Richard Lowe wrote: > Stephen Lau wrote: >> http://cr.opensolaris.org/~stevel/hg_branch/ > > As a random aside, I keep entirely forgetting that that exists. > >> Anyone want to take a look at the above and see if it makes sense? >> >> This uses hg-active (and by extension Cadmium) to determine the >> parenttip to determine what to diff against. The previous behaviour >> was depending on 'hg log <path>' to follow the branch, but it doesn't >> (and adding -f doesn't help since apparently for files, that only >> follows copies/renames - not branches). >> >> In any case, I haven't filed a bug to track it yet - but if the >> changes look reasonable I will. > > webrev.sh: > What happens if we get a file list without HG_PARENT? This needs to > continue to work with explicit file lists of both varieties (though it > may > not work as well...)
Right - I believe the old behaviour isn't quite right (it mostly works, but certainly fails in this case of a repository with branches - so clearly it needs some more work). I tossed it out with the intent of working on something better, but wanted to get this code reviewed first. So currently this won't work if we get a file list without HG_PARENT. What cases are you expecting that we can't get HG_PARENT? I suppose the first changeset is one; are there other cases? If we can't contact the parent repo (disconnected operation), we should still have one. (Sorry it's late, I might be missing something obvious) > webrev.sh:1464 > You set hg_parent (note the case), and never use it. g > Given that, this whole bit doesn't seem to have a purpose (does it?) No, that was more a copy/paste bit before I changed the later code - it doesn't need to exist since we eval it from the $FLIST. Thanks. > webrev.sh:1585 > hg cat exits with 0 even when it fails, this won't work. (but this > annoys me, so fixing hg would also be ok by me...) Ugh. That's sucky - yeah, I'd rather fix hg so it returns 0 upon failure of 'hg cat'. > webrev.sh:1943 > tip is almost always the wrong thing to use in a world with in-repo > branches, consider 'hg parent' Good catch. Thanks. cheers, steve -- stephen lau // stevel at sun.com | 650.786.0845 | http://whacked.net opensolaris // solaris kernel development
