[V3dot9] Re: About condenseChanges and Traits

[stéphane ducasse]

Ok I understand because what I see is that the condenseSources broke  
the getPreamble (so the internal structure of the sources)
and as I do not know exactly what was the problem with  
condenseChanges I have problems to spot it.

Now it seems that this piece of code which works well in 7053 does  
not after we loaded some packages.


VersionsBrowser new
	scanVersionsOf: 		 
(TraitDescription>>#noteRecategorizedSelectors:oldComposition:) 	 
class: TraitDescription
	meta: false
	category: nil
	selector: #noteRecategorizedSelectors:oldComposition:.

I'm trying to find out with damien.
This would indicate that there is a bug in mechanism to manage changes.

Stef



On 23 août 06, at 13:27, Pavel Krivanek wrote:

> Hi Stef,
>
> there were some corrupted methods in the 7054 after condenseSources.
> The test showed that the problem with the same methods is before
> condesneSources, in version 7053 after evaluation of condenseChanges.
> The test worked well in some version before, I don't know the exact
> version and I wasn't doing this test by myself.
>
> -- Pavel
>
> On 8/23/06, stéphane ducasse <ducasse at iam.unibe.ch> wrote:
>> Hi pavel
>>
>> I'm confused:
>>
>>
>> In your email
>>
>> "
>> Corrupted condenseChanges again in 7053 (was "[ANN]  
>> Squeak3.9g-7054...")
>> Hi all,
>> it seems that the problem with corrupted sources Squeak3.9g-7054 in
>> has the source in corrupted condenseChanges in 7053. This is the
>> result of the condenseChanges test:
>> "
>> You seems to implied that condenseChanges was the problem in 7053.
>> Now you are saying the inverse.
>> I will redo the experience but I would like somebody else to help.
>> So now you are saying that the tests are ok in 7053?
>>
>> Stef
>>
>> > Yes, I have taken the 7053 image
>> > an http://mail.google.com/mail/?account_id=pavel.krivanek% 
>> 40gmail.comd
>> > evaluated the testing code that stores all souces into the  
>> dictionary,
>> > condense changes and then test if the sources are still the  
>> same. One
>> > small problem is that the image must be saved and reopened after
>> > condensing of changes because condenseChanges doesn't do correct
>> > reopening of the changes file with the right encoding.
>> >
>> > AFAIK this test was working well. Maybe we should test previous  
>> images
>> > to see where it stoped work well. Maybe is problem only with the
>> > method category name of broken methods because, as far as I  
>> remember,
>> > it contains space character.
>> >
>> > -- Pavel
>> >
>>
>>
>