That's a bug in ReadStream >> #match:. The behavior shouldn't have 
changed.


Levente

On Sun, 28 Jun 2020, Jakob Reschke wrote:

> Fix proposed as Monticello-jr.726. The bug was triggered by the
> optimized ReadStream>>match: method introduced with
> Collections-ul.900, which does not set the stream to the end. The
> traitComposition is read from a ReadStream of the class definition
> string, hence the wrong read result if there are no traits in the
> definition.
>
> Am Sa., 27. Juni 2020 um 11:17 Uhr schrieb Jakob Reschke
> <forums.jakob at resfarm.de>:
>>
>> Am Sa., 20. Juni 2020 um 02:57 Uhr schrieb Levente Uzonyi
>> <leves at caesar.elte.hu>:
>> >
>> > Some class definitions
>> > don't have trait compositions, but the reader is happy to read the first
>> > two lines of the next class's definition as trait composition and class
>> > trait composition.
>> >
>>
>> I see the same thing happening when loading the INIFile package. This
>> breaks my smalltalkCI tests even further, because Monticello now
>> thinks that the class INIFile depends on the class INIFile and warns
>> about it during the load... Curiously it seems to work when I do this
>> in a Trunk image, although in smalltalkCI I also used a Trunk image,
>> hmm.