[Bug 309309] Re: selftest can't be run in a bound branch

[Vincent Ladeuil]

>>>>> "Marius" == Marius Kruger <amanic at gmail.com> writes:

    Marius> 2008/12/18 Vincent Ladeuil <v.ladeuil+lp at free.fr <v.ladeuil%2Blp at free.fr>>
    >> Replying to the list so that this get discussed there.
    >> 
    >> >>>>> "Marius" == Marius Kruger <amanic at gmail.com> writes:
    >> 
    Marius> * revert the nick behaviour to not read the nick from
    Marius> the master branch (for bound branches)
    >> 
    >> I didn't follow the thread where this change was discussed, can
    >> you summarize the rationale ?

    Marius> for or against?

Both :-) I'm personally quite neutral about this as long as the
tests pass :)

    Marius> Originally when I wanted to update the local nick
    Marius> when switching, people (I think Robert/John) was of
    Marius> the opinion that in stead of writing the nick, we
    Marius> should rather let checkouts refer to the master
    Marius> branch for its nick.  It might be possible in future
    Marius> to _only_ do this if the master branch is also
    Marius> local?!

I don't think that's the right approach as :

- you may not be able to easily test that the master branch is
  local,

- if the nick is defined as being the master's one, so be it, but
  'commit --local' tends to lose its interest if you can't use a
  bound branch without being connected anyway.

I mean, I routinely use 'bzr info' just to *check* if I'm using a
bound branch or a local one, I'm already annoyed that a unbounded
(as in previously bound) branch doesn't give a firm indication
that it is currently not bound (I use the push branch as an hint,
but that's not the same (or is it ?)).

    Marius> with your testing issues, John mentioned that he
    Marius> might reverse his opinion about this.

Let's wait for John and Robert comments about it then,

      Vincent