It is actually worse than described right now. Even the combining of the PRs wouldn’t pass because they can’t affect the images themselves, which need patching to pass. The images are tainted by a previous merge too. So pinning back to an older image allows one of the fixes to pass but the other to fail. If unpin the image then we get the a different failure.
We simply MUST be able to override these blocks when necessary to get things back into a working state. -Jake > On Aug 17, 2021, at 2:26 PM, Owen Nichols <onich...@vmware.com> wrote: > > For this situation in Geode repo, in addition to Squash, we also allow > Rebase. This would allow you to put both commits in the same PR to pass > checks, but still apply them to develop as separate commits. > > On 8/17/21, 2:20 PM, "Blake Bender" <bbl...@vmware.com> wrote: > > Hello everyone, > > Today I once again find myself between a rock and a hard place managing > incoming PRs into the geode-native project. We merged a PR that passed > checkin gates, then broke in the main CI pipeline. Additionally, our code > formatter took an update yesterday, and now disagrees with some of the code > that is already on the develop branch. I submitted a PR to fix the > formatting, but it now won't pass checkin gates because of the first break, > and said first break can't be reverted because it won't pass checkin gates > due to formatting. I could maybe solve this problem by combining the two > PRs, but then I'd pollute my Git history, which IMO is a bigger problem than > either of these issues. > > Sadly, this happens much more often than you'd think, and every time it > does it takes days to untangle this knot we've tied ourselves in, when it > should take seconds or minute of my time at most. I would like to propose > that we either unprotect develop on geode-native, and allow direct checkins > for specific circumstances like reverting PRs or fixing things like this > formatting issue. It's crazy to keep wasting my time trying to work around > something with such a simple solution at hand. > > Thanks, > > Blake > >
