We've reached the state where our case-folding logic is robust enough for
purge's needs and I've already queued the appropriate changes to purge.

Did I see on the list that there was some work being done on case-folding
systems? Do you think that will fix this issue?

It's the 'usually fine' wherein the problem lies. Not 'usually' deleting local
changes is not good enough here (it made people sad, bugs were filed). We really
need a solution to the core casefolding problem before purge is safe.

My point is that case changes are rare and status is usually fine, whereas 'hg
clean' does not work _in the normal case_ on such systems.

On the contrary, hg status is completely clueless about case changes.

Adjusting subject and topics.

Mac users report that if they have modified files in a repository (but all
non-ignorable files tracked) and run 'hg clean --all', Hg reports

Purging on name mangling filesystems is not fully supported.
abort: outstanding uncommitted changes

This effectively makes the 'clean' command useless in a cross-platform
development team.

'hg status' knows how to deal with such filesystems, so I think 'hg clean'
should too.