On Wed, May 13, 2020 at 01:57:31PM +0200, Jiri Benc wrote:
On Wed, 13 May 2020 07:22:45 -0400, Prarit Bhargava wrote:
> IOW this is a chicken-and-egg problem and I'm not sure how get around it.
In other words, patches are sent to the list only after they pass CKI?
That's unexpected, though I can see that it could make sense.
However, in that case, we need something that ensures that patches that
have not been sent to the list are never applied. I can imagine that a
CKI outage might make all Gitlab merge requests tagged as failed CI,
which would be overriden by the package maintainer in good faith to
unblock the development. But that would mean that many reviewers did
not have a chance to see the patches. We need to ensure this can't
happen.
It's actually not CKI, just a few smoke tests (config validation, making
sure the SRPM builds) run on the free Gitlab runners. It'd be great to
get CKI to do build tests in the future, though.
Having the bridge wait for CI is configurable and I think it's best to
just turn it off. It was most useful back when I was the only one making
merge requests so I didn't spam the list with changes that were broken
anyway.
- Jeremy