Possible bug: abandoning conflicted changes no longer resolves bookmark conflicts

[BatmanAoD]

At least as of v0.32 (but possibly also before), abandoning all but one change-ID from a conflicted commit does not automatically resolve the conflict for any bookmarks attached to that commit. jj bookmark list shows:

branch-name (conflicted):
  - <sha> hidden ...
  - <sha> ...

It seems weird that a "hidden" change would cause a bookmark conflict in the first place, but it seems even weirder that the abandon operation that actually hides the operation leaves the bookmark in a conflicted state.

This behavior would at least seem consistent if abandon didn't operate on bookmarks at all, but it seems inconsistent with the bookmark-deletion behavior¹.

Footnotes

1. I don't want to argue about this behavior here, but just so this discussion isn't taken as tacit approval of the bookmark-deletion behavior, I want to say explicitly that I'm still not a fan. ↩

[yuja]

It's technically correct that jj abandon doesn't (always) resolve conflicted bookmarks because there are still conflicts between old_target -> new_target1 and old_target -> (absent).

initial state:
  old_target
on conflict:
  old_target -> [new_target1, new_target2]
on `jj abandon new_target2`:
  { old_target -> [new_target1, new_target2] } + { new_target2 -> (absent) }
  = old_target -> [new_target1, (absent)]

I don't have a good model to make jj abandon automatically resolve conflicts.

[martinvonz]

The behavior changed in 0.26 when we made jj abandon also delete any bookmarks pointing to the target (https://github.com/jj-vcs/jj/releases/tag/v0.26.0). Before that release, it was probably at least more common that jj abandon would resolve a conflict. It would have happened if new_target2 was a descendant of old_target in Yuya's example.