Re: BC ABI stuff

[Bryce McKinlay <mckinlay at redhat dot com>]

Ranjit Mathew wrote:

> May I request that a section on the BC-ABI implementation
> be added to either the GCJ manual or the GCC Internals
> manual? You can perhaps copy-paste much of the material
> from your paper in the GCC summit.
>  
Definitely. I'm not sure if burying it in the gcc internals docs is the 
best place for it, perhaps we should start our own gcj document. I 
intend to work on this once more of the ABI details are fleshed out.

> Actually, it would be nice to have such a section for
> GCJ's implementation as a whole - class and object
> representation, interaction with the GC, etc.
>  

There is already some (possibly outdated) documentation of the internal 
runtime APIs etc scattered around the website. It would be cool to have 
it all in once place, though.

> > flag to indicate "don't try to read dependent classes, and don't
> > bother warning me about it either", but those things are being worked
> >    
>
> Sorry, I'm a bit rusty on this stuff, but why exactly? If
> I refer to a class directly (i.e. not via, say, Class.forName())
> isn't the compiler supposed to check that class?
>  

When compiling source code, yes. But when comping bytecode->native, it 
turns out that this is detrimental to GCJ's usability. There are just 
too many .jar files out there that are not shipped with, and do not even 
expect to have at runtime, all of their dependencies.

Regards

Bryce