[PATCH] mercurial: fixes update --clean to work on new branch (issue5003)

[Pierre-Yves David]

On 12/31/2015 11:48 AM, Sean Farley wrote:
>
> Augie Fackler <raf at durin42.com> writes:
>
>> On Wed, Dec 30, 2015 at 08:55:38PM +0100, liscju wrote:
>>> # HG changeset patch
>>> # User liscju <piotr.listkiewicz at gmail.com>
>>> # Date 1451505270 -3600
>>> #      Wed Dec 30 20:54:30 2015 +0100
>>> # Node ID a92bf43906f58c665c35bc745e6d83049a2e95c2
>>> # Parent  23541bdd1610c08af247f9c8719045cf247ce541
>>> mercurial: fixes update --clean to work on new branch (issue5003)
>>
>> I'm somewhat unconvinced. What's wrong with 'hg revert --all' in this case?
>>
>> (Other reviewers, please chime in - I'm not opposed to this, it just
>> seems a little quirky to me, and has implications for topics and some
>> other experimental work as well.)
>
> I'm inclined to think --clean should work on new branch since we have
> 'hg up -C' work on a bookmark without deactivating it.

Given that the "new" branch have not been commited yet, and --clean is 
about "discard uncommitted changes (no backup)", I think we should 
discard uncommited branch data in all case.

To rephrase, with --clean, the branch used for the update logic should 
-always- be the on of the first working copy parents. Uncommited branch 
data, (new or not), should be disregarded.

We could add a warning message for clarity.

Does that seems an appropriate and consistent behavior to you?

Cheers,

-- 
Pierre-Yves David