[MERGE][#299313] --local commits are accessing the network

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Vincent Ladeuil wrote:
> I didn't follow closely how we end up changing the default
> behavior of branch.nick *but* please take into account that it
> 'breaks' the test suite when run in a bound branch.
> 
> One of the reasons is that some tests use 'bzr --version' as a
> 'do very little' command and, when the bound branch requires some
> password (like any ssh one does)... selftest prompts and hangs..
> 
> I use an ssh agent but the test suite (rightly) reset the
> SSH_AUTH_SOCK.
> 
> bzr selftest -s bb.test_selftest.TestRunBzrSubprocess.test_run_bzr_subprocess
> 
> is one such test.
> 
>       Vincent
> 

That certainly sounds like a test case that needs a bit more isolation
from the outside world.

I also ran into problems with this in my "show the nick of the current
branch" script I run to generate my command prompt. Suddenly just doing
"cd bzr.dev" would take 5s+ to show the prompt because it was connecting
and probing for the nickname of http://bazaar-vcs.org/bzr/bzr.dev

I hacked around it in my script, to detect if there was an explicit
local nick and if not, just use "<notlocal>".

I suppose when Branch.get_nick(local=False) becomes officially
available, I can switch to that.

In the end, I probably have to reverse my opinion that using the nick of
the master is better than using the local nick. There is a lot more
overhead to grab the remote value, and 'nick' seems like something that
should be trivially fast to compute.

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkk1SrEACgkQJdeBCYSNAAPX4QCeKWlR3NtaV11JSTuJ8+Ys0kwB
Wu0An2IqYsKMLGTOGkjBqW18w4zK8r7H
=2XfT
-----END PGP SIGNATURE-----