Gentoo-LTO overlay, but I think that makes no difference for Firefox. * Checking for at least 13500 MiB disk space at "/var/tmp/portage/www-client/firefox-116.0/temp" ... [ ok ] * sys-devel/lld:16 is missing! Cannot use LLVM slot 16 ... * virtual/rust:0/llvm-15 is missing! Cannot use LLVM slot 15 ... * sys-devel/clang:14 is missing! Cannot use LLVM slot 14 ...
Builds fine w/ clang + lld, gcc + mold. Fails w/ clang + mold (with out without elfhack)
Unfortunately latest upstream elf-hack patches don't fix this, so it's not elf-hack related. (In reply to Amit Prakash Ambasta from comment #1) > > Omitting `-C > linker=/var/tmp/portage/www-client/firefox-116.0/work/firefox-116.0/build/ > cargo-linker` works correctly What's the proper fix for this? Why can't I find any upstream bug about it, is this Gentoo-related?
Definitely an issue in upstream. Filed an upstream bug as well at https://bugzilla.mozilla.org/show_bug.cgi?id=1847697
(In reply to Amit Prakash Ambasta from comment #7) > Definitely an issue in upstream. Filed an upstream bug as well at > https://bugzilla.mozilla.org/show_bug.cgi?id=1847697 -> https://github.com/rui314/mold/issues/653#issuecomment-1670274638. I can backport the fix to mold, I suppose.
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=0b8e1cc4c541041f00338d897d34fabe66ae87b8 commit 0b8e1cc4c541041f00338d897d34fabe66ae87b8 Author: Sam James <sam@gentoo.org> AuthorDate: 2023-08-09 03:10:49 +0000 Commit: Sam James <sam@gentoo.org> CommitDate: 2023-08-09 03:11:03 +0000 sys-devel/mold: backport DT_RELR fix Bug: https://bugs.gentoo.org/911591 Signed-off-by: Sam James <sam@gentoo.org> .../mold/files/mold-2.0.0-DT_RELR-dependency.patch | 161 +++++++++++++++++++++ sys-devel/mold/mold-2.0.0-r2.ebuild | 94 ++++++++++++ 2 files changed, 255 insertions(+)
(In reply to Sam James from comment #8) > (In reply to Amit Prakash Ambasta from comment #7) > > Definitely an issue in upstream. Filed an upstream bug as well at > > https://bugzilla.mozilla.org/show_bug.cgi?id=1847697 > > -> https://github.com/rui314/mold/issues/653#issuecomment-1670274638. > > I can backport the fix to mold, I suppose. Done. We should still backport the Firefox workaround though: https://bugzilla.mozilla.org/show_bug.cgi?id=1847697#c4.
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=b865b575dde25d3a7abfd7717444a167f4d2e631 commit b865b575dde25d3a7abfd7717444a167f4d2e631 Author: Joonas Niilola <juippis@gentoo.org> AuthorDate: 2023-08-09 06:42:19 +0000 Commit: Joonas Niilola <juippis@gentoo.org> CommitDate: 2023-08-09 06:42:19 +0000 www-client/firefox: fix 116 with mold linker Closes: https://bugs.gentoo.org/911591 Signed-off-by: Joonas Niilola <juippis@gentoo.org> www-client/firefox/Manifest | 2 +- www-client/firefox/firefox-116.0.ebuild | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
Seems to work now. Thanks all!
