[RFC] Branches that fix bugs

[Jonathan Lange]

On Fri, Jun 26, 2009 at 7:58 AM, John Arbash
Meinel<john at arbash-meinel.com> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Jonathan Lange wrote:
>> The Launchpad developer discussion list recently started talking about
>> how we link our Bazaar version control data to bugs.
>>
>> Right now, we have two or three ways of linking branches & revisions
>> to bugs, none of which are used all the time. In particular, not many
>> people are huge fans of the --fixes option, since it seems to require
>> you to enter the bug fixing information at the wrong time.
>>
>> In the discussion, this list of desirable traits came up, and I
>> thought it would make a good starting point for a discussion on the
>> Bazaar list:
>>
>>  - I want to be able to link a branch to a bug offline
>>  - I want to be able to say "this branch is intended to fix this bug"
>> on branch creation
>>  - I'd like to change my mind about bug branch links
>>  - I'd like to distinguish between 'working on this bug' and 'fixes this bug'
>>  - I'd like to say that a bug was fixed pretty much independently of
>> when I commit.
>
> I assume this also includes removing that declaration?
>

Yes, "Turns out this branch isn't going to fix that bug after all" is
often independent of a commit that changes code.

Not necessarily in the sense of removing it from history, since I
don't really want to assume that level of detail.

>>  - On mainline branches, I would like to know which revision introduced the fix.
>
> As in, which merged it, or which committed the fix? (I think you mean
> the former.)
>

I do mean the former.

>>  - I want to see as much of this data as possible through the bzr
>> command line (again offline), ideally as part of bzr log.
>>
>> Are these clear? Would you change or remove any of these? Is there
>> anything you would add?
>>
>> jml
>
>
> Potentially:
>  - I want to see what revisions were committed that were intended to be
> part of this fix. (bzr log --related-to lp:33455)
>
>  - It would be good to mark a *group* of commits as fixing a bug, since
> I pretty much never implement the whole fix in a single commit.
> (Generally my --fixes commit is actually the one where I go update NEWS
> about having fixed the bug, which obviously has no actual code fix in
> itself.)
>

Both of these are quite interesting.

I can definitely empathize with the NEWS commit. Some Launchpadders
even do an empty commit with the fixes option.

jml