Re: [patch i386]: For 4.6 add support of thiscall calling convention attribute for x86

[Kai Tietz <Kai dot Tietz at onevision dot com>]

gcc-patches-owner@gcc.gnu.org wrote on 09.04.2010 12:16:00:

> On Fri, Apr 9, 2010 at 10:59 AM, Kai Tietz <ktietz70@googlemail.com> 
wrote:
> > 2010/4/7 Dave Korn <dave.korn.cygwin@googlemail.com>:
> >> On 07/04/2010 15:45, Kai Tietz wrote:
> >>> 2010/4/7 Gerald Pfeifer <gerald@
> >>
> >>>> The one additional comment/question I have is whether it is "on the 
386"
> >>>> and "the Intel 386" or just "on 386" and "Intel 386".  Also, what 
is the
> >>>> official name for that?  Perhaps say x86, or is this specific to 
386 as
> >>>> opposed to later generations?
> >>>
> >>> Well, x86 or ?86 would be more correct in terms of processor-family. 
I
> >>> used here the term, which was already used for fastcall convention. 
So
> >>> I have no problem about modifiying it to x86 if people agree to 
this.
> >>
> >>  I'm perfectly happy if you want to keep the wording as it is, 
because like
> >> you say it just matches the language of the existing paragraph. 
>  Technically,
> >> it's all ia32 architectures, so x86 would be more precise, but 
> only change it
> >> if you're willing to update the fastcall paragraph to match.  OK 
> by me either way.
> >>
> >>> Dave: Thanks for catching this in target definition about
> >>> fastcall/thiscall. But I don't think we should mention here
> >>> regparm-directive. regparm is using different registers,
> >>
> >>  Boh!  Right, then I agree it would be better not to mention it; 
> the idea was
> >> to make things clearer, not confuse them with an unhelpful "it's 
> just like X,
> >> execept for A, and B, and ....." analogy.
> >>
> >>    cheers,
> >>      DaveK
> >>
> >>
> >
> > I applied patch at revision 158155 with suggested corrections. I kept
> > here in documentation old naming, but this is possibly something for
> > an new patch to rename things here to x86.
> 
> I get
> 
> Running target unix/-m32
> FAIL: g++.dg/other/pr35504.C execution test
> 
> now on x86_64.  Wy did you amend an existing testcase instead
> of adding a new one?  And, obviously, thiscall doesn't work.
> 
> Richard.
> 

Hi Richard,

this patch fix this.

ChangeLog

2010-04-09  Kai Tietz  <kai.tietz@onevision.com>

        * config/i386/i386.c (x86_this_parameter): Handle aggregate for
        __thiscall convention.

Tested for i686-pc-linux-gnu, i686-pc-mingw32, and i686-pc-cygwin.

Ok, for apply?

Regards,
Kai


Regards,
 i.A. Kai Tietz

|  (\_/)  This is Bunny. Copy and paste Bunny
| (='.'=) into your signature to help him gain
| (")_(") world domination.

Attachment: thiscall_aggregate.diff
Description: Binary data