launchpad-dev team mailing list archive

[Graham Binns <graham@xxxxxxxxxxxxx>]

On 6 January 2011 19:41, Graham Binns <graham@xxxxxxxxxxxxx> wrote:
> On 6 January 2011 18:59, Aaron Bentley <aaron@xxxxxxxxxxxxx> wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> On 11-01-06 01:49 PM, Robert Collins wrote:
>>> lpland is setting the message now, and I've seen some really confused
>>> ones - I think the bug is therein.
>>
>> I set the commit message when I create the merge proposal (which is our
>> recommended practise), so I view this as an anti-feature.  It just
>> prepends gibberish to the field and spams me as a result.
>>
>> I guess that's the downside of sharing software-- people will make
>> changes you don't want.
>>
>
> I had a thought about this when I was making dinner, and I think I've
> identified where things broke down (tl;dr: 50/50 PEBKAC / tool):
>
>  1. I ran bzr lp-land rather than utilities/ec2 land.
>  2. I realised my error when it asked me for a commit message and quit
> the editor.
>  3. I was then prompted ("The commit message has not been changed...Do
> you still want to use it, etc."). Instead of saying "No" to this, I
> CTRL-C'd.
>  4. I then ran utilities/ec2 land and wasn't prompted for a commit
> message (though I didn't think this odd; I should have).
>
> So, what appears to have happened - if my memory is right anyway - is
> that lp-land hasn't treated my KeyboardInterrupt in the same way as it
> would have tret me saying "No" to the prompt. (Note, I haven't yet
> looked at the code to confirm this; it's just my recollection of
> things). I'll check shortly.
>

Ah, so this looks like a bug in bzr itself, in
bzrlib.msgeditor.edit_commit_message_encoded()[1]. I'll file a bug
about this presently.

 [1] http://bazaar.launchpad.net/~bzr-pqm/bzr/bzr.dev/annotate/head:/bzrlib/msgeditor.py#L152

-- 
Graham Binns | PGP Key: EC66FA7D
http://launchpad.net/~gmb