Re: re-jigged for lang independence, Re: [Patch, Darwin, c*, ObjC* 1/2] Implement Darwin CFStrings.

[IainS <developer at sandoe-acoustics dot co dot uk>]

On 20 Oct 2010, at 09:48, Richard Guenther wrote:

> On Tue, Oct 19, 2010 at 7:14 PM, IainS <developer@sandoe-acoustics.co.uk 
> > wrote:
> > Hi Mike, Paulo
> > I wasn't quite quick enough...
> > .. sorry if I wasted reviewer's time.
> >
> > On 19 Oct 2010, at 18:01, Mike Stump wrote:
> >
> > > On Oct 19, 2010, at 1:48 AM, IainS wrote:
> > > > The changes need to be applied as one commit - hence one email.
> > >
> > > > OK for trunk?
> > >
> > > Ok.
> >
> > here is an update that addresses Paolo's comments.
> >
> > This is a language-independent variant (other than the ObjC  
> > interface).
> >
> > Essentially the same code (less replacing pushdecl_top_level()  with
> > DECL_CONTEXT=NULL_TREE).
>
> toplevel entities should have DECL_CONTEXT of the
> TRANSLATION_UNIT_DECL, not NULL these days.

So noted in the applied patch with a FIXME.
However, as you noted in irc this needs to be done once per TU
.. and currently each FE handles it differently (and none of them  
export the result where a builtin or other parser code can reach).

Would it be a good idea to introduce  
langhooks.decl.current_translation_unit?
(defaulting to NULL_TREE, I guess).
> RIchard.
>
> > Bootstrapped (incl.Java)  and checked ObjC* on i686-darwin9.
> >
> > I can't imagine why there would be any difference on a non-darwin  
> > platform -
> > but...
> > ... OK if it passes remaining regtests and on a non-darwin platform?


this took a while because of the testcycle...
... in addition to adding the  comment above two minor adjustments  
were needed:
to deal with different sized decls (use the lang hook instead of sizeof)
to make sure that builtin enum code is after all the machine ones.

applied as r165820
Iain

Attachment: 165817-cfsrings-amended.txt
Description: Text document