Re: [lto] PATCH: DIE readers for struct/union/class types, + other bug fixes

[Mark Mitchell <mark at codesourcery dot com>]

Sandra Loosemore wrote:
> Mark Mitchell wrote:
>
> > > +       /* FIXME: mess with propagating mutable/volatile/etc attributes
> > > +          back to parent struct type.  */
> >
> > I'd prefer this to be:
> >
> >   if (TYPE_MUTABLE_P || TYPE_VOLATILE_P)
> >     sorry (...);
> >
> > Blowing up for things we don't understand is going to make it easier 
> > to debug things later.  But, it's not a huge deal either way.
>
> Part of the problem here is that I haven't figured out yet where to 
> stuff the bits when parsing the attributes from the DWARF information.  
> E.g., DECL_MUTABLE_P is specific to the C++ front end.

Some of these things plain don't matter: if the middle-end has no 
knowledge of it, and it's not referenced by any langhooks, then it's not 
relevant to code generation.  The only impact is on generated debug 
information from the LTO front-end, and there we may eventually want to 
just copy the debug information we were given as input, rather than try 
to regenerate it.

> Same problem here.  At this point, those tags are long gone; what we 
> have is whatever tree structures they're going to expand into, which I 
> haven't figured out yet, so I don't know how to test for them.

OK, I understand.  But, if they're long gone, we don't really need the 
comment here; it's wherever we're not translating them into tree 
structure that we should handle them, right?

Thanks,

--
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713