Re: [ASM 5.1] Error when reading nested type of type variable bound in method signature

Hello all,

Just to let you know, as a followup I created last week the following bug
report in Eclipse's bugzilla:
https://bugs.eclipse.org/bugs/show_bug.cgi?id=500473
No reaction so far,

Regards,

On Mon, 29 Aug 2016 at 18:00 Michael Gumowski <
[email protected]> wrote:

> Hello,
>
> Sorry for the delay answering you (holidays...).
>
> Thanks for the hint about the compiler. It's seems to be the root cause of
> my problem.
> I'm indeed not able to reproduce the issue using solely javac (1.8.0_91),
> as I'm getting the signature you shows. The signature is correctly
> generated (and parsed). However, I was observing the issue when dealing
> with the example I provided in my eclipse IDE, or when building a project
> with a similar structure using "ant" to build it and "ecj" as compiler.
>
> I tried to compare the decompiled code I'm getting when using:
>
>    - jdk1.8.0_91+javac
>    - ecj-4.6 (eclipse standalone batch java compiler)
>
> When looking only to the method, I'm getting the following differences:
>
> jdk1.8.0_91:
>
>   // access flags 0x0
>>
>   // signature
>> (Lorg/foo/bytecode/TypeWithNestedGenericClass$Nested<Ljava/lang/String;>;)V
>>   // declaration: void
>> method(org.foo.bytecode.TypeWithNestedGenericClass$Nested<java.lang.String>)
>>   method(Lorg/foo/bytecode/TypeWithNestedGenericClass$Nested;)V
>>    L0
>>     LINENUMBER 5 L0
>>     RETURN
>>    L1
>>     LOCALVARIABLE this
>> Lorg/foo/bytecode/GenericTypeUsingNestedTypeFromBound; L0 L1 0
>>     // signature
>> Lorg/foo/bytecode/GenericTypeUsingNestedTypeFromBound<TU;>;
>>     // declaration:
>> org.foo.bytecode.GenericTypeUsingNestedTypeFromBound<U>
>>     LOCALVARIABLE nested
>> Lorg/foo/bytecode/TypeWithNestedGenericClass$Nested; L0 L1 1
>>     // signature
>> Lorg/foo/bytecode/TypeWithNestedGenericClass$Nested<Ljava/lang/String;>;
>>     // declaration:
>> org.foo.bytecode.TypeWithNestedGenericClass$Nested<java.lang.String>
>>     MAXSTACK = 0
>>     MAXLOCALS = 2
>>
>
> eclipse compiler ecj-4.6:
>
>   // access flags 0x0
>>
>   // signature (TU.Nested<Ljava/lang/String;>;)V
>>   // declaration: void method(U.Nested<Ljava/lang/String, )
>>   method(Lorg/foo/bytecode/TypeWithNestedGenericClass$Nested;)V
>>    L0
>>     LINENUMBER 5 L0
>>     RETURN
>>    L1
>>     LOCALVARIABLE this
>> Lorg/foo/bytecode/GenericTypeUsingNestedTypeFromBound; L0 L1 0
>>     // signature
>> Lorg/foo/bytecode/GenericTypeUsingNestedTypeFromBound<TU;>;
>>     // declaration:
>> org.foo.bytecode.GenericTypeUsingNestedTypeFromBound<U>
>>     LOCALVARIABLE nested
>> Lorg/foo/bytecode/TypeWithNestedGenericClass$Nested; L0 L1 1
>>     // signature TU.Nested<Ljava/lang/String;>;
>>     // declaration: U.Nested<Ljava/lang/String
>>     MAXSTACK = 0
>>     MAXLOCALS = 2
>>
>
>
> From that point, it seems to me that the issue is therefore on Eclipse
> side, as they apparently does not follow the JLS when dealing with such
> cases.
> I'm probably going to open an ticket on eclipse side.
>
> Thanks for your help,
>
> *Michael GUMOWSKI* | *SonarSource*
> *Software Developer @ Language Team*
> http://sonarsource.com
>
> On 15 August 2016 at 19:54, Eric Bruneton <[email protected]> wrote:
>
>> This type signature seems incorrect to me (see the signature grammar
>> specification:
>> https://docs.oracle.com/javase/specs/jvms/se7/html/jvms-4.html#jvms-4.3.4
>> : T
>> can only be followed by an identifier itself followed by ';')
>>
>> If I compile your example with javac (1.8.0_91) I get the following
>> signature
>> for this method:
>> (LTypeWithNestedGenericClass$Nested<Ljava/lang/String;>;)V
>>
>> How did you get "(TU.Nested<Ljava/lang/String;>;)V"?
>>
>
>


-- 
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