[m-rev.] for review: separate gcc and clang specific code in the runtime

Paul Bone pbone at csse.unimelb.edu.au
Thu Aug 11 15:43:02 AEST 2011


On Thu, Aug 11, 2011 at 03:20:19PM +1000, Julien Fischer wrote:
> 
> On Thu, 11 Aug 2011, Paul Bone wrote:
> 
> >On Wed, Jul 27, 2011 at 05:51:57PM +1000, Julien Fischer wrote:
> >>
> >>For review by anyone.
> >>
> >>runtime/mercury_atomic_ops.c:
> >>runtime/mercury_atomic_ops.h
> >
> >I'll review these files by modifing them myself.  What you've done is correct,
> >but there are some cases where clang appears to be compatible enough to support
> >the inline assembly and compiler intrinsics here.
> 
> Yes, however I think as a policy we ought to maintain a strict
> separation between GCC and clang here, i.e. if both of them support a
> particular extension then the condtion should be:
> 
>   #if defined(MR_GNUC) || defined(MR_CLANG)
> 
> In particular, we shouldn't be using __GNUC__ to also mean clang.
> 

Agreed, that's what I was intending to do.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 490 bytes
Desc: Digital signature
URL: <http://lists.mercurylang.org/archives/reviews/attachments/20110811/d9621438/attachment.sig>


More information about the reviews mailing list