On Thu, Feb 28, 2019 at 3:09 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> The only thing that's really clear is that some senior committers don't
> want to be bothered because they don't think there's a problem here that
> justifies any additional expenditure of their time. Perhaps they are
> right, because I'd expected some comments from non-committer developers
> confirming that they see a problem, and the silence is deafening.
I don't think that you can take that as too strong a signal. The
incentives are different for non-committers.
> I'm inclined to commit some form of Naylor's tool improvement anyway,
> because I have use for it; I remember times when I've renumbered OIDs
> manually in patches, and it wasn't much fun. But I can't force a
> process change if there's not consensus for it among the committers.
I think that that's a reasonable thing to do, provided there is
obvious feedback that makes it highly unlikely that the committer will
make an error at the last moment. I have a hard time coming up with a
suggestion that won't be considered annoying by at least one person,
though.
Would it be awful if there was a #warning directive that kicked in
when the temporary OID range is in use? It should be possible to do
that without breaking -Werror builds, which I believe Robert uses (I
am reminded of the Flex bug that we used to have to work around). It's
not like there are that many patches that need to assign OIDs to new
catalog entries. I would suggest that we put the warning in the
regression tests if I didn't know that that could be missed by the use
of parallel variants, where the output flies by. There is no precedent
for using #warning for something like that, but offhand it seems like
the only thing that would work consistently.
I don't really mind having to do slightly more work when the issue
crops up, especially if that means less work for everyone involved in
aggregate, which is the cost that I'm concerned about the most.
However, an undocumented or under-documented process that requires a
fixed amount of extra mental effort when committing *anything* is
another matter.
--
Peter Geoghegan
