Looking for a C++11 support Overview

Isaac Gutekunst isaac.gutekunst at vecna.com
Fri Mar 4 20:55:45 UTC 2016


I've tried using std::mutex, and it doesn't seem to be available in 
rtems 4.11. Is that true?

I have not tried 4.12 yet.


Isaac

On 03/02/2016 01:07 AM, Sebastian Huber wrote:
>
>
> On 01/03/16 22:32, Isaac Gutekunst wrote:
>>
>>
>> On 02/25/2016 03:37 AM, Sebastian Huber wrote:
>>> For proper C++11 support you need RTEMS 4.12. Below "fully supported"
>>> means as far as GCC/libstdc++ supports it.
>>>
>>> On 24/02/16 18:52, Isaac Gutekunst wrote:
>>>> Hi All,
>>>>
>>>>
>>>> Which C++11 features are supported by RTEMS?
>>>> Is there a table listing supported features, and limitations? Sorry if
>>>> I just missed it.
>>>>
>>>> I'm specially interested in
>>>>
>>>> * std::atomic
>>>
>>> Fully supported on targets that don't need libatomic.
>>>
>>>> * std::thread
>>>> * std::queue
>>>> * std::mutex
>>>> * std::condition_variable
>>>
>>> Fully supported.
>>>
>>>> * std::chrono
>>>
>>> Partially supported. For the details run the libstdc++ test suite and
>>> check the libstdc++ configure script which doesn't detect all RTEMS
>>> features (we have to fix several configure checks for this).
>>
>> I've never looked at gcc sources, so I'm probably doing something
>> stupid..
>>
>> I ran RSB with --no-clean, and looked into the build directory for
>> libstdc++, which for me is:
>>
>> rtems-source-builder/rtems/build/arm-rtems4.12-gcc-6-20160228-newlib-2.3.0.20160226-x86_64-linux-gnu-1/gcc-6-20160228/libstdc++-v3
>>
>>
>>
>> I can't seem to find any evidence of RTEMS. I was particular
>> interested in seeming something in config/os
>>
>>
>> Could you point me in the right direction? I'd like to know what
>> things work. I'd be happy to make some fixes if it doesn't seem too hard.
>
> Working with the GCC sources and build needs a bit of experience. The
> libstdc++ is customized via the output of the configure script. Some
> tests use a link-time test, which fails in case of RTEMS since several
> functions are declared in Newlib header files and defined in RTEMS, thus
> not link-time available during GCC build.
>
> I think the main problem is that gcc_no_link=no
> (config/no-executables.m4) since
>
> int
> main ()
> {
>
>    ;
>    return 0;
> }
>
> links. This leads GCC_TRY_COMPILE_OR_LINK() to use link-time configure
> tests.
>
That seems a bit beyond my skills for the time being.




More information about the users mailing list