Alan,

Just FYI: It has been recently noted that *all* float operations on the SH3
are more or less broken. The reason is that the VM got compiled with a flag
that confuses the word order of the float (I thought that a definition of
_SH3_ implies _MIPS_ since it's basically a mips variant) but the above
basically means that you'll have to recompile a VM on your own to work
reliably.

  - Andreas

> -----Original Message-----
> From: aldel at alum.mit.edu [mailto:aldel at alum.mit.edu]
> Sent: Tuesday, October 17, 2000 6:56 PM
> To: squeak at cs.uiuc.edu
> Cc: recipient list not shown
> Subject: Re: Running on HP Jornada 548
> 
> 
> Thanks for the help, Bob.  I was able to avoid the crash by simply
> changing Float>>truncated to always return 100 if the primitive fails.
> So it's pretty clear that that primitive (51) is broken on the SH3 VM.
> Maybe I'll try to dig into the VM source code a bit, although I
> currently have no idea what to look for there.
> 
> Squeak no longer crashes (except when I try to save the image, but
> that might be a memory limitation), but now I can't open a new window.
> I actually noticed this before, when I tried to open the "Play With
> Me" windows, but I thought I'd deal with one problem at a time.  What
> happens is that the "Play With Me" bar animates itself up to the top
> of the screen, then disappears, and then I just see the close button
> and the minimize button from the title bar, and not the rest of the
> window.  And those buttons don't respond; clicking on them just calls
> up the main menu, as if I had clicked on the background.  I get a
> similar result (sans the animation) if I try to move a window (which
> formerly crashed the VM) or open a browser or workspace.  So how do I
> debug this problem?
> 
> Also, I still don't know how to do a right- or middle-click or an
> alt-keystroke.  Does the WinCE VM provide a way to do these when
> there's no keyboard?
> 
> Thanks!
> 
> Alan deLespinasse
> 
> >Date: Sun, 15 Oct 2000 16:15:48 -0400 (EDT)
> >From: Bob Arning <arning at charm.net>
> >
> >Alan,
> >
> >This problem has been mentioned several times on the list, 
> but I have not seen an answer posted. Here is a reply I sent 
> to one of the earlier messages (which seemed to point to some 
> failure in floating point arithmetic primitives). Perhaps you 
> could try this out and see if it provides any additional 
> insight into the problem.
> >
> >On Sat, 26 Feb 2000 23:44:23 CST "Francois Le Coguiec" 
> <francois_le_coguiec at hotmail.com> wrote:
> >>Can someone please help me to understand the debugger walkback
> >>in the following attached zipped pictures?
> >
> >Francois,
> >
> >Here is what I can see from your images:
> >
> >In attempting to create a new color for part of a debugger 
> button, one of the color components remained a fraction after 
> a #rounded message, as in:
> >
> >(r * ComponentMax) rounded bitAnd: ComponentMask
> >
> >The resulting fraction did not understand the #bitAnd: 
> message. The real question is how #rounded returned a 
> Fraction. One strange possibility is in Float>>truncated 
> which, if the primitive fails, could return a fraction with:
> >
> >	^ self asTrueFraction.  "Extract all bits of the 
> mantissa and shift if necess"
> >
> >Some obvious questions:
> >
> >1. Can you repeat this error? If so, can you tell me how to 
> recreate it?
> >2. Do you know what caused the debugger to try to open in 
> the first place?
> >
> >If this can be recreated, the way I would attack it is to 
> add a method to Fraction (temporarily)
> >
> >bitAnd: something
> >
> >	Smalltalk at: #STRANGEBUG put: {self. thisContext shortStack}.
> >	^something
> >
> >This will provide a less fatal behavior and you can inspect 
> "Smalltalk at: #STRANGEBUG" afterwards to see a bit more 
> about what was happening.
> >
> >Cheers,
> >Bob
>