Re: SRF memory mgmt patch (was [HACKERS] Concern about

Tom Lane wrote:
> First let's see if we can figure out why the code is failing to fail
> as it stands. The fact that it's not dumping core says there's
> something we don't understand yet ...

I'm not sure if the attached will help figure it out, but at the very
least it was eye-opening for me. I ran a test on
dblink_get_pkey('foobar') that returns 5 rows. I had a breakpoint set in
ExecClearTuple. I found that ExecClearTuple was called a total of 32
times for 5 returned rows!

Relevant to this discussion was that ExecClearTuple was called three
times, with the same slot pointer, for each function call to
dblink_get_pkey. Once in SRF_PERCALL_SETUP (per_MultiFuncCall), once in
TupleGetDatum (ExecStoreTuple), and once in FunctionNext in the loop
that builds the tuplestore.

Unfortunately I have not been able to get back to a point where I see a
coredump :(. But, that did seem to be related to calling the function
with an inappropriate declaration (now it just gives me garbage instead
of dumping core, even though I reverted the per_MultiFuncCall changes I
made earlier). I'll keep messing with this for a while, but I was hoping
the attached info would lead to some more suggestions of where to be
looking.

Thanks,

Joe