[PATCH 3 of 3 STABLE] revert: look for copies information for all local modifications

[Pierre-Yves David]

On 11/26/2014 04:50 PM, Martin von Zweigbergk wrote:
>
>
> On Wed Nov 26 2014 at 4:44:00 PM Pierre-Yves David
> <pierre-yves.david at ens-lyon.org <mailto:pierre-yves.david at ens-lyon.org>>
> wrote:
>
>
>
>     On 11/26/2014 04:38 PM, Martin von Zweigbergk wrote:
>      >
>      > On Wed Nov 26 2014 at 4:33:16 PM Pierre-Yves David
>      > <pierre-yves.david at ens-lyon.__org
>     <mailto:pierre-yves.david at ens-lyon.org>
>     <mailto:pierre-yves.david at ens-__lyon.org
>     <mailto:pierre-yves.david at ens-lyon.org>>>
>      > wrote:
>      >
>      >
>      >
>      >     On 11/26/2014 04:29 PM, Martin von Zweigbergk wrote:
>      >      >
>      >      > On Wed Nov 26 2014 at 3:31:37 PM Pierre-Yves David
>      >      > <pierre-yves.david at ens-lyon.____org
>      >     <mailto:pierre-yves.david at ens-__lyon.org
>     <mailto:pierre-yves.david at ens-lyon.org>>
>      >     <mailto:pierre-yves.david at ens-____lyon.org
>     <mailto:pierre-yves.david at ens-__lyon.org>
>      >     <mailto:pierre-yves.david at ens-__lyon.org
>     <mailto:pierre-yves.david at ens-lyon.org>>>>
>      >      > wrote:
>      >      >
>      >      >     +Also true for move overwriting existing file
>      >      >     +
>      >      >     +  $ hg mv --force a b/b
>      >      >     +  $ hg revert b/b
>      >      >     +  $ hg status a b/b
>      >      >     +
>      >      >
>      >      >
>      >      > Could you remind me what the output would have been before
>     this
>      >     patch?
>      >
>      >
>      >     B would not have been removed.
>      >
>      >
>      > It is still not removed... which makes sense, but it's no difference.
>
>     What I wanted to write is:
>
>     Before this patch, a (the source of the rename) is left removed. After
>     this patch it is properly restored when 'b/b' (the destination of the
>     rename) is reverted.
>
>
> Interesting. I would not expect a to be restored since the user
> explicitly asked to restore b/b.

This is the behavior we have for move.

  If you move a → b. And call 'hg revert b', it move 'a' back in place

(I repeat this is the current behavior)

This patch just make it so it happen also in the case 'b' previously 
existed. This sound like a consistency improvement.

> Did you? The patch is not tagged with issue4458, so I assume it's not needed for mq.

MQ have nothing to do with issue4458. It is just that the bug was found 
while playing with MQ.


-- 
Pierre-Yves David