[Vm-dev] newDelta >0 failed in sqUnixMemory -- Was Re: [Pharo-project] Incompatibility between Pharo and Seasidehosting

[Bruce O'Neel]

Hi,

I downloaded Ian's image/changes bundle named unix-3.10-4.vmm.tar.gz from http://www.squeakvm.org/unix/
and looked into it a bit.

In ObjectMemory biasToGrow it looks like the problem is 

biasToGrow
	| growSize |
	growSize :=  growHeadroom*3/2 - (self sizeOfFree: freeBlock) 
	self growObjectMemory: growSize

there are no '.'s after the lines.  Shouldn't it look like:

biasToGrow
        | growSize |
        growSize :=  growHeadroom*3/2 - (self sizeOfFree: freeBlock).
        self growObjectMemory: growSize.

that?

It looks like the slang interpreter/translater must be coughing a bit on what is basically 
slight invalid syntax, right?

I guess that next up I have to figure out how to actually run vm maker :-)

cheers

bruce

On Fri, Feb 20, 2009 at 12:32:52PM +0100, Bruce O'Neel wrote:
>  
> Hi,
> 
> Um, yes, I know you've discussed this before, but....
> 
> I'm not sure it's just a gcc 4.3.2 problem.
> 
> I downloaded and poked at the tar file of 3.10-4's source.  I got this from 
> http://squeakvm.org/unix/.
> 
> The error message comes from uxGrowMemoryBy, but, I think it comes from 
> being called (two frames higher) in incrementalGC.
> 
> The code is (from the gnu-interp.c file, though interp.c is the same):
> 
>                 if (((((usqInt) ((longAt(foo->freeBlock)) & AllButTypeMask))) < (((usqInt) foo->growHeadroom))) && (foo->gcBiasToGrow > 0)) {
>                         /* begin biasToGrow */
>                         /* begin growObjectMemory: */
>                         foo->statGrowMemory += 1;
>                         limit = sqGrowMemoryBy(foo->memoryLimit, growSize);
>                         if (!(limit == foo->memoryLimit)) {
>                                 foo->memoryLimit = limit - 24;
>                                 initializeMemoryFirstFree(foo->freeBlock);
>                         }
>                         weDidGrow = 1;
>                 }
> 
> So we call sqGrowMemoryBy with growSize which is used to calculate newDelta in uxGrowMemoryBy.  
> 
> A search of incrementalGC though shows that growSize, a sqInt is declared it's never set.
> It turns out that with my system (Debian Lenny ppc, with gcc 4.3.2) growSize is zeroed.
> Maybe that's the 'problem' with gcc 4.3.2? 
> 
> If I set it to some number, say, 100000, then a print statement just before
> sqGrowMemoryBy comes out and squeak grows its memory, and, volia, no crash.
> 
> Now, this might just be bypassing the real problem, but, it's unclear to me how
> an uninitialized growSize would be a good idea.
> 
> This also explains why my previous workaround, using the flag -memory on the command line,
> caused the problem to not occur.
> 
> cheers
> 
> bruce
> 
> On Thu, Feb 19, 2009 at 09:18:49AM -0800, John M McIntosh wrote:
> > look in the archive and on the vm-dev list I think people agree this  
> > is a gcc 4.3 issue.
> > 
> > On 19-Feb-09, at 4:15 AM, Bruce O'Neel wrote:
> > 
> > > Hi,
> > >
> > > Is it possible this error?
> > >
> > > sqUnixMemory.c:172: uxGrowMemoryBy: Assertion `newDelta >= 0` failed.
> > > Aborted
> > >
> > > I get this following the procedure below on a Debian Lenny PPC  
> > > system with Squeak 3.10-4.
> > > Gcc is 4.3.2.  I do not get this on a Ubuntu hardy ppc system with  
> > > the same version of squeak.
> > > Gcc is slightly older at 4.2.3.
> > >
> > > I've worked around this, for the moment, by adding -memory 200m to  
> > > the squeak
> > > command line.
> > >
> > > I spent some weekend debugging this, but, I didn't find where the  
> > > problem it is comes
> > > from.
> > >
> > > cheers
> > >
> > > bruce
> > 
> > --
> > = 
> > = 
> > = 
> > ========================================================================
> > John M. McIntosh <johnmci at smalltalkconsulting.com>
> > Corporate Smalltalk Consulting Ltd.  http://www.smalltalkconsulting.com
> > = 
> > = 
> > = 
> > ========================================================================
> > 
> > 
> > 
> > 
> > _______________________________________________
> > Pharo-project mailing list
> > Pharo-project at lists.gforge.inria.fr
> > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project