Attached is a patch that fixes three Python reference leaks in
PLy_traceback(): the objects returned by PyErr_Fetch() are owned by the
caller, so their reference count should be decremented. The memory leak
can be reproduced as follows:

create function import_fail() returns text as
'import foosocket
return "succeeded, that wasn''t supposed to happen"'
LANGUAGE plpythonu;

create function test_py() returns void as $$
begin
for i in 1 .. 1000 loop
perform import_fail();
end loop;
return;
end;$$ language plpgsql;

select test_py();

On my system, each invocation of test_py() leaks about 2MB. With the
patch applied, each invocation leaks about 500KB. So obviously there are
some more leaks here -- I searched briefly for additional problems, but
couldn't see anything obvious.

I don't have time at the moment to track down the remaining problems, so
I'd like to apply the patch as-is: I'll come back to the remaining
memory leaks later, unless someone beats me to it.

Barring any objections, I'll apply this patch to HEAD and back branches
tomorrow.

-Neil



---------------------------(end of broadcast)---------------------------
TIP 1: if posting/reading through Usenet, please send an appropriate
subscribe-nomail command to majordomo@postgresql.org so that your
message can get through to the mailing list cleanly

On Sun, 2006-02-19 at 20:34 -0500, Neil Conway wrote:
> Attached is a patch that fixes three Python reference leaks in
> PLy_traceback(): the objects returned by PyErr_Fetch() are owned by the
> caller, so their reference count should be decremented.

Applied to HEAD and back branches. I also noticed a minor bug in
PLy_modify_tuple(): we don't own a reference to `platt', so we shouldn't
try to decrement its refcount. I only bothered fixing the latter bug in
8.0+

-Neil



---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match