SystemTracer issue

Yoshiki Ohshima ohshima at is.titech.ac.jp
Fri Feb 23 01:05:31 UTC 2001


  Hello,

> At 12:56 22.02.2001 -0800, Yoshiki Ohshima wrote:
> >The another mystery is that occasionary a MethodContext
> >which is an invocation of Semaphore>>waitTimeoutMSecs: cause
> >out-of-bounds error.  Calliing "Delay shutDown" seems
> >prevent it, but it has another side-effect...
> 
> That has puzzled me too recently (a concurrent stack access, I guessed), 
> and I also wrapped the tracing doit with Delay shutDown/startUp, but what 
> exactly is that "side effect" you mention?

  Well, when something go wrong while tracing, the user
cannot interact the system anymore because nothing is
scheduled.

> Also, I see a dormant hiccup in #writeContext: at the line
> 
>     length _ BlockContext instSize + aContext basicSize.
> 
> where actually both MethodContexts and BlockContexts come 
> there.

  Yes, I noticed it but I will let it go as long as it is
fine:-)

  Another "hiccup" is that the first line of
#writeSet:useIdentity: should read 

	obj basicSize > 0 ifTrue: [self halt. ...

not

	self basicSize > 0 ifTrue: [...

.

> But I currently have other problems with traced images (yes, the darn 
> #cannotReturn of the clamped sender context hits hard, and my mind boggles 
> looking at EmergencyEvaluators...).

  You mean you're trying to fix it, or trying to produce 
an image in special format?

  -- Yoshiki





More information about the Squeak-dev mailing list