Re: [BUGS] BUG #5412: test case produced, possible race condition. - Mailing list pgsql-hackers

Rusty Conover <rconover@infogears.com> writes:

On Apr 14, 2010, at 2:31 PM, Tom Lane wrote:

There is another slightly odd thing here, which is that the stack trace

Rusty provided clearly shows the crash occurring during processing of a

local relcache invalidation message for the truncated relation.

The test case program was the smallest thing I could write to reproduce the problem consistently on my machine, but I couldn't reproduce it consistently on other machines and architectures.  I'm glad Heikki was able to also see the crash on his hardware.  I can take Heikki's patch back out and get a new stack trace from the test program if that would be useful to you.

Yeah, I'm curious to know if the stack trace is the same for crashes in
the watered-down test case.  AFAICS, relcache invals triggered by the
TRUNCATE itself ought to be serviced before we get out of the TRUNCATE;
and after that, the xact is holding exclusive lock on the table so
there's no way that we should get a remote inval on the specific relation
(not to mention that the trace is clearly about a local inval not an
incoming remote inval).  So I don't understand why the trace says that
it's happening during a subsequent INSERT.

Mind you, Heikki's patch looks like a good idea in any case; but I'm
not sure that there isn't something else going on here.

regards, tom lane