Re: binary metadata

[Bryce McKinlay <mckinlay at redhat dot com>]

Tom Tromey wrote:

> Bryce> Having the compiler use it is an interesting idea, however -
> Bryce> the metadata is complete enough, so really its just down to some
> Bryce> design/technical challenges in having the compiler read from the
> Bryce> object files.
>
> The compiled code doesn't have the deprecated tag, and perhaps won't
> have @Retention(CLASS) annotations either.  (The new @Deprecated
> annotation seems to have @Retention(RUNTIME), which looks weird, but
> would let this work.)  That's a barrier to directly using the .so in
> the compiler.
Good point, I'd forgotten about those. I think if we are serious about 
having the compiler read .so's, however, these could be included in the 
metadata and just ignored by the runtime. The extra space when using 
pointer-free metadata shouldn't be too bad.

Regards

Bryce