I have sent you the original 2 versions of Publish Subscribe.
The WeakKeyDictionaries (publishers) have always been weak dependents of
WeakArray,
so that is not likely to be the source of the problems you have been having.
Finalization was handled this way right from version 0.
I think the problems are more likely to involve some sort of deadlock or
other synchronization problem. If you don't need process-safe code, the
quickest fix is simply to remove the #protected: block from the 3 or 4
methods
in PostOffice that use it. Then test using the class side tests
individually,
but don't run the 'start' process - that test is virtually guaranteed to
cause
problems.


I am going to look at the synchronization stuff in detail soon, but have
some
exams looming that I must take care of first.

Let me know if this fixes things.


>Peter:
>
>I'm working on my event system, again, and need at least a partial fix for
>you stuff.  Unfortunately, I tossed the older, working versions.  The
>problem seems to have started when you made the publishers dictionary a
weak
>dependent of WeakArray (if I recall correctly).  I don't remember how you
>handled finalization before that.  I don't need Smalltalk-process-safe
code,
>now, just something to get me by.
>
>(I don't know whether you received my previous e-mail:  the latest
>Publish-Subscribe change when tested will hang a clean 2.4c /w updates.)
>
>Thanks.
>
>Shaping
>
>