On Mon, Oct 07, 2013 at 02:46:49PM -0700, Nick Kledzik wrote:
> 
> On Oct 3, 2013, at 1:37 PM, Joerg Sonnenberger <[email protected]> 
> wrote:
> 
> > On Thu, Oct 03, 2013 at 10:36:20AM -0700, Nick Kledzik wrote:
> >> Getting started:
> > 
> > There are a few high-level items to discuss after this gets committed.
> > The biggest question for me is whether to expose the unw_* API at all.
> > I'd prefer to get rid of that as it creates a number of complications
> > and it is not required for the desired goal.
> 
> Can’t that be a platform decision?  That is, those functions could be made
> attribute hidden on platforms that don’t want to expose them.

This matters e.g. for the way floating point and vector registers are
exposed. Without the unw_* API, they can be represented easily as fixed
size objects without attached semantics. The only required operations
are copying from and to the stack as well as the context manipulation.

Joerg
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to