While I'd like to see support for hotpatching, the other Clang bugs
block it from being useful. If Mozilla is close to working, let them
have it.
Thanks for asking :).
-Austin
On Thu, Sep 29, 2011 at 21:39, Charles Davis <cdavis at mymail.mines.edu>
wrote:> Howdy Austin,
>
> I'm about to sacrifice the Hotpatch attribute (which I added a long
time ago to start fixing PR10212 before you even filed it) so the Mozilla guys
can have their ReturnsTwice attribute. The thing is, I didn't hook Hotpatch
up to anything yet, and I don't expect to get to it anytime soon; that's
why I'm giving it to them for now. If you don't want this to happen, you
better speak up now.
>
> Chip
>
> Begin forwarded message:
>
>> From: Patrick Walton <pwalton at mozilla.com>
>> Subject: Re: [patch] Add the returns_twice attribute
>> Date: September 29, 2011 4:43:52 PM MDT
>> To: Rafael Ávila de Espíndola <rafael.espindola at gmail.com>
>> Cc: llvm-commits <llvm-commits at cs.uiuc.edu>, Charles Davis
<cdavis at mymail.mines.edu>
>>
>> On 9/29/11 3:42 PM, Rafael Ávila de Espíndola wrote:
>>> The attached patch reuses the hotpatch attribute (which Charles
Davis
>>> says is unused) to create the returs_twice attribute. This will be
used
>>> to implement the C attribute with the same name.
>>>
>>> It is needed by valgrind, since it has its own setjmp like
function.
>>>
>>> See PR11038 for more information.
>>
>> I have a patch to expand the attributes field to 64 bits that seems to
work, so that's fine with me.
>>
>> Patrick
>
>
--
-Austin
