Hello all,
That was a fairly unusual case of a very old PR that had passed all of the checks at the time and managed not to have any conflicts with the current state of master, but didn’t actually fit the current state of the code. We could go back
to a merge strategy that requires manual rebases so that no code that isn’t actually up to date with (and hasn’t run CI with) the current state of master gets in; we could, GitHub actions permitting, invalidate CI runs of a certain age. Both of those would
certainly take care of this kind of situation though _always_ requiring a manual rebase on the very tip of master would like be a pain. If it’s possible, I’d lean towards some sort of staleness in CI, but fortunately, with our drive to clear out the
PR queue, we also don’t have that many old PR’s sitting around in this condition.
Best,
Colin
From: theia-dev <theia-dev-bounces@xxxxxxxxxxx> On Behalf Of
Thomas Mäder
Sent: Thursday, March 16, 2023 8:10 AM
To: theia-dev@xxxxxxxxxxx
Subject: [theia-dev] Master breakage
Hi folks,
we recently merged https://github.com/eclipse-theia/theia/pull/10338,
which seems to have broken master. Mistakes happen, but I'm wondering if we can improve our process? Did the linter check fail and we merged anyway, or is there some technical improvement we can make?
I'd also would encourage anyone who is involved in a "master is broken"-situation to write a quick mail to this very list to alert the community. It keeps everyone with failing
PR's from having to investigate what's wrong themselves.