[llvm] r208148 - llvm-cov: Implement --no-output

David Blaikie dblaikie at gmail.com
Wed May 7 09:15:29 PDT 2014


On Wed, May 7, 2014 at 9:12 AM, Justin Bogner <mail at justinbogner.com> wrote:
> David Blaikie <dblaikie at gmail.com> writes:
>> On Wed, May 7, 2014 at 2:00 AM, Timur Iskhodzhanov <timurrrr at google.com> wrote:
>>> I've worked around that in r208171.
>>>
>>> Please take a look and come up with a more permanent fix.
>>
>> This is essentially the semi-permanent fix (Justin can just remove the
>> comment). Reid's done this in other places for the same reason.
>>
>> One discussion we had is that we could modify STLExtras.h to check
>> compiler version and just bring in the MSVC version of make_unique
>> into the llvm namespace if it's available, and otherwise define our
>> own. Then there wouldn't be an ambiguity.
>
> I've removed the comment and fully qualified the two other calls to
> make_unique in the same function in r208219. Two questions:
>
> 1. I guess we don't have buildbots with the right version of MSVC to hit
>    this?  I didn't get any email.
>
> 2. Why wasn't the zero argument call to make_unique ambiguous?

I think it's ADL - you'd need at least one std-namespace argument to
cause the std::make_unique to be considered a candidate (since we
don't have any "using namespace std" in the code, generally).



More information about the llvm-commits mailing list