Re: [GENERAL] Still big problems with pg_dump!

Andrew Sullivan <andrew(at)libertyrms(dot)info> writes:
> On Tue, Sep 17, 2002 at 04:25:45PM +0200, Wim wrote:
>> ERROR: AllocSetFree: cannot find block containing chunk 4c5ad0
>>
>> Postgres is running on solaris 8...

> Someone else suggested that this would not be the error when you have
> written bad data to the disk (I thought you could have this if the
> controller was flakey and wrote bad data in the past. Maybe I'm
> wrong. Probably).

Actually, what it looks like to me is a memory clobber; I don't think
bad data on disk would be likely to lead to this particular type of
failure. But writing one byte too many into a string, and thereby
zeroing the high-order byte of an adjacent pointer, could lead to
exactly this message when we later try to pfree() the pointer.

I am wondering if Wim is running into that same Solaris snprintf() bug
that we discovered awhile back --- it was not clear if the bug still
exists in Solaris 8, but the symptoms sure match. See
http://archives.postgresql.org/pgsql-bugs/2002-07/msg00059.php

It would be useful to see a stack traceback from the point of the error,
if possible.

regards, tom lane