| Submitter | Sean Farley |
|---|---|
| Date | Jan. 22, 2013, 7:05 p.m. |
| Message ID | <CANH+MCiem0OHCw44WT6k561Dj3CoWq=Ogwr5Q4xJ1mW=5b7wSQ@mail.gmail.com> |
| Download | mbox | patch |
| Permalink | /patch/713/ |
| State | Superseded |
| Headers | show |
Comments
On 22 janv. 2013, at 20:05, Sean Farley wrote: > On Tue, Jan 22, 2013 at 1:01 PM, Matt Mackall <mpm@selenic.com> wrote: >> On Tue, 2013-01-22 at 13:47 -0500, Augie Fackler wrote: >>> On Tue, Jan 22, 2013 at 1:45 PM, Sean Farley >>> <sean.michael.farley@gmail.com> wrote: >>>> On IRC, Kevin and Pierre-Yves were thinking that this is a bug in >>>> 'revrange' and that ":null" should return [-1] instead of [0,-1]. >>>> Thoughts? >>> >>> >>> Interesting. I'm somewhat inclined to agree, as nothing is an ancestor of null. >> >> Agreed. This result is nonsense: >> >> $ hg log -r :null >> changeset: 0:9117c6561b0b >> user: mpm@selenic.com >> date: Tue May 03 13:16:10 2005 -0800 >> summary: Add back links from file revisions to changeset revisions >> >> changeset: -1:000000000000 >> user: >> date: Thu Jan 01 00:00:00 1970 +0000 >> >> As is this one, which is wrong in a different way: >> >> $ hg log -r '(:null)' >> changeset: 0:9117c6561b0b >> user: mpm@selenic.com >> date: Tue May 03 13:16:10 2005 -0800 >> summary: Add back links from file revisions to changeset revisions > All those error probably comes from the revlog.rev trying to be smarter than xrange. xrange(start, end, step) - generate from start - in "step" order - until end is reach (not included) changelog.revs(start, stop) - generate from start - in the order that allow to reach end (reverse if stop start) - end is included As a result cl.revs is: - unable to generate empty list - can accidentally generate a non sensical reversed output instead of an empty list > I have a simple patch for this but think I need to split it up (both > to be applied on top of Bryan's patch). I wanted to be sure that the > change to using "repo['tip']" was correct for filtering and also to > check to see if there was a better way than "if end == -1", repo['tip'] is correct regarding filtering (But I sent a fix about it today) len(repo) - 1 is probably not wrong either as it will be filtered by changelog.rev The best option is probably to leave end to None and leave changelog.revs do it's jobs I do not see a better option than checking end==nullrev as changelog.rev is totally unaware that nullrev is valid iteration. As Matt pointed out, revset is also unaware of nullrev
Patch
diff --git a/mercurial/scmutil.py b/mercurial/scmutil.py --- a/mercurial/scmutil.py +++ b/mercurial/scmutil.py @@ -644,11 +644,13 @@ def revrange(repo, revs): continue if _revrangesep in spec: start, end = spec.split(_revrangesep, 1) start = revfix(repo, start, 0) - end = revfix(repo, end, len(repo) - 1) + end = revfix(repo, end, repo['tip']) + if end == -1: + start = -1 rangeiter = repo.changelog.revs(start, end) if not seen and not l: # by far the most common case: revs = ["-1:0"] l = list(rangeiter) # defer syncing seen until next iteration