LeakSanitizer has encountered a fatal error when running PDFium tests |
||||
Issue descriptionHappened a few times randomly: https://build.chromium.org/p/tryserver.client.pdfium/builders/linux_xfa_asan_lsan/builds/5297/steps/embeddertests/logs/stdio https://build.chromium.org/p/client.pdfium/builders/linux_xfa_asan_lsan/builds/2458/steps/embeddertests/logs/stdio https://build.chromium.org/p/client.pdfium/builders/linux_xfa_asan_lsan/builds/2376 (log expired) [----------] Global test environment tear-down [==========] 351 tests from 34 test cases ran. (51741 ms total) [ PASSED ] 351 tests. YOU HAVE 12 DISABLED TESTS Tracer caught signal 11: addr=0x7f5236e33000 pc=0x1061de0 sp=0x7f5235d24d50 ==11342==LeakSanitizer has encountered a fatal error. ==11342==HINT: For debugging, try setting environment variable LSAN_OPTIONS=verbosity=1:log_threads=1 ==11342==HINT: LeakSanitizer does not work under ptrace (strace, gdb, etc) step returned non-zero exit code: 1
,
Dec 11 2017
,
Dec 11 2017
I can't get this to repro locally. My LSAN configuration is definitely working as it catches a leak inserted on purpose.
,
Dec 11 2017
You may have to try many times before it occurs, as it happens on the bots in a flaky manner. The earliest build above was from Nov 29th. Before that, the last time we rolled DEPS for anything build related was in September. I don't know if there is anything to roll back.
,
Dec 11 2017
Is it a full moon tonight? Here's another one: https://build.chromium.org/p/tryserver.client.pdfium/builders/linux_xfa_asan_lsan/builds/5311/steps/embeddertests/logs/stdio
,
Dec 11 2017
This is an LSAN tool failure, rather than a memory leak. Leaks when detected look different in the log, it actually shows the stack trace. I tried 20 times locally running all embedder tests, if we're seeing 3/7 failure today, even if it was "luck", I should have gotten at least a failure.
,
Dec 12 2017
Yes, I thought that was obvious, which is why I CC'd kcc@. Should we consider filing a bug upstream? https://github.com/google/sanitizers/issues/ It happened a few more times today. Maybe we want to get ssh access to our bots, and try running it there with LSAN_OPTIONS set and see if we can repro and gather more details first.
,
Dec 12 2017
If you can give us a reproducer, yes, please file a bug. It's unlikely we'll be able to do anything w/o a reproducer.
,
Dec 14 2017
Should we remove lsan from the trybot commit blockers? I'm having to run trybots 2-3 times before the lsan bot finishes without this error.
,
Dec 14 2017
10/21 failures today. Does a reproducer need to be local? It's fairly easy to reproduce on the bot..
,
Dec 14 2017
I still recommend the plan in the comment 7. If someone gets bot access, they can login to the bots, run with LSAN_OPTIONS set, or get additional guidance from kcc@ on how to get even more info / debug the problem.
,
Feb 6 2018
Hasn't happened for a month, closing. |
||||
►
Sign in to add a comment |
||||
Comment 1 by dsinclair@chromium.org
, Dec 11 2017