Re: Problem with ftview on DEC Alpha (traced to t2gload.c)

[David Turner]

Hi Tom,

> 
> Hi,
> 
> Yes, I think that changing the various header/source files to use
> FT_Int32 (FT_UInt32) for 32 bit quantitites, FT_Int16... for 16 bit
> quantities would *definitely* ease library maintenence.  One could
> then use autoconf magic on UNIX-like boxes to get the correct types.
> For instance, on the DEC alphas I work with, FT_Int32 would be
> "typedef signed int FT_Int32;", provided one has a freetypeh.in (or
> some such) for autoconf to run sed on...
> 
You don't need autoconf to detect the size of integers. The ANSI header
file <limits.h> is sufficient, as proven by the code below taken from
<freetype/config/ftconfig.h>. Fortunately, we only target ANSI compilers,
and this works on all platforms :-)

When it comes to compiling the library, autoconf is mainly useful to
determine wether an "mmap" implementation is present on Unix, and provide
the libtool/install stuff (well, at least in theory, it's still not
completed in "freetype2/config/unix").

Note that I  want to separate the demo programs from the main FreeType 2
source hierarchy. They could also use autoconf/automake more deeply in
order to compile the executables (detect X11, etc..).. It'd be nice if we
could do this quickly, because I want to make the last beta next week
(with the auto-hinter !!, if everything goes well..)

Cheers,

- David

----------------------------------------------------------------------
/* We use <limits.h> values to know the sizes of the types.  */
#include <limits.h>

/* The number of bytes in an `int' type.  */
#if   UINT_MAX == 0xFFFFFFFF
#define FT_SIZEOF_INT  4
#elif UINT_MAX == 0xFFFF
#define FT_SIZEOF_INT  2
#elif UINT_MAX > 0xFFFFFFFF && UINT_MAX == 0xFFFFFFFFFFFFFFFF
#define FT_SIZEOF_INT  8
#else
#error "Unsupported number of bytes in `int' type!"
#endif

/* The number of bytes in a `long' type.  */
#if   ULONG_MAX == 0xFFFFFFFF
#define FT_SIZEOF_LONG  4
#elif ULONG_MAX > 0xFFFFFFFF && ULONG_MAX == 0xFFFFFFFFFFFFFFFF
#define FT_SIZEOF_LONG  8
#else
#error "Unsupported number of bytes in `long' type!"
#endif

... later ...

  typedef signed short    FT_Int16;
  typedef unsigned short  FT_UInt16;

#if FT_SIZEOF_INT == 4

  typedef signed int      FT_Int32;
  typedef unsigned int    FT_UInt32;

#elif FT_SIZEOF_LONG == 4

  typedef signed long     FT_Int32;
  typedef unsigned long   FT_UInt32;

#else
#error "no 32bit type found - please check your configuration files"
#endif



> >
> > Other than that, I don't think we really need to make FT_Pos a 32-bit
> > quantity, just take care of what we're doing in the library. But
> > maybe this would ease the engine's maintenance ?
> >
> 
> Regards,
> 
> Tom