A scenario was described during development of "resolve conflicts" PR #725:
the _local_update signal of the MagicFile state-machine goes from _conflicted -> _conflicted without doing anything
should we queue a LocalSnapshot, or is the user editing the file to ultimately resolve the conflict?
what if a remote party resolves the file meantime, and so we get an incoming update that is also a resolution?
(Could immediately re-conflict with the contents of the local edit -- that is, have a LocalSnapshot ready to go, apply the remote resolve but then produce a new conflict (hang on, we can't "produce a conflict" -- the other side has generated their new version) -- so yeah if we produce a LocalSnapshot at this moment, we could ensure its parent is our version before the incoming resolve (makes sense) and thus this is seen as a conflict by all parties)
That last thing might be the most-correct. That is:
continue to do nothing when we're conflicted and see a local change
_if_we get an incoming resolve while conflicted AND we have changed local contents, then:
queue a LocalSnapshot (with our current latest version as the parent, not the incoming resolve)
produce a new conflict locally (preserve "our" contents, write the resolution into conflict-marker
when our snapshot is uploaded, all other participants will see this as a conflict (I think?? right?)
...maybe the user completes editing the file and deletes the conflict-markers before any incoming resolution appears; this means we just create our own conflict-resolution Snapshot
A scenario was described during development of "resolve conflicts" PR #725:
_local_updatesignal of theMagicFilestate-machine goes from_conflicted -> _conflictedwithout doing anythingLocalSnapshotready to go, apply the remote resolve but then produce a new conflict (hang on, we can't "produce a conflict" -- the other side has generated their new version) -- so yeah if we produce a LocalSnapshot at this moment, we could ensure its parent is our version before the incoming resolve (makes sense) and thus this is seen as a conflict by all parties)That last thing might be the most-correct. That is:
LocalSnapshot(with our current latest version as the parent, not the incoming resolve)