bug#68681: 29.2; Re: 66589 Core dump

[Eli Zaretskii]

> Date: Wed, 24 Jan 2024 15:57:37 +1100
> From: Robert Jenssen <robertjenssen@ozemail.com.au>
> 
> I am getting random crashes with emacs. I am not sure if it is related
> to bug #66589 but the backtrace seems to pass through some similar
> territory.

Are these crashes only when you run Emacs under valgrind?  Or does it
also crash when run just under GDB (or outside GDB)?

FWIW, it doesn't look to me similar to bug#66589 at all.

> I ran valgrind with --undef-value-errors=no because at startup emacs
> gives many  "Conditional jump or move depends on uninitialised
> value(s)" warnings. Is this to do with the garbage collector?

Could be.  Did you read the part of etc/DEBUG that describes how to
run Emacs under valgrind?

> ==113842== Syscall param pselect6(sig) points to unaddressable byte(s)
> ==113842==    at 0x6B743CF: pselect64_syscall (pselect.c:34)
> ==113842==    by 0x6B743CF: pselect (pselect.c:56)
> ==113842==    by 0x5D9D33: really_call_select (thread.c:612)
> ==113842==    by 0x564ECF: flush_stack_call_func1 (alloc.c:5207)
> ==113842==    by 0x5DA1F8: flush_stack_call_func (lisp.h:4233)
> ==113842==    by 0x5DA1F8: thread_select (thread.c:644)
> ==113842==    by 0x5F5BB5: xg_select (xgselect.c:205)
> ==113842==    by 0x5BD387: wait_reading_process_output (process.c:5691)

I don't see a problem here: AFAIU, it is completely legitimate to call
pselect with SIGMASK argument a NULL pointer (which we do in this
scenario).

I think we need a backtrace from GDB when Emacs crashes, and not under
valgrind, to make progress with this bug report.

Thanks.