[pqm-submit:MERGE] Remove default --message

[John Arbash Meinel]

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

James Henstridge wrote:
> On 16/03/2008, John Arbash Meinel <john at arbash-meinel.com> wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>>  Hash: SHA1
>>
>>  The existing pqm-submit plugin grabs the commit message from the last
>>  commit on the branch. However, this is *rarely* appropriate for a merge
>>  commit.
> 
> That fits in with my experience with pqm-submit, so no complaints from
> me.  The last few commits on most of my branch are usually cleanups
> related to reviewer comments and don't have log messages that
> represent the primary purpose of the branch.
> 
> Furthermore, Launchpad's PQM requires messages in a form that doesn't
> match most of my commit messages :)
> 
> 
>>  It also adds another fix, which only allows a non-exact path when it is
>>  using the '.' path.
>>
>>  The impetus for this was because Andrew just successfully did:
>>
>>  bzr pqm-submit "This is a nice long commit message"
>>
>>  However, it thought that was a filename in the current working tree, and
>>  sent a message for the last commit on the branch.
> 
> That does seem like a bug.  So you're changing it to do open() rather
> than open_containing() if a path is provided?
> 

Well, it already had a check for "if relpath and not tree", I just
changed it to "if relpath and not tree and location != '.'".

I believe what people want is the ability to be in a subdir and do "bzr
pqm-submit". I'm fine with that. What we are turning off is any case of
"bzr pqm-submit 'foo'" that doesn't directly refer to a branch.

So... effectively plain .open() but I just worked it into the way we are
doing it now.

It is a bit complex because we want to open a working tree if it is
available so that we can check it is clean. The way I would usually do
that is:

> 
>>  I didn't want to push this automatically without telling people, because
>>  it does disallow 'bzr pqm-submit' with no -m/--message.
> 
> I'd say go ahead and do it.  If there are complaints, we could add an
> option to use the last commit as the merge message (which users could
> turn on by default with an alias if they want the old behaviour).
> 
> James.
> 

Well, Ian Clatworthy has mentioned that he does use the last commit
message. I'll follow up with him a bit, because atm, I think that is
actually incorrect workflow. (Since it won't attribute the merge to him,
etc.)

I'll push.

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFH3pGdJdeBCYSNAAMRAgEeAKCNoGutGvfU76ojwRqjCfay1VGyZwCeISSB
aJr3Lq1GqjJJAUl+jyXD9SA=
=OL9h
-----END PGP SIGNATURE-----