Re: exit code -1073741819

"Shuo Liu" <delphet(at)excite(dot)com> writes:
> The log shows the following message:

> CurTransactionContext: 0 total in 0 blocks; 0 free (0 chunks); 0 used
> ExecutorState: 122880 total in 4 blocks; 1912 free (9 chunks); 120968 used
> ExprContext: 8192 total in 1 blocks; 8176 free (0 chunks); 16 used
> ExprContext: 8192 total in 1 blocks; 8176 free (0 chunks); 16 used
> ExprContext: 8192 total in 1 blocks; 8000 free (1 chunks); 192 used
> ExprContext: 8192 total in 1 blocks; 8000 free (1 chunks); 192 used
> ExprContext: 8192 total in 1 blocks; 8096 free (1 chunks); 96 used
> SPI Exec: 0 total in 0 blocks; 0 free (0 chunks); 0 used
> SPI Proc: 8192 total in 1 blocks; 2616 free (0 chunks); 5576 used
> ExecutorState: 57344 total in 3 blocks; 35776 free (7 chunks); 21568 used
> ExprContext: 0 total in 0 blocks; 0 free (0 chunks); 0 used

The above is a fragment of a memory stats dump, which normally would
only be emitted if you had an out-of-memory situation. However the
part of it that you've shown doesn't indicate any particularly heavy
memory usage. Were there any large numbers in the preceding similarly-
formatted lines? Was there anything possibly relevant in the log
entries before that?

> 2007-07-10 12:25:57 LOG: server process (PID 2004) exited with exit code -1073741819

I suppose you're on Windows? This is what is currently printed for an
Access Violation trap on Windows. The fact that it came out partway
through a stats dump is pretty suspicious; it suggests that the trap
happened as a result of trying to scan the memory context bookkeeping
data, which implies a memory clobber of some sort.

So you're looking at a bug, but there's much too little data here to
guess what the bug is. Can you get a debugger stack trace? Or put
together a self-contained test case for someone else to poke at?

Actually the *first* thing to do is make sure you are up to date on
both Postgres and PostGIS versions. No sense in spending a lot of time
chasing down a bug if it's already been fixed.

regards, tom lane