#306 Remove 64 to 32 bit conversions

[Fredrick Meunier]

libspectrum has lots of integer conversions between integer types that are commonly 32 and 64 bit on 64 bit platforms, e.g. libspectrum_bzip2_inflate( const libspectrum_byte bzptr, size_t bzlength, libspectrum_byte outptr, size_t outlength ) has unsigned int length2 which is assigned length2 = *outlength. On 64 bit Mac OS X size_t is unsigned long which is 64 bit while length2 is unsigned int.

These conversions should be located and removed.

[Fredrick Meunier]

The proper approach here is not completely obvious - Spectrum file formats are by definition 32 bit as they embed 32 bit length block and file lengths. However libspectrum uses size_t for a lot of the values that end up being read/written to those files and in memory block compression/decompression functions that would fail if more than 4GB was required for any block/file lengths.

My inclination is to formally make libspectrum a 32 bit library for these purposes and not worry about the limits that will impose - any dissenting thoughts out there?

[Sergio Baldoví]

On 64 bit Mac OS X size_t is unsigned long which is 64 bit while length2 is unsigned int.

Is not 'unsigned int' 64 bits long? (pun not intended). Building with standard C99 might mitigate these conversions.

My inclination is to formally make libspectrum a 32 bit library for these purposes and not worry about the limits that will impose

Sounds reasonable.

[Fredrick Meunier]

Is not 'unsigned int' 64 bits long? (pun not intended). Building with standard C99 might mitigate these conversions.

OS X uses the LP64 model for 64 bit programming, so int is 32 bit while longs and pointers are 64. C89/C99 will both act the same way.

[Sergio Baldoví]

I messed up the whole thing. Interesting reading.