RFR: 8339190: Parameter arrays that are capped during annotation processing report incorrect length

[Jan Lahoda]

On Wed, 23 Oct 2024 11:45:15 GMT, Nizar Benalla <nbenalla at openjdk.org> wrote:

> Can I please get a review for this change.
> Passes tier1, currently running tier 2-3.
> 
> The regression test creates temporary files in `tmpdir` to check that the warning is emitted correctly.
> I've also added the new warning to `example.not-yet.txt` as the example would require a very large file.
> 
> Here is the result of running `javap -c -p -v` on `ClassAnnotationWithLength_65536.class`  and `ClassAnnotationWithLength_65537.class` respectively. Results are the same before/after the change.
> 
> 
> static int x;
>     descriptor: I
>     flags: (0x0008) ACC_STATIC
>     RuntimeVisibleAnnotations:
>       0: #14(#15=s#16,#17=I#18,#19=[J#20])
>         CustomAnno(
>           value="custom"
>           count=42
>           arr=[-1l]
>         )
> 
> 
> 
>   static int x;
>     descriptor: I
>     flags: (0x0008) ACC_STATIC
>     RuntimeVisibleAnnotations:
>       0: #14(#15=s#16,#17=I#18,#19=[J#20,J#20])
>         CustomAnno(
>           value="custom"
>           count=42
>           arr=[-1l,-1l]
>         )
> 
> 
> TIA

I am sorry, but producing warnings is IMO a wrong solution. While the Java language does not impose limits, the classfile format does, and if we cannot generate a classfile that corresponds to the input Java file, javac should not produce anything and should finish up with a compile-time error. That is what is done for other limits as well. Like, for example, here:
https://github.com/openjdk/jdk/blob/964d8d2234595afaf4dfe48ea5cacdbfd3792d03/src/jdk.compiler/share/classes/com/sun/tools/javac/jvm/Gen.java#L940

(Although this check for annotations may need to be done in `ClassWriter`.)

-------------

PR Comment: https://git.openjdk.org/jdk/pull/21663#issuecomment-2432411334