[Scummvm-devel] Build of ScummVM 0.8.2 on Solaris 10 x86

[Jonathan Gray]

On Wed, Feb 08, 2006 at 02:35:51PM +0100, Laurent Blume wrote:
> Max Horn wrote:
> >Actually, the problem in this particular case is that the external header
> >file  mpeg2dec/mpeg2.h seems to assume that int8_t etc. is defined, when
> >it really should include the standard header inttypes.h. I consider this a
> >bug in that header / lib. The Solaris "int_types.h" header is odd, BTW, I
> >wonder if that is based on an older version of the recent C(++) spec :-).
> 
> They say it should be standard. If it's not, I can try to open a case 
> with them, so it gets better in the future:
> 
>  * This file, <sys/int_types.h>, is part of the Sun Microsystems 
> implementation
>  * of <inttypes.h> defined in the ISO C standard, ISO/IEC 9899:1999
>  * Programming language - C.
> 
> >Anyway, our workaround for that "bug" in mpeg2dec is flawed. The best
> >immediate workaround, IMO, would be to replace those typedefs in
> >animation.h by #define's.
> >
> >On the long run, we may want to consider moving from our uint8, int32 etc.
> >types to the new standard compliant types defined in inttypes.h. Of
> >course, we'd still check for the presence of that header, and if it is not
> >there, we'd still use our own typedefs. But I think that when possible,
> >it's usually better to use the existing standards as much as possible.
> >Alas, this is not a quick short term solution!

Actually you are supposed to include stdint.h not inttypes.h for it,
inttypes.h gets you other things like printf format strings macros and the like
on top of it which are uneeded.

> 
> I asked a friend of mine who is a developer about it, and I now 
> understand what's happening.
> 
> The int16_t, int32_t redefinitions are not a problem, because they are 
> reusing exactly the same declaration as is done already on Solaris.
> 
> For int8_t, however, there is a case in int_types.h:
> 
> #if defined(_CHAR_IS_SIGNED)
> typedef char                    int8_t;
> #else
> #if defined(__STDC__)
> typedef signed char             int8_t;
> #endif
> #endif
> 
> And since _CHAR_IS_SIGNED is defined in isa_defs.h, it's always there by 
> default.
> 
> So what I see as an immediate solution would be to have configure put 
> that in config.h:
> 
> #if defined(_CHAR_IS_SIGNED)
> typedef char        int8;
> #else
> typedef signed char int8;
> #endif
> 
> Would it be ok in the short term?

No, it should try compile something with the types and include
stdint.h