This patch affects creation of new shared repositories, right? At the moment we don’t have any Rust code doing that.

This version of the patch looks good, thanks!

Looks good, but there seems to be yet another edge case within the edge case… *sigh*. See inline comment.

Looks good, thanks!

Apparently Phabricator doesn’t show commit author separately from submitter, but this was written a few months ago by Georges Racinet. I’ve reviewed it, rebased it, adapted it to compile again, and added the config to disable it by default. Some tests are failing when running this for hg status (hence disabling) but this is still a good start.

I haven’t measured memory at all but I’d expect the effect to be very small or not noticeable since this function is typically called for printing the return value to stdout then deallocating it immediately. We’re not accumulating many "relativized" paths in memory at any given time.

Yes, absolutely. The new configuration keys and their values too. However this is split between several patches that I feel are not quite settled yet, so it’ll be easier to start writing docs after this patch series has been merged.

Marmoute makes a good point about the default to looking up hg in $PATH being fragile, making it easy to accidentally use not the expected installation of Mercurial.