On Fri, Mar 07, 2025 at 11:41:38AM -0500, Andres Freund wrote:
> On 2025-03-07 10:36:35 -0600, Nathan Bossart wrote:
>> On Fri, Mar 07, 2025 at 11:32:28AM -0500, Andres Freund wrote:
>> > Is it possible that the signal number we're getting called for is above
>> > PG_NSIG? That'd explain why the source value is something fairly random?
>> >
>> > ISTM that we should add an Assert() to wrapper_handler() that ensures that
>> > the
>> > signal arg is below PG_NSIG.
>>
>> We have such an assertion in pqsignal() before we install wrapper_handler
>> for anything. Is there another way it could be getting called with a
>> different signo?
>
> Who the hell knows :).
>
> One potential way would be that we got SIGNAL_ARGS wrong for the platform and
> are interpreting some random thing as the signal number. Or something went
> wrong in the windows signal emulation code. Or ...
>
> It seems cheap insurance to add it both places.
Good enough for me. I'll commit/back-patch to v17 the attached soon.
--
nathan
diff --git a/src/port/pqsignal.c b/src/port/pqsignal.c
index 5dd8b76bae8..79b50486175 100644
--- a/src/port/pqsignal.c
+++ b/src/port/pqsignal.c
@@ -87,6 +87,8 @@ wrapper_handler(SIGNAL_ARGS)
{
int save_errno = errno;
+ Assert(postgres_signal_arg < PG_NSIG);
+
#ifndef FRONTEND
/*
