This dicussion reminds me of a possible memory leak in plpgsql's code. In case you are interested in it;
in pl_comp.c, plpgsql_build_variable takes a pointer to a PLpgSQL_type structure, which is always a malloc'ed instance(since we always use plpgsql_build_datatype function). The switch statement in plpgsql_build_variable function elicits that its reference is only kept in case the type structure represents a PLPGSQL_TTYPE_SCALAR, otherwise it is not kept and needed in case its either PLPGSQL_TTYPE_ROW or PLPGSQL_TTYPE_REC.
So is it intensional or a memory leak?
Thank you
On 9/27/05, Tom Lane <tgl@sss.pgh.pa.us> wrote: "Qingqing Zhou" <zhouqq@cs.toronto.edu> writes:
> "Tom Lane" <tgl@sss.pgh.pa.us> wrote
>> No, because you're thinking in terms of the backend environment, and
>> generally in the backend the answer to "when to use malloc directly"
>> is "never".
> Well, except before MemoryContext mechanism is set up? For example, the
> functions( e.g., GUC, vfd) used during bootstrap.
I think you need to take another look at the startup sequences. Those
modules are not run before MemoryContextInit. In any case, the odds
of running out of memory before we get to MemoryContextInit are so small
that I don't have a problem with crashing if it happens.
regards, tom lane
---------------------------(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