Propagates bookmarks termination

[Sean Farley]

On Fri, Sep 21, 2012 at 12:44 AM, Martin Geisler <martin at geisler.net> wrote:
> Pierre-Yves David <pierre-yves.david at logilab.fr> writes:
>
>> I'm using bookmarks for my Mercurial developpement for a few months.
>> My workflow is as follow:
>>
>>     At work
>>
>>      $ hg bookmarks somefeature
>>      hack hack hack
>>      $ hg push -B somefeature http://hg-lab.logilab.org/wip/hg/
>>
>>     at home
>>
>>      $ hg pull http://hg-lab.logilab.org/wip/hg/
>>      hack hack hack
>>      $ hg amend
>>      $ hg push http://hg-lab.logilab.org/wip/hg/
>>
>>     eventually
>>
>>      $ hg patchbomb -r '::somefeature and draft()'
>>
>> Since 2.3 hg pull synchronize all bookmarks. this part works well! But
>> when somefeature is queued into core troubles start
>>
>>   At home
>>
>>     $ hg pull main
>>     $ hg bookmark -d somefeature
>>     $ hg push -B somefeature http://hg-lab.logilab.org/wip/hg/ # to propagate the kill to hg-lab
>>
>> and from work I need
>>
>>     $ hg pull http://hg-lab.logilab.org/wip/hg/ -B somefeature
>>
>> Otherwise an hold version of the bookmarks remains dandling somewhere.
>>
>> It would be nice if the pull could propagate the deletion of the
>> bookmark the same is does with addition.
>
> Isn't this trivially impossible? Since bookmarks have no history, you
> don't know if remote is missing the bookmark because it was deleted
> there, or if it's missing the bookmark because it never went there in
> the first place. The missing history is exactly what you're adding with
> the obsolete work, isn't it? :-)

Impossible? As a hack, why couldn't something like this be done:

for b in local_bookmarks:
  append b to pull_list

<tests this out>

... and now I see. Le sigh. Could we put an obsolete marker on the
bookmark data?