RFC: dealing with dead, anonymous feature branches

[Augie Fackler]

On May 19, 2010, at 5:58 PM, Martin Geisler wrote:

> Augie Fackler <durin42 at gmail.com> writes:
> 
>> On Mon, May 17, 2010 at 10:13 AM, Martin Geisler <mg at aragost.com> wrote:
>>> Augie Fackler <durin42 at gmail.com> writes:
>>> 
>>> Now that I think of it, the scenario where T adds a cross branch tag
>>> for Y is already a problem today: if I only clone C, then my .hgtags
>>> file has a dangling reference to Y.
>>> 
>>> In other words: the consequences of dead branches are not new. All that
>>> changes is the heads that are pushed/pulled by default and you can
>>> already get yourself into the same problems today with 'hg clone -r'.
>> 
>> Sort of. You're proposing to not propagate dead branches *by default*,
>> which is a significant behavior change, especially in this tagging
>> case.
> 
> Yeah, but only if people create their tags while being on the "wrong"
> branch.
> 
>> Most users I know just do 'hg clone' without trying to be clever about
>> -r.
> 
> Sure, I also never use -r when cloning, or #branch for that matter.
> About #branch -- Henrik said on IRC that they use the #branch syntax a
> lot and that they basically never clone an entire repository. So he felt
> that pushing the abandoned branches to the server is okay, since people
> wont pull them again anyway.

Having talked to Martin about this some in IRC, I've come to understand that dead branches are unreleated to closed branches and are a way of solving the history mutation problem somewhat elegantly. +1 to the concept.

> 
> -- 
> Martin Geisler
> 
> See my Mercurial presentation: http://vimeo.com/11497288