Is it memory leak or not?

Dmitry Igrishin <dmitigr@gmail.com> writes:

Maybe I'm wrong, but anychar_typmodin() of
src/backend/utils/adt/varchar.c of PostgreSQL 12.1 does not pfree()s
the memory allocated by ArrayGetIntegerTypmods(). Probably, I'm
missing something. Could anybody please clarify on that?

It is a leak, in the sense that the pointer is unreferenced once the
function returns. But we don't care, either here or in the probably
thousands of other similar cases, because we don't expect this function
to be run in a long-lived memory context. The general philosophy in
the backend is that it's cheaper and far less error-prone to rely on
memory context cleanup to reclaim (small amounts of) memory than to
rely on manual pfree calls. You can read more about that in
src/backend/utils/mmgr/README.

regards, tom lane

On Mon, 10 Feb 2020, 18:59 Tom Lane, <tgl@sss.pgh.pa.us> wrote:

Dmitry Igrishin <dmitigr@gmail.com> writes:

Maybe I'm wrong, but anychar_typmodin() of
src/backend/utils/adt/varchar.c of PostgreSQL 12.1 does not pfree()s
the memory allocated by ArrayGetIntegerTypmods(). Probably, I'm
missing something. Could anybody please clarify on that?

It is a leak, in the sense that the pointer is unreferenced once the
function returns. But we don't care, either here or in the probably
thousands of other similar cases, because we don't expect this function
to be run in a long-lived memory context. The general philosophy in
the backend is that it's cheaper and far less error-prone to rely on
memory context cleanup to reclaim (small amounts of) memory than to
rely on manual pfree calls. You can read more about that in
src/backend/utils/mmgr/README.

I see. Thank you very much!