patchzilla idea

[Joe Walker]

On 29/03/2011 11:04, Max Kanat-Alexander wrote:
> On 03/29/2011 02:27 AM, Joe Walker wrote:
>> #1: Hello is spelled incorrectly
>> #2: World is spelled incorrectly
>>
>> The patches for those bugs are necessarily dependent, however the bugs
>> are not.
> 	Ah ha. I see what you are talking about, now. But those are really just
> conflicts, not dependencies. They apply without fixing only if you
> choose to do them in order, consciously. In that case, you could record
> their blocking relationship in the bug tracker.
True, you could. However that would mean you'd need to know. If the 2 
bugs were assigned to different people, you might not know - unless you 
had a system to work it out, ...

>> Also, I've seen several cases where the one problem == one patch rule
>> has not worked in practice. There could be 2 alternative implementations
>> of part of the solution, 2 distinct steps that are functionally
>> distinct, etc.
> 	Okay. I'm sure there are such cases, but I would like to see some
> specific ones. The examples you specified may not be the best:
>
> 	Two alternative implementations means that only one will be checked
> in--that's not so much of a problem. It would avoid a certain number of
> permutations in your auto-rebaser, but I don't know if it's a
> common-enough situation to warrant a whole application.
>
> 	Two distinct steps should definitely be two bugs, perhaps linked back
> to the original with the original being changed into a meta-bug.

But that starts to make the bugs dependent on the implementation 
details. I'm not sure they should be.

Bug #3: Joe can't spell color
Attachement #1: replaces instances of colour in module X
Attachement #2: replaces instances of colour in module Y

It could be useful to have them in separate patches because the modules 
have different owners, or just to keep things clear.
It doesn't seem right to have to create 2 separate bugs, (or even 3 if 
you're going to have a parent) in order to keep patches separate.

>> This was born out of a discussion in which all sorts of sacred cows were
>> being slaughtered. Including 'move everything to github' and so on. It
>> was an attempt to have a more manageable jump!
> 	Ah, okay. :-) I'm always for manageable jumps. :-) ("move everything to
> github" is definitely bad idea in so many ways.) I suspect that patch
> management may want to be a discussion you have with VCS vendors instead
> of with us, though--I suspect they have developed far more tooling
> around it than we would have even thought of.
There's the trouble - it's a solved problem with github pull requests 
(which is part of the reason for the clamor!) and Hg branches are too 
heavyweight to handle this. Plus we're used to working with patches...

Joe.