You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
After an extensive discussion on Discord about "What prePR means to you", I think we roughly came to this consensus:
prePR should only do polish tasks, which are essentially things that can be fixed by the task itself. Additionally, the polish check should be a separate job in CI, like suggested in Project verification job #93.
We should try and relax the CI-requires-approval-for-new-contributors, since it really kills the feedback loop. I'm pretty sure there's a mode to only require it for new GH users, and I bet an admin can set this for the whole org.
Roughly, prePR/polish tasks should meet this criteria:
it happens without any configuration or control
it's infallible when it runs
once you run it locally, the check in ci will never fail
things that check something and fix it if needed on their own
As much as CI failures are annoying, at least maintainers can help. We don't want prePR to discourage contributors from creating a PR in the first place by failing with e.g. mima issues that they don't understand or know how to fix.
There's also a lot of interest in an idea to ask CI to run prePR for you like in #69, but it's possibly tricky to implement.
The text was updated successfully, but these errors were encountered:
Note that some scalafix are "polish" (e.g., organize imports), and some require intervention (e.g., no fs2.Compiler[Sync]). And some of those that are "polish" are semantic rules, and therefore relatively slower.
After an extensive discussion on Discord about "What
prePR
means to you", I think we roughly came to this consensus:prePR
should only do polish tasks, which are essentially things that can be fixed by the task itself. Additionally, thepolish
check should be a separate job in CI, like suggested in Project verification job #93.Roughly,
prePR
/polish tasks should meet this criteria:As much as CI failures are annoying, at least maintainers can help. We don't want
prePR
to discourage contributors from creating a PR in the first place by failing with e.g. mima issues that they don't understand or know how to fix.There's also a lot of interest in an idea to ask CI to run
prePR
for you like in #69, but it's possibly tricky to implement.The text was updated successfully, but these errors were encountered: