[PATCH] parsers: it should check the other interesting before decreasing ninteresting (issue3984)

[Chingis Dugarzhapov]

On Thu, Jul 25, 2013 at 5:01 PM, elson <elson.wei at gmail.com> wrote:

> Yes, in the original version, ancestors.py, the ancestors are sorted
> before to get the deepest one (line 77).
> But in parser.c, the ancestors are not sorted. That's why you get the
> different result.
> I was wondering which one is correct. Shouldn't we get the most recent,
> the largest number, revision? If the original is correct, I will send
> another patch to fix it.
>

According to old behavior, deepest function (ancestor.py:71) was returning
*both* 4 and 6, as they are both "deepest". And only manual check
(revlog.py:716) was filtering the winner (rev 4). It's not the case of your
patch ("C-style analogue of deepest" returns only 6).



> And by the way, does the "root" mean rev 0?
>

I suppose yes.



> In rev 18988, It will cause exception according to this bug. It's normal.
>


Yes, but then backward-compatibility check was dropped, for unknown (at
least for me) reasons.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://selenic.com/pipermail/mercurial-devel/attachments/20130725/5bd2629b/attachment.html>