Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!zaphod.mps.ohio-state.edu!usc!snorkelwacker.mit.edu!bloom-beacon!eru!hagbard!sunic!mcsun!hp4nl!dutrun!winfwrd
From: winfwrd@dutrun.UUCP (Paul van der Weerd)
Newsgroups: comp.lang.smalltalk
Subject: Re: anyone using time-sliced scheduling w/st80?
Message-ID: <15211@dutrun.UUCP>
Date: 22 Mar 91 14:33:22 GMT
References: <HUGGINS.91Mar12085614@dino.ti.com> <4861@ns-mx.uiowa.edu> <3095@laura.UUCP> <15113@dutrun.UUCP> <3109@laura.UUCP>
Reply-To: winfwrd@dutrun.tudelft.nl.UUCP (Paul van der Weerd)
Organization: Delft University of Technology, The Netherlands
Lines: 62

In article <3109@laura.UUCP>, huba@ls5.informatik.uni-dortmund.de,
(Hubert Baumeister) writes:
>...
>I suppose you use Smalltalk 80 v2.5 or earlier. These versions assume that
>when drawing on a window this window is completely visible. To achieve this the
>TextCollectorViews use the method displaySafe: aBlock in ControlManager. This
>method tops the window and executes aBlock.
>When both TextCollectorViews want to display themselves at the same time
>there is a race condition in displaySafe:.

Indeed the ControlManager gets lost (Smalltalk-80 v2.5). So I extended the
example with extra protection:

| p1 p2 t1 t2 displayMutex |
"Protect the display code."
displayMutex _ Semaphore forMutualExclusion.
"Define some textcollectors."
[t1 := TextCollector new.
TextCollectorView open: t1 label: 'Process 1'] fork.
[t2 := TextCollector new.
TextCollectorView open: t2 label: 'Process 2'] fork.
"Define some processess."
p1 := [[true] whileTrue:
	[displayMutex critical: [t1 show: 'A']]] newProcess.
p2 := [[true] whileTrue:
	[displayMutex critical: [t2 show: 'B']]] newProcess.
"Run processes alternately."
(p1 priority: 3) resume.
(p2 priority: 3) resume.
300 timesRepeat: [
(Delay forMilliseconds: 200) wait.
(Processor suspendFirstAt: 3) resume].
p1 terminate.
p2 terminate.

This works fine. Also the appearance of the display during execution is okay.

Hubert Continues:
>In Smalltalk Release 4.0 this works fine since there it is possible to draw on
>partly hidden windows and thus it needs no hack like the one described above.

Hearing the word "hack" I added a classVariable DisplayMutex to
StandardSystemView and changed the class initialization method in:

!StandardSystemView class methodsFor: 'class initialization'!

initialize

	FillInHoles _ true.
	DisplayMutex _ Semaphore forMutualExclusion! !

This also works fine (no loss of control), but the display becomes a mess
during execution (use restore display afterwards).
Is this the kind of hack you had in mind?

--
Paul van der Weerd                  Tel. +31 15 782523
Delft University of Technology      Fax. +31 15 786522
Dept. of Computer Science           Email winfwrd@dutrun.tudelft.nl
Julianalaan 132
2628 BL  Delft
The Netherlands
...................................................................