[Tracker-discuss] [issue415] Mercurial integration: regression in diff generation

[Martin v. Löwis]

Martin v. Löwis <martin at v.loewis.de> added the comment:

> “utterly broken” is a very strong opinion.  The git diff format
> supports things that unified diff does not and has one drawback to my
> knowledge (I forgot exactly what—it was one missing piece of
> information that you mentioned during one of the python-dev threads
> about Mercurial migration or Roundup-Mercurial integration).
> Everyone recommends always using git diffs.

That's because nobody (except for the Rietveld integration) needs to
apply patches automatically, apparently - else people would have
noticed that the format is just unusable, for this kind of application.

Notice that this holds for Mercurial's implementation of git-style
diffs. The ones generated by git don't have this issue.

> 
>> As for changes that are reverted: what changes specifically? (or:
>> what versions should it have compared instead of comparing
>> 0c1c9bb590a9 and bc362109eed8?)
> I’m sorry I don’t know exactly.  I have this comment from the bug
> report:
> 
>> I'm not sure how the bugtracker patch mechanism works, but the
>> patch it produced included a lot of changes > that I didn't make
>> (changesets already committed to the master repo).

I think that claim is incorrect. Unless it can be demonstrated with
specific chunks that are there but ought not, I propose to close
this issue as invalid.

_______________________________________________________
PSF Meta Tracker <metatracker at psf.upfronthosting.co.za>
<http://psf.upfronthosting.co.za/roundup/meta/issue415>
_______________________________________________________