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
I know this was mentioned somewhere on Zulip, but figured it's worth an issue to discuss.
It seems like using an octopus merge when handling >1 PR may be better than the current strategy of doing individual rollup commits. Some reasons:
It enforces the desired strategy for rollup merges, i.e. it will not allow a "complex merge that needs manual resolution". We already have to be sure that rollups contain only orthogonal branches, I have a mild suspicion that letting git handle the resolution may make this easier to implement.
Easier to understand history. It is currently kind of tricky to navigate around the git log and see (1) what was merged & tested together, and (2) what the source of the rollup is, since the rollup PR is not linked. Using a single merge commit to represent the rollup PR/test makes this easier to understand
More concise git history. This is obviously a low priority thing, but not having a separate merge commit for every PR helps keep history smaller and much less noisy
The text was updated successfully, but these errors were encountered:
If we want to keep the PR messages, we could put them all together into the merge commit.
That being said - I don't think there is much value to making the PR body exist in the git log. Linking to the PR in the merge commit is good, but the PRs themselves are often pretty bulky and contain information that isn't relevant. The commit messages really should be able to stand their own - but commit message quality is of course a separate conversation that does need to be had.
Related, there doesn't seem to be much value to including the branch name in the merge commit. A lot of times it's just noise like patch-1 that adds nothing of value, especially since the PR link handles this.
I know this was mentioned somewhere on Zulip, but figured it's worth an issue to discuss.
It seems like using an octopus merge when handling >1 PR may be better than the current strategy of doing individual rollup commits. Some reasons:
The text was updated successfully, but these errors were encountered: