Re: Trouble with plpgsql on 7.4.6

On Sat, 20 Nov 2004 11:17:48 -0500
Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> "D'Arcy J.M. Cain" <darcy(at)druid(dot)net> writes:
> > I thought that this would have sent everything to both the log and
> > the screen but I found that the syslog has much more detail. I have
> > attached that output.
>
> We still need to look at the stderr output. All this says is that
> you're getting repeated abort()s. It's unlikely that anything would
> be coded to abort() without emitting any gripe at all --- but the
> gripe is not appearing in syslog output, so stderr is the next place
> to look.

The stderr was in the previous message. No gripes there either other
than in the startup after the failure.

> Also see about getting a stack trace from one of the core dumps.

I did look at the core file and here is what I saw:

#0 0x483cafeb in kill () from /usr/lib/libc.so.12
#1 0x483cd0af in __libc_mutex_catchall_stub (m=1212478892)
at /usr/src/lib/libc/thread-stub/thread-stub.c:112
#2 0x4843f0f7 in free (ptr=<incomplete type>)
at /usr/src/lib/libc/stdlib/malloc.c:1149
#3 0x081b3efc in AllocSetDelete (context=<error type>) at aset.c:464
#4 0x081b468a in MemoryContextDelete (context=<error type>) at
#mcxt.c:192

In fact it is calling the thread stubs but the applicatino is threaded.
At least, that's what I see from this code in NetBSD.

#define CHECK_NOT_THREADED_ALWAYS() \
do { \
if (__isthreaded) \
DIE(); \
} while (/*CONSTCOND*/0)

#if 1
#define CHECK_NOT_THREADED() CHECK_NOT_THREADED_ALWAYS()
#else
#define CHECK_NOT_THREADED() /* nothing */
#endif

I am going to follow up with the NetBSD team but I pass this on in case
it suggests anything. Note that this still only causes problems on
7.4.6 and not under 7.4.3.

--
D'Arcy J.M. Cain <darcy(at)druid(dot)net> | Democracy is three wolves
http://www.druid.net/darcy/ | and a sheep voting on
+1 416 425 1212 (DoD#0082) (eNTP) | what's for dinner.