Re: [Patch / Request For Testers] libstdc++/34032

[Paolo Carlini <pcarlini at suse dot de>]

I have another comment...
> Eric Niebler wrote:
>   
>> I can confirm that that patch fixes the problem with <iostream> and
>> ::isblank. Of course, it means that some standard C++0x functions are
>> not available under -std=c++0x even though they are implemented on
>> cygwin, but I don't think there is much gcc can do about that. Someone
>> needs to fix cygwin's C headers.
>>     
> Thanks Eric. I agree with your analysis. Note that there is more about
> it: as I'm going to write in a comment in acinclude.m4, such targets
> define snprintf and co as extensions for a reason: those implementations
> are pre-standard, not fully-conforming C99 faciltities. snprintf is a
> common example: pre-standard the return value was different, didn't tell
> the user the required size in case the buffer was exceeded.
>   

Something we could do, as an enhancement, we could make the autoconf
tests for C99 facilities more fine grained, and thus enable selectively
the C99 facilities actually available. That would certainly make sense
but note that there are pitfalls: if we want to avoid AC_TRY_RUN tests,
individual tests can be fooled rather easily (consider again the example
of snprintf individually: without running a test, to my best knowledge
it's very hard to tell a real C99 implementation from the pre-standard
one...)

Paolo.