Hiroshi Inoue <Inoue(at)tpf(dot)co(dot)jp> writes:
>> I suspect you will find that these crashes occur during the window just
>> after
> Sorry what does 'just after' mean ?
> Isn't it during the semop() ?
>> the semop() call in IpcSemaphoreLock() --- see the comment
If an interrupt during the semop led to a crash, it would be easy to
reproduce. I suspect that the crash condition arises only when the
interrupt occurs in a narrow time window ... such as the few
instructions just before or just after the semop call. It's just a
hunch though.
regards, tom lane