2010/9/21 Eliot Miranda <eliot.miranda at gmail.com>:
>
>
> On Tue, Sep 21, 2010 at 1:14 AM, Nicolas Cellier
> <nicolas.cellier.aka.nice at gmail.com> wrote:
>>
>> 2010/9/21 Nicolas Cellier <nicolas.cellier.aka.nice at gmail.com>:
>> > 2010/9/21 Casey Ransberger <casey.obrien.r at gmail.com>:
>> >> Currently we have three more errors than we had in the last build that
>> >> I
>> >> rolled. I'm looking at a run on #10532.
>> >>
>> >> In general I'd recon it's an improvement, except for the errors (I
>> >> don't
>> >> mind new failures as long as the total number of tests has gone up by
>> >> an
>> >> equal or greater amount, but errors just give me the willies.) Also,
>> >> the
>> >> errors are to do with some stuff that looks like it could be hairy.
>> >> Blocks
>> >> and finalization. If these can be blessed by more experienced folks
>> >> than me
>> >> as "not that big a deal" then I would be quite happy to roll a build
>> >> tomorrow night.
>> >> Here's the numbers:
>> >>
>> >> 2803 run, 2762 passes, 9 expected failures, 28 failures, 4 errors, 0
>> >> unexpected passes
>> >>
>> >> 25 new tests
>> >>
>> >> 9 expected failures (no change)
>> >>
>> >> 28 failures (2 more than the previous image)
>> >>
>> >> 4 errors (3 more than the previous image)
>> >>
>> >> Errors:
>> >>
>> >> EventManagerTest>>
>> >>
>> >> #testBlockReceiverNoArgs
>> >>
>> >> #testBlockReceiverOneArg
>> >>
>> >> #testBlockReceiverTwoArgs
>> >>
>> >
>> > Oh, I see, it was me introducing clean-up from Juan for
>> > http://bugs.squeak.org/view.php?id=7352 in
>> > http://source.squeak.org/trunk/Kernel-nice.482.mcz
>> >
>> > Obviously, WeakActionSequence now restricts element class to
>> > MessageSend and WeakMessageSend...
>> > I see three solutions:
>> >
>> > 1) implement the 3 compatibility messages into classes of "valuable"
>> > objects (like BlockClosure)
>> > 2) wrap those valuable object into a (MessageSend receiver:
>> > valuableObject selector: #value) at creation time
>> > 3) revert the changes... and reintroduce the bug :(
>> >
>> > I don't like 1 because it's spreading very specific messages in the
>> > system.
>> > I don't like 3 because there's enough bugs in the image.
>> > I much prefer 2)
>> >
>> > Any other idea?
>> >
>>
>> On the other hand there is already a protocol 'events-support' in
>> BlockClosure with two messages #asMinimalRepresentation #isValid, so
>> 1) was the previous option adopted in Squeak.
>
> +1.  But is 'events-support' correct or is 'finalization support' better?

Oh, my change is relative to events #when:evaluate: #when:send:to:
rather than finalization.
I did not address WeakFinalizersTest>>#testNewFinalizationSupported,
cf. recent queries from Igor for this one.

Nicolas

>>
>> Note that the second message #isValid is not used anymore after Juan's
>> clean-up and could be removed.
>>
>> Nicolas
>>
>> > Nicolas
>> >
>> >> WeakFinalizersTest>>#testNewFinalizationSupported
>> >>
>> >> --
>> >> Casey Ransberger
>> >>
>> >>
>> >>
>> >>
>> >
>>
>
>
>
>
>