Re: [Boost-testing] Teeks clang tests all C++98?

Subject: Re: [Boost-testing] Teeks clang tests all C++98?
From: Tom Kent (lists_at_[hidden])
Date: 2017-02-28 12:28:20

On Sun, Feb 26, 2017 at 8:44 PM, Tom Kent <lists_at_[hidden]> wrote:

>
>
> On Thu, Feb 23, 2017 at 4:04 PM, Tom Kent <lists_at_[hidden]> wrote:
>
>>
>>
>> On Wed, Feb 22, 2017 at 1:16 PM, John Maddock via Boost-Testing <
>> boost-testing_at_[hidden]> wrote:
>>
>>> Nearly all the Teeks clang tests are in C++98 mode in-spite of their
>>> names, for example http://www.boost.org/developme
>>> nt/tests/develop/developer/output/teeks99-02-dc3-5-14-Docker
>>> -64on64-boost-bin-v2-libs-config-test-config_test-test-clang
>>> -linux-3-5~c++14-debug.html is listed as "C++14" but is actually C++98
>>> (check the __cpluplus version).  Which means the tests all sort of
>>> duplicate each other :(
>>>
>>
>> The change seems to have applied, the result page you referenced is now
>> different:
>> http://www.boost.org/development/tests/develop/developer/
>> output/teeks99-02-dc3-5-14-Docker-64on64-boost-bin-v2-
>> libs-config-test-config_test-test-clang-linux-3-5~c++14-debug.html
>>
>> Does that look like you'd expect? Is the error expected?
>>
>>
> It looks like that gets error is related to a pre-c++11 version of
> libstdc++ installed on the VM. I've attempted to update that
> teeks99-02-dc3.5-14 runner with a new version, we'll see how it goes.
>
> That raises the bigger issue of libstdc++ vs libc++ (or libstdcxx,
> STLport, etc). I'm thinking that maybe the clang instanced (>=3.5) should
> be switched to libc++? Any thoughts? Maybe a question for the developers
> list?
>

I've got a modern version of libstdc++ (6) installed on that machine, and
the config test now passes :-)

However, the affected clang regression runs jumped from taking 2hrs to
6.5hrs...is this expected? Are there some huge, intense tests that only
kickoff in C++11/14 mode? I had assumed from following the list that the
c++11/14 specific code was limited to just a handful of libraries.

Does anyone know if there is any test timing data available so that I could
look and find any big offenders?

Tom