On 31 January 2013 06:31, Bert Freudenberg <bert at freudenbergs.de> wrote:
>
> Am 30.01.2013 um 19:50 schrieb Frank Shearar <frank.shearar at gmail.com>:
>
>> On 30 January 2013 18:33,  <commits at source.squeak.org> wrote:
>>> Frank Shearar uploaded a new version of Kernel to project The Inbox:
>>> http://source.squeak.org/inbox/Kernel-fbs.735.mcz
>>>
>>> ==================== Summary ====================
>>>
>>> Name: Kernel-fbs.735
>>> Author: fbs
>>> Time: 30 January 2013, 6:33:47.078 pm
>>> UUID: 18ffff61-cfcf-4843-8c64-caea3ceb3fbc
>>> Ancestors: Kernel-fbs.734
>>>
>>> Actually, return a Message, and find out the receiver of the Message from the signalercontext.
>>>
>>> Also, _returning_ the value of #subclassResponsibility means that you can return to the original caller the value of your just implemented method.
>>>
>>> =============== Diff against Kernel-fbs.734 ===============
>>>
>>> Item was added:
>>> + ----- Method: ContextPart>>asMessage (in category 'converting') -----
>>> + asMessage
>>> +       | sender selector args |
>>> +       sender := self sender.
>>> +       selector := sender method selector.
>>> +       args := Array new: selector numArgs.
>>> +       1 to: selector numArgs do: [ :i | args at: i put: (sender tempAt: i)].
>>> +       ^ Message selector: selector arguments: args.!
>>>
>>> Item was removed:
>>> - ----- Method: ContextPart>>asMessageSend (in category 'converting') -----
>>> - asMessageSend
>>> -       | sender selector args |
>>> -       sender := self sender.
>>> -       selector := sender method selector.
>>> -       args := Array new: selector numArgs.
>>> -       1 to: selector numArgs do: [ :i | args at: i put: (sender tempAt: i)].
>>> -       ^ MessageSend receiver: self receiver selector: selector arguments: args.!
>>
>> <snip>
>>
>> Diffs serve as a helper for reviewers. To that end, the diffs ought to
>> actually show what changes would be applied to trunk should the change
>> be accepted. This diff, for instance, shows the removal of
>> #asMessageSend and the addition of #asMessage, but really the change
>> applied to trunk will be just the addition of #asMessage.
>>
>> In other words when something undergoes a few rounds of review (and
>> I'd think this should be the _norm_) the reviewer must reconstruct a
>> series of diffs to get an idea of how trunk will change.
>>
>> Wouldn't it be better to diff against trunk rather than against the
>> mcz's ancestor? (*)
>>
>> frank
>
> You should move obsolete versions out of the inbox. The diff is against the latest ancestor in inbox or trunk.

OK, but if we do that then we lose ancestry. I know what you're
hinting at, and I'd also like to see the equivalent of a git squash -
"take these versions in this branch and squish them together so that
the main line only sees a single commit". Right now that's pretty
onerous: it'd be a manual task, fraught with opportunities to mess up.
Er, actually, am I talking nonsense? Given some series of mczs, once
the chain/latest has passed peer review, I could

* load the latest versions into a clean up-to-date image
* file out the versions (so we flush the ancestry)
* load the fileouts into a clean up-to-date image
* commit to trunk, with suitable commit messages.

That's nearly automatable, except for the commit messages.

> We could change that to prefer trunk over inbox as diff base, agreed. Care to submit a squeaksource patch? :)

*ahem*. OK - that's http://source.squeak.org/SqueakSource-SqF/ right?

frank

> - Bert -