SystemTracer is broken again (was: Re: [squeak-dev] The Trunk: Collections-nice.386.mcz)

Levente Uzonyi leves at elte.hu
Sun Oct 3 03:18:38 UTC 2010


On Sun, 3 Oct 2010, Levente Uzonyi wrote:

> On Fri, 1 Oct 2010, Levente Uzonyi wrote:
>
>> On Fri, 1 Oct 2010, Nicolas Cellier wrote:
>
> snip
>
>>> 
>>> Check whether #newHash: is defined too.
>> 
>> Both #hashMappedBy: and #identityHashMappedBy: is used by the SystemTracer. 
>> #newHash: is defined by the SystemTracer. I'll move #hashMappedBy: and
>
> Note that the method is actually called #newHashFor:, not #newHash:.
>
>> #identityHashMappedBy: there.
>
> I moved #hashMappedBy: and #identityHashMappedBy: to SystemTracer 
> (SystemTracer-ul.18 on SqueakSource). While I was testing the changes I found 
> that SystemTracer is broken again. The tracing finishes, but the resulting 
> image doesn't work. I tried it with a Squeak 4.1 image too and found that 
> it's broken as well. Something changed between January and April which broke 
> SystemTracer. I someone would like to test/fix it, just evaluate [ 
> SystemTracer2 writeImage: 'test.image' ]. Then run the VM with the new 
> test.image file.

I found the cause: since Squeak 4.1 Smalltalk is a SmalltalkImage and not 
a SystemDictionary. But the SystemTracing package didn't add #wordSize: to 
SmalltalkImage. So there was a DNU in the startup process of the new 
image. Fixed in SystemTracing-ul.19.


Levente

>
>
> Levente
>
>> 
>> 
>> Levente
>> 
>
> snip
>
>



More information about the Squeak-dev mailing list