Re: is this correct code?

  Le 09/12/2010 11:54, Jochen Theodorou a écrit :
> Hi all,
>
> I am currently trying to find an bytecode problem in a strange 
> situation. The code is fine if verified and normally run, but fails if 
> debugged in IntelliJ with:
>
> java.lang.VerifyError: (class: 
> org/codehaus/groovy/ast/builder/AstBuilderTransformation, method: 
> <clinit> signature: ()V) Attempt to split long or double on the stack
>
> No clinit looks like this:
>
>>   static <clinit>()V
>>     LDC 0
>>     INVOKESTATIC 
>> org/codehaus/groovy/runtime/typehandling/DefaultTypeTransformation.box (J)Ljava/lang/Object; 
>>
>>     CHECKCAST java/lang/Long
>>     ASTORE 0
>>     ALOAD 0
>>     INVOKESTATIC 
>> org/codehaus/groovy/runtime/typehandling/DefaultTypeTransformation.longUnbox 
>> (Ljava/lang/Object;)J
>>     PUTSTATIC 
>> org/codehaus/groovy/ast/builder/AstBuilderTransformation.__timeStamp__239_neverHappen1291891067850 
>> : J
>>     ALOAD 0
>>     POP
>>     LDC 1291891067849
>>     INVOKESTATIC 
>> org/codehaus/groovy/runtime/typehandling/DefaultTypeTransformation.box (J)Ljava/lang/Object; 
>>
>>     CHECKCAST java/lang/Long
>>     ASTORE 1
>>     ALOAD 1
>>     INVOKESTATIC 
>> org/codehaus/groovy/runtime/typehandling/DefaultTypeTransformation.longUnbox 
>> (Ljava/lang/Object;)J
>>     PUTSTATIC 
>> org/codehaus/groovy/ast/builder/AstBuilderTransformation.__timeStamp : J
>>     ALOAD 1
>>     POP
>>     RETURN
>>     RETURN
>>     NOP
>>     MAXSTACK = 2
>>     MAXLOCALS = 2
>
> And I see two potential problems.
>
> First there is the LDC 1291891067849, which should produce an int or 
> long? I quite never understood how a constant can be long and int at 
> the same time here. Anyway, next is a method call that takes an long 
> and we may or may not have one. Now I know that there is actually no 
> way to represent a real long constant over 1l, so this is probably 
> right and what the java compiler would have done as well.
>
> But before that there is also a LDC 0 followed by a box call that 
> takes a long. This is the only place I can imagine could cause a 
> problem. And I recently had the case that I did "LDC 0; LSTORE 5" and 
> the verifier complained... sometimes. I could fix it there by instead 
> having "LCONST_0; LSTORE 5".
>
> Now I am wondering about some things:
> (1) why is it valid to use LDC one time and at another not?
> (2) if it is not valid, why is it sometimes a VerifyError and 
> sometimes not?
> (3) shouldn't asm do a strict check here (I am using asm 3.2) and 
> point this out as error?
>
> bye blackdrag
>

Hi Jochen, ASM doesn't print if it's a LDC with a long or with an int 
(perhaps it should),
but in the constant pool the constant storing a long is different from 
the one storing an int
so there is no problem here.

I think the problem is at the end of your dump,
RETURN
RETURN
NOP
is a weird sequence, while if think it's a valid one from the verifier 
point of view,
it seems that it disturb IntellJ.

Rémi

-- 
You receive this message as a subscriber of the [email protected] mailing list.
To unsubscribe: mailto:[email protected]
For general help: mailto:[email protected]?subject=help
OW2 mailing lists service home page: http://www.ow2.org/wws