On Mon, Oct 29, 2018 at 07:11:35AM +0100, Fabien COELHO wrote:
> Michaël suggests that there is no issue of external tool using the internal
> function, so I'm fine with this version.
>
> I have switched the patch to ready for committer.
One catch with this refactoring is that for example this combination
does not result in an error on HEAD, but it does with the patch:
pg_restore -l -C -1
Anyway, the fact that we save the caller from one exit_horribly()
knowing that opening the archive is completely useless makes the move
worth it in my opinion. RestoreArchive should complain about things
which depend on the opened archive, which is the only thing it does
now. I would not risk back-patching it though.
At the same time, I have checked the set of TAP tests for pg_restore and
the cross-option checks are all covered, so no need to go crazy on this
side.
For the archive's sake, the original commit 3a819b07 which did the
option check in RestoreArchive comes from here:
https://www.postgresql.org/message-id/496B6B40.1010909@hagander.net
And committed, thanks Daniel and Fabien!
--
Michael