Re: Please check Salsa CI before your next upload (Re: debian/watch version 5)
Hi,
For the record, devscripts is green again thanks to Yadd's fix about
1h after I my email noting it:
https://salsa.debian.org/debian/devscripts/-/commits/main
> I did in fact look at CI (and also ran autopkgtest and saw a failure).
>
> But I judged that I should upload anyway, given the inconsistent state
> sid was in wrt the transition I was working on.
I think it would have been better if you had held off the upload. It
wasn't such an urgent "production issue" that you could not have had a
pause for some hours or half a day.
Instead of just uploading, you could have raised the issue of
autopkgtest being broken and waited to see whether Yadd has time to
fix it immediately or not. As we can see in hindsight, he fixed it in
an hour from my email. Intentionally uploading broken stuff would only
be motivated if there was some effort to fix it first and the
conclusion that "we can't wait more" came after actually spending at
least a half day waiting. I earlier also sent you an email with the
list of commits missing from the changelog, which you then chose not
to incorporate into your final changelog update, and I sent a private
email to you asking to hold off the upload until devscripts is fixed
but seems that arrived 15 minutes late. If you had waited for one
extra hour, both the autopgktest and changelog would likely have been
fixed and the upload correct.
Also, we could have avoided this in the first place if more people in
the past 3 days had sent their submissions to devscripts as Merge
Requests instead of pushing directly to main.
Many people here seem to rehash the argument that they should be
allowed to do whatever they want and trust their own judgment. My
argument is that people don't make mistakes intentionally, and using
"judgment" to have things reviewed and tested *only* when the
submitter suspects there could be a mistake is going to miss most of
the cases when people do accidental mistakes. Another argument that
gets repeated here is that people don't want to spend time on waiting
for tests or reviews because of the increased cost. Yes - there is
some extra effort required upfront, but there are also savings from
avoiding mistakes that affect the whole archive. Waiting for CI to
pass or for another person to check out your change and approve it
will slow you down a bit, but if you don't do it, you may risk slowing
down tens of other people who tried to do something else and then got
affected by the mistake. Your workflow decision is not fully your own
personal decision, but it has consequences for other people too, in
particular when working on a package like devscripts.
Reply to: