[PATCH] Add return value attribute to C interface

[Eric Christopher]

He's only somewhat active, but I'm using some of the same stuff these
days (or will).

The idea behind the C interface is "As needed" so that we don't have
to worry about the maintainability of it. Is there some use you have
for being able to do this via the C interface? Basically you'll need
to support arbitrary attributes if necessary. Having a good idea of an
interface for this is at least one reason why we haven't added it to
the C API other than lack of users.

-eric


On Wed, Jul 2, 2014 at 12:01 AM, deadal nix <deadalnix at gmail.com> wrote:
> I see. Is Bill Wendlign still active ? Maybe he has some comment on this.
>
> So what needs to be done is basically have a C interface to manipulate
> attribute sets ? Is there any pitfall I may not be aware of, or is this not
> done simply because nobody has the time to do it ?
>
>
>
>
> 2014-07-01 11:25 GMT-07:00 Reid Kleckner <rnk at google.com>:
>
>> The larger picture is that LLVM IR supports arbitrary string attributes,
>> and a bitmask is no longer an appropriate representation for attributes.
>> Bill Wendling never extended the C API to support this.  Furthermore, we
>> have more than 32 known, single-bit, enum attributes, so even if we ignore
>> string attributes, a 32-bit bitmask is insufficient.
>>
>>
>> On Mon, Jun 30, 2014 at 8:18 PM, deadal nix <deadalnix at gmail.com> wrote:
>>>
>>> OK,
>>>
>>> Can you explain me the larger picture here ? I can probably come up with
>>> something. Being able to specify attributes on return value is important to
>>> me, and I can't use the C++ API.
>>>
>>>
>>> 2014-06-30 11:36 GMT-07:00 Reid Kleckner <rnk at google.com>:
>>>
>>>> I'm not excited to add new APIs that use LLVMAttribute.  See this
>>>> comment from the end of the enum in Core.h:
>>>>     /* FIXME: These attributes are currently not included in the C API
>>>> as
>>>>        a temporary measure until the API/ABI impact to the C API is
>>>> understood
>>>>        and the path forward agreed upon.
>>>>     LLVMAddressSafety = 1ULL << 32,
>>>>     LLVMStackProtectStrongAttribute = 1ULL<<33,
>>>>     LLVMCold = 1ULL << 34,
>>>>     LLVMOptimizeNone = 1ULL << 35,
>>>>     LLVMInAllocaAttribute = 1ULL << 36,
>>>>     LLVMNonNullAttribute = 1ULL << 37,
>>>>     LLVMJumpTableAttribute = 1ULL << 38,
>>>>     */
>>>>
>>>> We haven't addressed this for over a year.
>>>>
>>>> If you're interested in fleshing out the C API attribute support, do you
>>>> mind proposing something that solves this problem?
>>>>
>>>>
>>>>
>>>> On Mon, Jun 30, 2014 at 11:24 AM, deadal nix <deadalnix at gmail.com>
>>>> wrote:
>>>>>
>>>>> Pong \o/
>>>>>
>>>>>
>>>>> 2014-06-26 22:00 GMT-07:00 deadal nix <deadalnix at gmail.com>:
>>>>>
>>>>>> ping ?
>>>>>>
>>>>>>
>>>>>> 2014-06-24 16:02 GMT-07:00 deadal nix <deadalnix at gmail.com>:
>>>>>>
>>>>>>> ping !
>>>>>>>
>>>>>>>
>>>>>>> 2014-06-20 18:39 GMT-07:00 deadal nix <deadalnix at gmail.com>:
>>>>>>>
>>>>>>>> ping ping ?
>>>>>>>>
>>>>>>>>
>>>>>>>> 2014-06-17 19:11 GMT-07:00 deadal nix <deadalnix at gmail.com>:
>>>>>>>>
>>>>>>>>> ping ?
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> 2014-06-14 23:47 GMT-07:00 deadal nix <deadalnix at gmail.com>:
>>>>>>>>>
>>>>>>>>>> As per title. Right now, the API do not provide any way to set
>>>>>>>>>> attributes for return values of function.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> llvm-commits mailing list
>>>>> llvm-commits at cs.uiuc.edu
>>>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>>>>
>>>>
>>>
>>
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>