https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103985
--- Comment #5 from Dr. Thomas Orgis ---
OK, I tested with gcc-12-20220109. I repeated the test job, got 17102
successful runs and no crash. So one can hope that this particular one has been
fixed at some point. All good? Is the specific change
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103985
--- Comment #4 from Dr. Thomas Orgis ---
I can try to roll a build and squeeze some test runs in. Naturally, this will
take a few days to get some statistical confidence and it depends on cluster
load, too.
So a snapshot of 12 would be
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103985
--- Comment #3 from anlauf at gcc dot gnu.org ---
Can you try gcc-11 or gcc-12?
There were some fixes for asynchronous I/O (pr100352) touching the locking
mechanism.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103985
--- Comment #2 from Dr. Thomas Orgis ---
You mean every call to a write(), even to a buffer variable, should be a thread
synchronization point? That would mean that I swiftly should get rid of those
in the performance-critical paths. I hoped
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103985
--- Comment #1 from Andrew Pinski ---
There could be a lock missing when it comes to the write. I remember in the
past there were some issues with writes not being thread safe (maybe they still
are not).