[patch] Correctly encode retained readonly properties

[Nico Weber]

On Tue, May 7, 2013 at 1:39 PM, jahanian <fjahanian at apple.com> wrote:

>
> On May 7, 2013, at 1:20 PM, Nico Weber <thakis at chromium.org> wrote:
>
> Thanks for the review!
>
>
> On Tue, May 7, 2013 at 1:12 PM, jahanian <fjahanian at apple.com> wrote:
>
>> Patch is OK. But I am concerned about producing ABI incompatibility with
>> older clangs.
>>
>
> That's a valid concern. It _does_ restore ABI compatibility with gcc; some
> code in Chromium apparently relied on the original behavior.
>
> Also, similar problem exists
>> with the ‘copy’ attribute (and probably others I haven’t tried).
>>
>
> 'copy' and 'retain' can't be used together as far as I know.
>
>
> Sorry for  not being clear. I meant when attribute is ‘copy’ (instead of
> ‘retain’).
>

Ah, good point. Tweaked patch attached.


>
> - Fariborz
>
>
>
>> Let me run this by our runtime guys to see its impact.
>>
>
> Thanks!
>
>
>>
>> - Fariborz
>>
>> On May 7, 2013, at 12:25 PM, Nico Weber <thakis at chromium.org> wrote:
>>
>> Hi,
>>
>> clang omits the '&' for retained properties if the property is also
>> readonly. The attached patch fixes this – it makes clang match gcc's
>> behavior and fixes PR15928.
>>
>> Ok?
>>
>> Nico
>> <clang-retain.patch>_______________________________________________
>> cfe-commits mailing list
>> cfe-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>>
>>
>>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20130507/e3362aec/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: clang-retain.patch
Type: application/octet-stream
Size: 1773 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20130507/e3362aec/attachment.obj>