[loom:MERGE] fixing "main" part of setup.py

[Alexander Belchenko]

Aaron Bentley пишет:
> Alexander Belchenko wrote:
>> Consider this trivial patch.
> 
>> I have a question: what's plans to teach `bzr send` understand looms?
> 
> send does not fundamentally need to support looms, because it uses
> revision-ids to specify the merge.
> 
> To send just the current thread, you can do "bzr send -r thread:.."

Hm, I did not know about -rthread:

>> And I had troubles to figure out how to `bzr send` it. Currently I did:
> 
>> bzr send -r-2..-1 . . 
> 
> You should have done:
> bzr send -r-2..-1
> http://bazaar.launchpad.net/%7Ebzr-loom-devs/bzr-loom/trunk

Sometime I have a limited internet access, so I don't want to go every time to inet, when I need to 
bzr send.

> 
> Or if you had a local mirror of trunk, you could have used that as the
> submit_branch.

Yes, my bottom thread is exact local mirror. That's why I ask my question about support looms in 
send command. I don't want and I think I don't need to have separate mirror just for make send happy.

> 
> You should never specify the same directory for the source branch and
> the submit_branch.  The submit_branch is used to determine which
> revisions need to be sent.  Since you specified a submit_branch that
> already has all the necessary revisions, no revisions were sent.

I'm still think it's a wrong.

> 
> You should never specify . as a public branch.  This is an optional
> parameter.  There's no need to specify it at all.  But if you do specify
> it, you must specify a real public location, not a local path.
> 
>> And here is the patch. Is it mergeable patch actually?
> 
> No.  The bundle doesn't contain the necessary revisions, because you
> specified the wrong submit_branch.

Aaron, with all my respects to you and other core dev guys, I think current bundle/send behavior is 
the pain in the ass. Old dumb way to bundle revisions that user requested is better and 
understandable. At least for me personally. I'd like to have a knob to force old behavior. 
Unfortunately I see that using format 0.9 does not help here.