warning about divergent renames

[Brodie Rao]

On Oct 11, 2010, at 8:23 AM, Faheem Mitha <faheem at email.unc.edu> wrote:

> 
> 
> On Mon, 11 Oct 2010, Brodie Rao wrote:
> 
>> On Oct 11, 2010, at 4:27 AM, Dirkjan Ochtman wrote:
>> 
>>> On Fri, Sep 10, 2010 at 21:29, Dirkjan Ochtman <dirkjan at ochtman.nl> wrote:
>>>> On Fri, Sep 10, 2010 at 20:34, Matt Mackall <mpm at selenic.com> wrote:
>>>>> This is another instance of an issue that needs full historical context
>>>>> to progress.
>>>>> There originally wasn't a warning, then I added one after various people
>>>>> complained. Notably there's a warning in the book about Mercurial not
>>>>> doing -more- in this case.
>>>> I suspected something like that. Do you know/remember what the edge
>>>> cases are why people would want to be aware of this? I think it has to
>>>> do with merging, right? When you merge a file that has modifications
>>>> on one side and a divergent rename on the other?
>>> Matt, can you respond to this? I saw Dan sent in some patches relating
>>> to this. For those of us trying to follow along at home, was this
>>> discussed at the sprint? What was the outcome?
>> 
>> So, there are a couple of cases where the warning happens when doing hg update:
>> 
>> 1. A changeset either copies or renames the same file to different destinations.
>> 
>> 2. One or more changesets *just* copy the same file to different destinations.
>> 
>> copies.copies() wasn't distinguishing between copies and renames when recording divergent renames, so you'd get warnings for all of those cases.
>> 
>> Dan's patch makes it only display the warning for renames. That gets rid of #2 completely but still warns you on actual divergent renames.
> 
> I'm confused about the wording of 1 and 2. Doesn't 1 include 2 as well? It would make sense to be if 1 was just renames.

Sorry, that's what I meant.

And for some context, this came up at the sprint when Augie was hacking on online help. The Mercurial repo now has a changeset that triggers the warning.