On 25/02/2014 15:53, Aaron Ballman wrote:
On Tue, Feb 25, 2014 at 10:51 AM, David Chisnall
<[email protected]> wrote:
On 25 Feb 2014, at 15:13, Alp Toker <[email protected]> wrote:
Changing the semantics means pain for our users until those compilers go away,
and compilers have a tendency to stick around :-/
And headers expecting the old behaviour. FreeBSD and OS X both ship system
headers that use __has_attribute(), shooting off the feet of our users is
simply not acceptable.
There's some misinformation happening which I'll clear up -- the
semantics are the same as always. If you provide __has_attribute with
a plain identifier, it behaves the same as it always has. The only
difference this patch proposes is allowing additional syntax where you
specify more than just a plain identifier.
Right, the discussion relates to making the proposed syntax easy to use
in a way that won't break existing clang deployments further down the line.
Existing headers using __has_attribute() are unaffected by this either way.
Alp.
--
http://www.nuanti.com
the browser experts
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
