On Fri, Nov 02, 2018 at 04:39:07PM +0900, Amit Langote wrote:
> Agreed that they're two independent issues, although it wouldn't be such a
> bad idea to fix them in one go, as they're both issues related to the
> handling of ON COMMIT actions on tables in inheritance trees.
I have pushed 0001 which fixes the bug reported on this thread down to
v10, after tweaking a bit the patch after more review. I was testing
heap_truncate_one_rel() a bit more deeply and it actually happens that
it can work with matviews. Re-reading the thread and sleeping on it,
Tom was been actually suggesting to move the check one level down to
heap_truncate_one_rel(), which actually makes more sense. So I have
changed the patch so as a check on RELKIND_PARTITIONED_TABLE is done
instead of RELKIND_RELATION which is what has been proposed until now.
Regarding the second patch, could you start a second thread? The scope
is not only related to partitioned tables but also to inheritance trees
so this goes way back in time, and I think that we could attract a
better audience about the problem. I don't mind starting a thread
myself, not without your authorization as you wrote a patch to deal with
the problem. My apologies, I have not looked at what you are proposing
yet.
--
Michael