Andreas Raab wrote:

> > sometimes it had (leaving off the addresses)
> >     Object>become:
> >     (someClass)>doesNotUnderstand:
> >     Behavior>removeSelectorSimply:
>
> This is pretty strange from my point of view. Why should result
> #doesNotUnderstand: in a #become: operation?!

Quite so. It strikes me that perhaps the stack is getting clobbered. Often no
stack dump is generated. When it does, the #doesNotUnderstand: line shows up
with various classes as the receiver or perhaps not at all. By looking
statically at the code, I would expect to see
    Object>become:
    MethodDictionary>removeKey:ifAbsent:
    Behavior>removeSelectorSimply:
because MethodDictionary does a "self become: copy" in #removeKey:ifAbsent:
(called from #removeSelectorSimply:) right after it does a 
#removeDangerouslyKey:ifAbsent:.

Ah ha! In reviewing the crash.dmp again, I see that one time recently it did
give the stack as shown above,  rather than a #doesNotUnderstand: line.
Strange, but odd.

> Good! Can you send me copy?

Sure. The image and changes file zip up to 2.5MB. Would you have any trouble
getting that as an email attachment? (I promise to try and remember not to
send it to the mailing list ;->)

> > I'm currently running in Windows NT 4.0 SP 3, using the VM "v2.0.1.10
> > 1998/07/23 19:26:53 raab Exp".
>
> If it's been in all version back to 1.XY then it's relatively unlikely
> that this is some platform specific problem (nearly everything has
> changed since then ;-) Are you able to reproduce your problem on any other
> platform?

Some of my earlier (non repeatable) times were on NT 3.51 before it was
upgraded to 4.0. I haven't tried it on anything else, since I'm currently
surrounded by Windows machines.

-------------------------------------------
Bill Dargel            wdargel at shoshana.com
Shoshana Technologies
100 West Joy Road, Ann Arbor, MI 48105  USA