[BC] Block Closures, Version 2

John M McIntosh johnmci at smalltalkconsulting.com
Tue Feb 19 00:43:13 UTC 2002


>
>Acting like a new user, I downloaded a fresh 3.2gamma-4743 image and VM
>source from squeak.sourceforge.net.  Then I downloaded and filed-in
>CGeneratorEnhancements-ajh.cs, BCIntepreter-ajh.cs, and
>VMMaker-3-2-version5.4.cs in that order.  Then I opened VMMakerTool,
>made all plugins internal, and generated all.  Then I compiled the
>vm.  Then I download the pre-converted SqueakBC3.image (and changes)
>and ran it on the just built vm, and it worked.
>
>Can you please try the above recipe and see if it works for you?
>
>Thanks,
>Anthony

Ok, I did this exact procedure.

The only change for this VM over the other ones is that in the 64 bit 
conversion routines it uses instantiateSmallClasssizeInBytesfill 
versus the no-fill allocation that Rob had earlier suggested.

However the VM still crashes in the same manner, using the image you 
posted to the swiki

Now a clue might be how free space is allocated/used or limited or not.

As you pointed out in another note
?
In the bug above it looks like fp is off by two words because method =
longAt(fp-4) = a MethodContext2 instead of a CompiledMethod2.?


I still find it interesting that this object, wrong or not is the 
last object in youngspace before the freeBlock which implies it was 
just allocated. I'll note that we allocate 845,904 bytes of memory 
out of young space before the failure occurs.

-- 
--
===========================================================================
John M. McIntosh <johnmci at smalltalkconsulting.com> 1-800-477-2659
Corporate Smalltalk Consulting Ltd.  http://www.smalltalkconsulting.com
===========================================================================



More information about the Squeak-dev mailing list