Re: Java: [BC] Implement type assertion table

[Bryce McKinlay <mckinlay at redhat dot com>]

Tom Tromey wrote:

> Bryce> Index: link.cc
> [ ... ]
> Bryce> +void
> Bryce> +_Jv_Linker::verify_type_assertions (jclass klass)
>
> Everything in _Jv_Linker should be independent of the execution engine
> (there's one bogus exception).  That's why all the interpreter-related
> code is in interpret.cc.  So it seems a little weird to put this code
> here.
>  

Doing the work to lay out indirect-dispatch tables and such is also 
"linking", but I agree that having the distinction between generic class 
layout and that which is specific to the BC-ABI may make sense. However, 
there is already  native-code/BC-ABI-specific functions are already in 
_Jv_Linker - link_symbol_table(), link_exception_table(), etc. We could 
separate these out into some other class/file, specific to the BC-ABI?

> OTOH if we plan to support this field in Class independent of the
> engine, then we might as well just inline the code into
> _Jv_Linker::verify_class().
>  
No, I do think this is native code/BC-ABI specific.

> Bryce> +      if (assertion_code == 0)
> Bryce> +        // End of the table.
> Bryce> +        return;
>
> Should probably use JV_ASSERT_END_OF_TABLE instead of "0" here.
>  
Right.

Regards,

Bryce