Summary: | www-client/firefox-66.0 - cargo: thread 'main' panicked at 'stack backtrace: ??? | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Luke Bratch <luke> |
Component: | Current packages | Assignee: | Mozilla Gentoo Team <mozilla> |
Status: | RESOLVED OBSOLETE | ||
Severity: | normal | CC: | gmiramir, hlein, ivanhoe, jaglover, luke |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
build.log.gz
build-backtrace.log.gz build-nobacktrace.log.gz build-backtrace-success.log.gz backtrace.log build-rust-bin.log.gz Build log firefox-66.0.2-nopanic.log.gz firefox-66.0.2-nopanic-finished.log.gz Successful build using MAKEOPTS=-j1 |
Description
Luke Bratch
2019-03-19 14:27:11 UTC
Created attachment 569768 [details]
build.log.gz
Had to attach build log gzipped due to file size.
> I see the warnings at the start of the log about clang/llvm slots 7 and 8, but I get the same warnings on another box that compiles firefox-66.0 without problem so I don't think that's the problem. Right, as long as this block ends with > * Will use LLVM slot X! you are fine -- no error. I cannot reproduce with - gcc-8.3.0 (you used 8.2.0) - rust-1.32 (same) - cbindgen-0.7.1 (same) - clang:6 (same) - llvm:6 (same) Please re-install rust with symbols enabled. If you are still able to reproduce, show us the error with the following env variables set: RUST_BACKTRACE=1 RUST_LOG=debug Created attachment 569912 [details]
build-backtrace.log.gz
Thanks for your reply.
I have rebuilt Rust with FEATURES="nostrip" and attempted to build Firefox with RUST_BACKTRACE=1 RUST_LOG=debug and it fails with different errors, please see the attached.
Just to confirm, my other machine that works fine (also amd64) works with the same packages including gcc-8.2:
- gcc-8.2.0-r6
- rust-1.32
- cbindgen-0.7.1
- clang:6
- llvm:6
Thanks
Luke
Created attachment 569914 [details]
build-nobacktrace.log.gz
In case it is useful, here is Rust rebuilt with FEATURES="nostrip" and Firefox re-attempted without RUST_BACKTRACE or RUST_LOG set. Same error as in the initial report.
Yes, you cannot build entire firefox with RUST_LOG=debug set. But you can set RUST_LOG=debug when the error occurs and then re-run to get details:
> cd $(portageq get_repo_path / gentoo)/www-client/firefox
> ebuild firefox-66.0.ebuild clean compile
> ...now building will fail...
> RUST_BACKTRACE=1 RUST_LOG=debug ebuild firefox-66.0.ebuild compile
This should give us the details we are looking for.
If not, please show us build.log only with RUST_BACKTRACE=1 set...
Created attachment 569964 [details]
build-backtrace-success.log.gz
In an unexpected twist, continuing the failed build with ebuild compile actually let the compile finish. Now installed and qmerged successfully.
I let it run with RUST_BACKTRACE=1 RUST_LOG=debug anyway and have attached the results.
I have another (very similar) machine that is failing as well, so I will test letting it fail and resume with ebuild compile on that machine with symbol-less rust and without RUST_BACKTRACE or RUST_LOG to see what happens.
*** Bug 680990 has been marked as a duplicate of this bug. *** Created attachment 569972 [details]
backtrace.log
Extracted backtrace.
Are you using rust-bin? If not, can you please install rust-bin in parallel, switch via `eselect rust` and try again? Created attachment 569974 [details]
build-rust-bin.log.gz
Done. Was using dev-lang/rust-1.32.0 before, now using dev-lang/rust-bin-1.32.0:
$ eselect rust list
Available Rust versions:
[1] rust-bin-1.32.0 *
[2] rust-1.32.0
Same panic as before though.
FYI one difference compared to yesterday is that since firefox-66.0-r1 is now in Portage I had to unmask and install media-libs/libaom-1.0.0-r1 due to a failure during ./configure. This log is for firefox-66.0-r1.
I haven't had a chance to test the build/fail/resume trick on a second machine yet as it is switched off at home.
Created attachment 570110 [details]
Build log
Have tried to upgrade in three computers, all fail.
~ $ emerge -pv firefox
These are the packages that would be merged, in order:
Calculating dependencies... done!
[ebuild U ] www-client/firefox-66.0-r1::gentoo [65.0.2::gentoo] USE="gmp-autoupdate hwaccel lto screenshot system-av1%* system-harfbuzz system-icu system-jpeg system-libevent system-sqlite system-webp -bindist -clang -custom-cflags -custom-optimization -dbus -debug -eme-free -geckodriver -hardened -jack (-neon) -pgo -pulseaudio (-selinux) -startup-notification -system-libvpx -test -wayland -wifi" CPU_FLAGS_X86="avx2" L10N="-ach -af -an -ar -as -ast -az -bg -bn-BD -bn-IN -br -bs -ca -cak -cs -cy -da -de -dsb -el -en-GB -en-ZA -eo -es-AR -es-CL -es-ES -es-MX -et -eu -fa -ff -fi -fr -fy -ga -gd -gl -gn -gu -he -hi -hr -hsb -hu -hy -id -is -it -ja -ka -kab -kk -km -kn -ko -lij -lt -lv -mai -mk -ml -mr -ms -nb -nl -nn -or -pa -pl -pt-BR -pt-PT -rm -ro -ru -si -sk -sl -son -sq -sr -sv -ta -te -th -tr -uk -uz -vi -xh -zh-CN -zh-T
(In reply to Luke Bratch from comment #6) > I have another (very similar) machine that is failing as well, so I will > test letting it fail and resume with ebuild compile on that machine with > symbol-less rust and without RUST_BACKTRACE or RUST_LOG to see what happens. I have now done this - it worked. A machine with nothing special done other than resuming the compile once it fails using "ebuild /usr/portage/www-client/firefox/firefox-66.0-r1.ebuild compile" works fine. Also suffered from similar cargo error on firefox 66 builds (but could not find any panic print). Interesting observation: - With "-j8", firefox 66.0.1 built fine for me on my i7-6700K. - With "-j1" afterwards, the same build again fails reproducably (just like the original poster had used "-j1" in MAKEOPTS). Might be an option to try out. Maybe some race condition or dependency issue. Created attachment 571178 [details]
firefox-66.0.2-nopanic.log.gz
firefox-66.0.2 also fails on the same machine, and again can be resumed by issuing an ebuild compile.
