But then we do use --hidden (at least on L30). Also there's no output for the command on L28 (which is where I suspect you wanted to avoid --hidden), should there be? (This may be because the PR is still in draft, let me know if I shouldn't be commenting on this yet)
While it seems to produce correct results according to the graph (and not crashing), looking at the contents, I'm pretty sure they're not actually correct. I'll need to spend more time on this to identify all the ways it's breaking (there's more than one) and possible fixes (or maybe I should create an issue detailing the problem(s)?)
If users are not proactive about resolving content divergence prior to issuing other commands, such as rebase, the repository can be put into a state that resembles:
o commit2 [content-divergent]
|
o commit1 [content-divergent]
|
o base
And running hg evolve --content-divergent
may attempt to relocate one of these commits to be a child of itself. If the relocation target would be ourself, just don't try to relocate.
Kyle Lippincott (212ca7a4) at 18 Mar 00:36
content-divergent: do not attempt to relocate node on top of itself
... and 1 more commit