10.02.2016, 22:05, "Tim Blechmann" <[email protected]>:
>>>   we have seen funny toolchain bugs, with 10.10 sdk and 10.8 deployment
>>>   target on 10.8, where std::exceptions could not be caught ... compiling
>>>   against 10.8 sdk solved that issue.
>>
>>   Regardless of whether that is a valid approach or not, the fact that people
>>   are doing that poses a problem for us. If we write code against the latest
>>   toolchain, it may not compile against older ones people might be using.
>
>  this implies bumping the minimum requirements. for some use cases this
>  might be acceptable, for others it isn't.
>
>>   That can be because we used new features that aren't present in older ones,
>>   even if properly runtime checked.
>>
>>   Another problem is that the compilers in the old toolchains are older, with
>>   older libc++ and with bugs that we aren't testing for.
>
>  ... or newer toolchains can introduce regressions ...
>
>>   If we're going to upgrade our Xcode in all our builds, I would advise we 
>> also
>>   make it mandatory for everyone else too. Check during configure against the
>>   minimum SDK version, that being the *current* at the time of release.
>
>  so what happens when you are hit by a toolchain regression? i'm not
>  voting for having code compiling with gcc-4.2 or any of the llvm-gcc
>  flavors which were around during that time. however qt is a framework
>  and it should support more than one toolchain so that people who are hit
>  by toolchain bugs still have an alternative ...


I can imagine the next solution: additionally run CI using open source Clang, 
corresponding to minimum Xcode version you are going to support.


-- 
Regards,
Konstantin
_______________________________________________
Development mailing list
[email protected]
http://lists.qt-project.org/mailman/listinfo/development

Reply via email to