On Mon, Apr 29, 2019 at 01:35:12PM +0000, Daniel Gustafsson wrote:
> Good point, I've updated the patch to include those as well.
I have been reviewing this patch, and the change in pg_waldump is
actually a good thing, as we could finish with a crash if strdup()
returns NULL as the pointer gets directly used, and there would be an
assertion failure in open_file_in_directory().
parseAclItem() in dumputils.c gets changed so as we would not return
false on OOM anymore. I think that the current code is a bug as
parseAclItem() should return false to the caller only on a parsing
error so the caller can get confused between the OOM on strdup() and a
parsing problem.
In short, as presented, the patch looks acceptable to me. Are there
any objections to apply it on HEAD?
--
Michael