Open sam-mccall opened 3 years ago
Happened again in http://45.33.8.238/linux/50888/step_9.txt
Seems to fail about once a month, which isn't all that rarely.
There's also bug 50773.
Maybe it's time for a look into the threading bits here?
Happened again in this build http://45.33.8.238/linux/48398/step_9.txt
Same stacks as in comment 1.
Before I ran gdb:
thakis@dotc:~/src/hack$ ps aux | grep clangd ... thakis 2313354 0.0 0.0 218744 32540 pts/1 Sl+ 10:01 0:00 /usr/local/google/home/thakis/src/llvm-project/out/gn/obj/clang-tools-extra/clangd/unittests/./ClangdTests --gtest_filter=TUSchedulerTests.Cancellation
I then ran kill 2313354
(which is what in the end caused the test to fail, but only after the test had been running for > 2h).
(this was the build http://45.33.8.238/linux/45981/summary.html )
Happened again today (but not since then I think, so somewhat rare).
I attached to the hanging process in gdb before killing the process:
Attaching to process 2313354 [New LWP 2313397] [New LWP 2313398] [Thread debugging using libthread_db enabled] Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". futex_wait_cancelable (private=0, expected=0, futex_word=0x7ffcd0bbf8c8) at ../sysdeps/nptl/futex-internal.h:186 186 ../sysdeps/nptl/futex-internal.h: No such file or directory. (gdb) bt
()
(gdb) threads info Undefined command: "threads". Try "help". (gdb) info threads Id Target Id Frame
186 in ../sysdeps/nptl/futex-internal.h (gdb) bt
(gdb) thread 3 [Switching to thread 3 (Thread 0x7f87b439b700 (LWP 2313398))]
186 in ../sysdeps/nptl/futex-internal.h (gdb) bt
Let me know if there's additional tings I should capture next time.
This is another instance of this hang: https://reviews.llvm.org/D122251#3528961
(But haven't seen clangd unit tests hang in a very long time before that -- less frequently than once / month for sure.)
Saw this once locally last week, and once today here: http://45.33.8.238/macm1/37396/step_9.txt
(Both times on an M1 mac.)
Extended Description
From Nico on llvm/llvm-project#48342
(I don't think it's the same cause, but definitely a bug)