Re: [Stlport-devel] WCHAR_MAX and USHRT_MAX handling

[<fra...@fr...>]

There are already several issues of 5.1.2 fixed in STLPORT_5_1 branch. I 
am waiting for your README.evc8 file to release 5.1.3 :-)

It looks like you meet the same issue as the one reported here:
https://sourceforge.net/forum/forum.php?thread_id=1690171&forum_id=490891

    This fatal error only occur when wchar_t is a built-in type, 
surprising no!

    Yes I think this test is done to check that WCHAR_MAX is testable, I 
agree that comparison with USHRT_MAX is surprising, we should perhaps 
have a test for each macro.

    AFAIK, conclusion of discussion about WCHAR_MAX was that STLport 
should try to fix native definition. It is already done for DMC for 
instance. Maybe we should always redefine those macro to be sure.

    I also agree that static assertion is weird, I really do not 
remember how I finally come to this check. It is going to be difficult 
to check how it behaves on an Alpha platform as I was using one of the 
Sourceforge compiler farm host for that and this service has been removed...

Bests


Ulrich Eckhardt wrote:
> Greetings!
>
> I'm currently trying to get 5.1.2 to compile using VC8 and for MS Windows CE5. 
> I'm getting an error when compiling the unittests in file cwchar.cpp:
>
>   cwchar.cpp(11): fatal error C1017: invalid integer constant expression
>
> The problem there is the use of WCHAR_MAX, which is defined but not to a 
> numeric macro suitable for comparisons. Instead, it is defined 
> to '((wchar_t)-1)'. I'm wondering how this ever worked before?
>
> Now, the file where the error occurs is basically undocumented. Is it really 
> the intention to check that the WCHAR_MAX and WCHAR_MIN are defined and 
> defined properly? If neither are defined, this file compiles okay, so that is 
> obviously not the point. Then, why does it compare anything with USHRT_MAX? 
> If this is just supposed to make sure it can be used in comparisons, the 
> value compared with could be arbitrary. I also remember some discussions in 
> the forum or the mailinglist concerning the correct definition of WCHAR_MAX 
> but what was the outcome? Was it that the value should be suitable for 
> preprocessor comparisons? And shouldn't a test for those values exist anyways 
> somewhere?
>
>
> Lastly, why is this in stl/_cwchar.h:
>
>   _STLP_STATIC_ASSERT(((wchar_t)-1 > 0) && (WCHAR_MIN < 0))
>
> The comment above says that wchar_t for GCC/Alpha is unsigned, so how can the 
> minimal value be less than zero?
>
> I hope you can shed some light on the issue.
>
> cheers
>
> Uli
>
>