Created attachment 219524 [details] Partial build log Overview: Attempt to build mail/thunderbird from ports tree rev 553560 using poudriere on 12.2-RELEASE failed. Steps to reproduce: 1. Create a 12.2-RELEASE poudriere jail on a 12.2-RELEASE host 2. Run 'poudriere bulk' specifying the jail and a port list including mail/thunderbird Actual result: build failure as shown in attached log snippet Expected result: a successful build System: FreeBSD <hostname> 12.2-RELEASE FreeBSD 12.2-RELEASE r366954 GENERIC amd64 Note: Tried updating /usr/ports/mail/thunderbird to rev 554726 with no change.
Comment on attachment 219524 [details] Partial build log Don't see in log installation of the rust. > Installing icu-67.1,1 Why old icu? Quarterly?
Created attachment 219529 [details] Partial rust build log
A (hopefully useful?) snippet of the rust log is attached now. Note that this is from a build a few days earlier using ports rev 553560. Let me know if you need more - the whole thing is 7.5 MB. > Why old icu? Quarterly? Nope, I don't do quarterly. It's the version I got when I did 'svn update' a couple of weeks ago.
(In reply to Steve Peurifoy from comment #0) > [r122amd64-default-job-01] Extracting harfbuzz-2.7.2_1: .......... done > ===== > ... > ... > [style 0.0.1] cargo:rerun-if-changed=/wrkdirs/usr/ports/mail/thunderbird/work/.build/dist/include/mozilla/ServoTypes.h > [style 0.0.1] cargo:rerun-if-changed=/wrkdirs/usr/ports/mail/thunderbird/work/.build/dist/include/mozilla/dom/VisualViewport.h > error: build failed The actual error is somewhere within "..." (stripped). Provide full log if you can't locate it. (In reply to Steve Peurifoy from comment #3) > Let me know if you need more - the whole thing is 7.5 MB. Compress or upload to a pastebin. Base system has xz(1) and zstd(1). For example, $ fetch http://beefy6.nyi.freebsd.org/data/121amd64-default/554788/logs/firefox-82.0.3,2.log $ zstd --ultra -22 firefox-82.0.3,2.log $ xz -9ek firefox-82.0.3,2.log $ du -Ah firefox*.log* 13M firefox-82.0.3,2.log 233K firefox-82.0.3,2.log.xz 243K firefox-82.0.3,2.log.zst Old logs: http://beefy6.nyi.freebsd.org/data/latest-per-pkg/firefox/
(In reply to Jan Beich from comment #4) > Old logs: http://beefy6.nyi.freebsd.org/data/latest-per-pkg/firefox/ Replace "firefox" with "thunderbird" or any other package you want old build logs for. Maybe compare those logs with local ones to get an idea why the build may fail for you.
Created attachment 219533 [details] thunderbird build log (compressed)
Created attachment 219534 [details] rust build log (compressed)
Compressed full build logs for thunderbird and rust attached.
Comment on attachment 219533 [details] thunderbird build log (compressed) > process didn't exit successfully: `.../webrender-e1eff54c5c57c23c/build-script-build` (signal: 11, SIGSEGV: invalid memory reference) Could be a number of reasons: faulty hardware, kernel bug, broken lang/rust, aggressive optimizations, FFI ABI mismatch, etc. Find where core(5) file was dumped and use lldb(1) (or gdb(1)) to get a stacktrace from it. Make another build and get a stacktrace again. If each stacktrace looks different then it's likely due to hardware/kernel.
OK, appears I'll have to change poudriere.conf to preserve wrkdirs and my build machine is currently attempting to build chromium. Experience suggests that it will be a while, but I'll do this when I can.
So after setting 'SAVE_WRKDIR=yes' and firing off another build, I got... a successful build of thunderbird. For a number of reasons, I seriously doubt that this is hardware. The one difference that stands out to me is that in the two failed instances, builds of other ports were happening on other cores of the machine. In the successful build, it was tbird all by itself. Sample size too small to be conclusive though obviously. Nothing suspicious in dmesg/syslog at the time of the failures other than the actual signal 11 notifications. This will have to wait until I get a repeat occurrence unless someone else trips over the same thing. Thanks.
Since everything has changed and I can no longer duplicate this, it makes no sense to leave it open.