On Fri, Feb 21, 2025 at 9:14 PM Andres Freund <andres@anarazel.de> wrote:
> I'd also say that breaking CI and BF is probably something I'd consider more
> urgent, as that could indicate the commit was just generally less well tested.
This is possible, but I think we oversell it by quite a lot. It's
quite possible for someone to put hundreds of hours into a patch and
forget to check CI before pushing; and it's also quite possible for
someone to do a terrible job vetting a patch where CI is clean. When
somebody is already stressed out about their patch breaking the
buildfarm, the very last thing they need is somebody who hasn't read
the patch or understood what the problems are to show up and say "hey,
maybe this patch should be reverted for all time and never considered
again ever!". It's just making a very stressful situation more
stressful. And it's cheap. If somebody shows up and says "hey, this
was improvidently committed for the following six design-level
reasons," that is abundantly fair and deserves major respect. Such
reviews take real time, thought, and work. Idly speculating that
someone's failure to check CI is a sign that they've also done
everything else wrong takes almost no work at all. I hate that we do
that to people. As much as I hate it when it happens to me, I think I
hate it even more when it happens to other people. It's a terrible way
to treat people who have poured their heart and soul into becoming
committers and who really care about the project, at least until we
beat the caring out of them.
But the real point of my previous email is that I just do not think
it's reasonable to expect people to fix complex programming problems
within hours. As much as I can be grumpy about CI, anything that goes
wrong with CI should in theory be something you can avoid ever having
to deal with on a Friday night no matter when you choose to commit,
because you can test things in advance. I know the BF has more
configurations than CI, but instead of having LESS capability to test
things in advance, it has NONE. Twenty years ago, post-commit testing
was probably the best you could hope for, but today it isn't.
--
Robert Haas
EDB: http://www.enterprisedb.com
