[llvm] r211309 - Add StringMap::insert(pair) consistent with the standard associative container concept.

[David Blaikie]

On Fri, Jun 20, 2014 at 11:41 AM, Agustín K-ballo Bergé
<kaballo86 at hotmail.com> wrote:
> On 19/06/2014 09:37 p.m., Rafael Espíndola wrote:
>>
>> Looks like this broke the build with gcc 4.7 in some bots and I reverted
>> it.
>>
>> On 19 June 2014 16:08, David Blaikie <dblaikie at gmail.com> wrote:
>>>
>>> Author: dblaikie
>>> Date: Thu Jun 19 15:08:56 2014
>>> New Revision: 211309
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=211309&view=rev
>>> Log:
>>> Add StringMap::insert(pair) consistent with the standard associative
>>> container concept.
>>>
>>> Patch by Agustín Bergé.
>>>
>>> Modified:
>>>      llvm/trunk/include/llvm/ADT/StringMap.h
>>>      llvm/trunk/lib/Support/StringMap.cpp
>>>      llvm/trunk/unittests/ADT/StringMapTest.cpp
>>>
>
> It seems this is due to a regression in gcc 4.7 when movable-only types are
> used in conjunction with `std::pair`. I was able to reproduce this with 4.7,
> but not with 4.6 nor 4.8. Interestingly, this problem doesn't show if the
> copy-constructor is defined private and deleted, it happens only when the
> copy-constructor is declared private and not defined.
>
> Looks like the test case should have been used `LLVM_DELETED_FUNCTION`.

I'm OK with using LLVM_DELETED_FUNCTION - it's better anyway from a
diagnostic perspective and if it happens to workaround a buggy old
compiler, I'm OK with that too.

> Is
> this an acceptable workaround (assuming it works), or is using movable-only
> types in pairs just something to avoid because of this gcc 4.7 regression?
>
> Regards,
> --
> Agustín K-ballo Bergé.-
> http://talesofcpp.fusionfenix.com
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits