[squeak-dev] The Inbox: Tools-fbs.338.mcz

Frank Shearar frank.shearar at angband.za.org
Fri Apr 15 19:28:07 UTC 2011


On 2011/04/15 18:45, Hannes Hirzel wrote:
> Ok, so then let's continue discussing if your contribution should be
> included into the trunk or if it should be in a loadable package on
> Squeaksource.

Hey? How could it possibly be a loadable package when it rewrites the 
guts of Browser? Tools-fbs.338 and friends are a refactoring of stuff 
already in the image. And I do mean "refactor" - no new functionality's 
added, except for fixing the fact that the Browser currently keeps 
losing your place whenever you add system categories, classes, method 
categories and methods.

(But yes, I _would_ like Tools to be a loadable package - we're still a 
long way from being able to do that though!)
>
> --Hannes
>
> On 4/15/11, Frank Shearar<frank.shearar at angband.za.org>  wrote:
>> On 2011/04/15 18:21, Hannes Hirzel wrote:
>>> On 4/11/11, Frank Shearar<frank.shearar at angband.za.org>   wrote:
>>>> On 2011/04/11 01:46, Levente Uzonyi wrote:
>>>>> On Fri, 8 Apr 2011, Frank Shearar wrote:
>>>>>
>>>>>> When I load it into a virgin image I follow the order above. I
>>>>>> wouldn't stake my life on it being entirely 100% correct. There might
>>>>>> be some interaction between the Morphic and Tools packages, in
>>>>>> particular, because the Morphic version has changes to how Browsers
>>>>>> open.
>>>>>>
>>>>>> I have, for instance, noticed that my image "goes away" sometimes when
>>>>>> I load Morphic. I can hit alt-. and abandon the process, and
>>>>>> everything Just Works.
>>>>>
>>>>> I finally set up my Trunk clone, where I could test the integration of
>>>>> your changes. I didn't face the problem you described, but I found
>>>>> another one. MessageSets (senders, implementors, etc) opened before the
>>>>> update will have their messageListIndex instance variable uninitialized.
>>>>> This means that #messageListIndex will return nil instead of an integer.
>>>>> So if I had some senders/implementors browsers open before the update, I
>>>>> get a debugger for each of them after it. Implementing #messageListIndex
>>>>> as
>>>>>
>>>>> ^messageListIndex ifNil: [ 0 ]
>>>>>
>>>>> solves this issue. The state of the browsers will not be preserved this
>>>>> way, but it's not worth writing complex migration code for this IMHO.
>>>>
>>>> Ah yes, because of instvar move: the new MessageSet instvar will be nil,
>>>> and nothing's yet set it to a valid value (i.e., an index).
>>>>
>>>> See version 339!
>>>>
>>>> frank
>>>>
>>>>
>>>
>>>
>>> What is the decision on this?
>>>
>>> And: Do we need to keep the old versions in the inbox?
>>
>> Levente said that _all_ versions should be into Treated when everyone's
>> happy, so that we have complete and entire Monticello histories.
>>
>> Of course, I can't comment on the decision :)
>>
>> frank
>>
>>
>
>
>




More information about the Squeak-dev mailing list