Unlike in the initial report, I no longer see a panic, much like what Robert mentions in the previous post.
Attached is the log from when it failed (before resuming).
Created attachment 571180 [details]
firefox-66.0.2-nopanic-finished.log.gz
And here's the finished log from after issuing the ebuild resume and it ultimately succeeding.
Using -j2 instead of -j1 and it builds. First time ever I see parallel build is required for successful build. (In reply to Saul Peebsen from comment #16) > Using -j2 instead of -j1 and it builds. First time ever I see parallel build > is required for successful build. Wild. Same here. Some time ago I had added package.env settings to set MAKEOPTS="-j1" for firefox - I think ~2018-09-10, because 62.x could not compile with my default "-j10". But no version of www-client/firefox-66.x would build with -j1. Upped it to -j4 and it worked! In case it matters: sys-devel/clang-8.0.0 sys-devel/gcc-8.3.0 sys-devel/llvm-8.0.0 dev-lang/rust-1.33.0 Created attachment 571684 [details] Successful build using MAKEOPTS=-j1 The claim that you cannot build current firefox with MAKEOPTS=-j1 is wrong: www-client/firefox-66.0.2 with USE="abi_x86_64 amd64 cpu_flags_x86_avx2 elibc_glibc gmp-autoupdate kernel_linux l10n_de lto pgo screenshot system-av1 system-harfbuzz system-icu system-jpeg system-libevent system-libvpx system-webp userland_GNU" real 203m33,001s user 239m41,843s sys 11m40,439s GCC 8.3.0 and rust-1.33 was used: > Portage 2.3.62 (python 3.7.3-final-0, default/linux/amd64/17.0, gcc-8.3.0, glibc-2.29-r1, 4.19.30-gentoo x86_64) > ================================================================= > System uname: Linux-4.19.30-gentoo-x86_64-AMD_Ryzen_Threadripper_2950X_16-Core_Processor-with-gentoo-2.6 > KiB Mem: 131953288 total, 123205776 free > KiB Swap: 16777212 total, 16777212 free > Timestamp of repository gentoo: Tue, 02 Apr 2019 12:44:16 +0000 > Head commit of repository gentoo: 0553bb476d25269069361ef972888fe53ecac2e0 > > sh bash 5.0_p3-r1 > ld GNU ld (Gentoo 2.32 p1) 2.32.0 > app-shells/bash: 5.0_p3-r1::gentoo > dev-lang/perl: 5.28.0::gentoo > dev-lang/python: 2.7.16::gentoo, 3.6.8::gentoo, 3.7.3::gentoo > dev-util/cmake: 3.14.1::gentoo > sys-apps/baselayout: 2.6-r1::gentoo > sys-apps/openrc: 0.41.2::gentoo > sys-apps/sandbox: 2.17::gentoo > sys-devel/autoconf: 2.13-r1::gentoo, 2.69-r4::gentoo > sys-devel/automake: 1.16.1-r1::gentoo > sys-devel/binutils: 2.32::gentoo > sys-devel/gcc: 8.3.0::gentoo > sys-devel/gcc-config: 2.0::gentoo > sys-devel/libtool: 2.4.6-r5::gentoo > sys-devel/make: 4.2.1-r4::gentoo > sys-kernel/linux-headers: 5.0-r1::gentoo (virtual/os-headers) > sys-libs/glibc: 2.29-r1::gentoo Wrong is a strong word. Anyhow, something has been changed and FF builds with -j1 now, I can confirm. I had similar problems, once a panic and then none, and I did an ebuild … compile and then install and then merge. This was with J1, gcc 7.3 and (I think) rust 1.33. I hit this today on an armv7a (cubieboard2). MAKEOPTS=-j1 and as others have reported, manually restarting with ebuild seems to have gotten past it (I wont know for hours if it finishes). Clang-8, rust-bin-1.34.2, building firefox-68.1.0 I can confirm this is occurring (again/still?) on amd64 for firefox-68.1.0, clang-8, rust-1.34.2. Could it be a resource issue? I vaguely recall some rust build on my machine failed with a weird panic and something like "received kill signal" until I hiked my swap way up. emerge --info: https://pastebin.com/jRpAnXxw The version this bug was reported against was removed. If you are still experiencing the issue with >=www-client/firefox-78.3.1 or newer please file a new bug or re-open. |