[cfe-dev] local/unnamed enums as diagnostic parameters

[Richard Smith]

Generally, the factor blocking us using C++11 features within Clang is
lack of support for the feature from system compilers that might be
used to build Clang. If every compiler we care about supports this
feature (in its C++98 mode, where applicable -- and someone will need
to verify that), then I'm fine with disabling the warning, but I'd
prefer to do that at the LLVM level rather than at the Clang level.

Chris: does that sound reasonable to you?

On Fri, Jun 21, 2013 at 7:24 AM, Rafael Espíndola
<rafael.espindola at gmail.com> wrote:
> This is an extension to c++98 and OK in c++11.  I think it is a common
> extension and we can just disable the warning. Richard, what do you
> think?
>
> On 18 June 2013 04:12, Evgeniy Stepanov <eugeni.stepanov at gmail.com> wrote:
>> Hi,
>>
>> we've got a bunch of warnings on bootstrap bots since a few days ago
>> related to the use of local or anonymous enums in Sema::Diag(). This
>> results in template operator<< of SemaDiagnosticBuilder being
>> instantiated with an unnamed type, which, AFAIK, is invalid. Seems
>> like the right fix is cast such values to (int),
>>
>> http://lab.llvm.org:8011/builders/clang-x86_64-debian-selfhost-rel/builds/4751/steps/compile.llvm.stage2/logs/warnings%20%2810%29
>>
>> What really concerns me is the fact that self-hosted build of clang
>> has not been warning-free for several days, and no one seems to care.
>> And the bots do not enforce it. If this is still a goal (and I believe
>> it should be) then I suggest turning on -Werror on one or all
>> self-hosted bots and paying attention to regressions.
>> _______________________________________________
>> cfe-dev mailing list
>> cfe-dev at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev