Re: SIGABRT on 7.4.5

Joe Conway <mail@joeconway.com> writes:
> I have been trying to troubleshoot a PL/R related issue on-and-off for a 
> few weeks now, but this morning ran into what appears to be a more 
> general issue related to 7.4.5 on x86_64. A full backtrace is below, and 
> in it you can see that PL/R is never reached (but PL/pgSQL is). What is 
> really odd is that the abort appears to happen because 
> CacheMemoryContext is invalid.

It could hardly not have been set up yet, so I'm betting that this
indicates a memory-stomp problem somewhere that happens to be hitting
part of the CacheMemoryContext structure.

> I do *not* see this same failure on a 
> different machine with a 32 bit build

Not too surprising; could be related to struct sizes, padding, trying to
cram a 64-bit pointer into a 32-bit field, who knows what.

> Any ideas? I'll request permission to provide the full script to 
> reproduce this to someone offlist if it will help.

The odds are that it couldn't be reproduced except on the same
architecture.  I do have access to some x86_64 machines at Red Hat,
though, so I can take a look if you can send me a test case.
        regards, tom lane