PATCH: fix building invalid IR lowering initializer list

[Richard Smith]

On Tue, Mar 17, 2015 at 5:20 PM, Nick Lewycky <nlewycky at google.com> wrote:

> On 16 March 2015 at 19:24, Nick Lewycky <nlewycky at google.com> wrote:
>
>> On 16 March 2015 at 19:17, Richard Smith <richard at metafoo.co.uk> wrote:
>>
>>> Thanks, LGTM!
>>>
>>
>> Thanks, r232454!
>>
>
> Got reverted and root caused in PR22940, here's another attempt at the
> patch which doesn't break the testcase in PR22940. Please review!
>

It might be cleaner to move the getBitCast calls into the caller of
createReferenceTemporary.


> Nick
>
> On Mon, Mar 16, 2015 at 5:19 PM, Nick Lewycky <nlewycky at google.com> wrote:
>>>
>>>> On 16 March 2015 at 16:19, Richard Smith <richard at metafoo.co.uk> wrote:
>>>>
>>>>> On Mon, Mar 16, 2015 at 4:00 PM, Nick Lewycky <nlewycky at google.com>
>>>>> wrote:
>>>>>
>>>>>> The attached patch fixes a case where we produce an invalid store
>>>>>> where the value type is not the same as the pointee type of the pointer.
>>>>>> This occurs when lowering to a global variable because the GV must have the
>>>>>> type of a pointer to its initializer, yet the initializer is not quite the
>>>>>> same that we use for the value. Fix it up by adding a cast.
>>>>>>
>>>>>> Ben, I've cc'd you because I think this is a regression from your
>>>>>> r231564.
>>>>>>
>>>>>> Please review!
>>>>>>
>>>>>
>>>>> Do the SD_Thread and SD_Static cases need the same handling? It looks
>>>>> like both of those cases can also return a Value with the wrong Type.
>>>>>
>>>>
>>>> They do! Patch attached.
>>>>
>>>> Nick
>>>>
>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20150317/dcd9dffe/attachment.html>