introduce mp_set_sint and mp_set_uint functions with precise types

[minad]

This is a bit of a compromise now.

• The API is biased toward 32 bit (mp_get_sint, mp_get_uint, ...).
• 64 bit variants are provided (mp_get_sint64)
• Unsigned getters return two complement for negative numbers
• There are an unsigned getters for the magnitudes only (mp_get_mag, mp_get_mag64) - these are just like the current mp_get_long mp_get_long_long
• There are aliases for long and unsigned long (mp_get_ulong, mp_get_slong, ...) since long is problematic if porting between 32/64 bit platforms. Furthermore since the current API uses long, this would make things easier. But maybe these should also go to not again introduce legacy cruft.
• mp_set and mp_init_set are NOT deprecated in order to not penalize programs on platforms with int < 32 bit. This is still the most efficient function.

Why are these changes needed?

• signed setters/getters are a new addition
• unsigned setters return two complement such that round tripping and casting works seemless with c integers
• the new setters never fail, the current setters return mp_err
• the new functions use precise types (the api is robust under platform changes)

/**** NEW ****/
/* get integer, set integer and init with integer (int32_t) */
int32_t mp_get_sint(const mp_int *a) MP_WUR;
void mp_set_sint(mp_int *a, int32_t b);
mp_err mp_init_sint(mp_int *a, int32_t b) MP_WUR;

/**** NEW ****/
/* get integer, set integer and init with integer (int64_t) */
int64_t mp_get_sint64(const mp_int *a) MP_WUR;
void mp_set_sint64(mp_int *a, int64_t b);

/**** CHANGED API: More precise replacements for mp_set_int, mp_set_long, mp_get_int, two complement ****/
/* get integer, set integer and init with integer, behaves like two complement for negative numbers (uint32_t) */
#define mp_get_uint(a) ((uint32_t)mp_get_sint(a))
void mp_set_uint(mp_int *a, uint32_t b);
mp_err mp_init_uint(mp_int *a, uint32_t b) MP_WUR;

/**** CHANGED API: More precise replacements for mp_set_int, mp_set_long, mp_get_int, two complement ****/
/* get integer, set integer and init with integer, behaves like two complement for negative numbers (uint64_t) */
#define mp_get_uint64(a) ((uint64_t)mp_get_sint64(a))
void mp_set_uint64(mp_int *a, uint64_t b);

/***** CHANGED API: More precise replacement for mp_get_long/long_long ****/
/* get magnitude */
uint32_t mp_get_mag(const mp_int *a) MP_WUR;
uint64_t mp_get_mag64(const mp_int *a) MP_WUR;

/**** COMPATIBILITY API, needed? ****/
/* get integer, set integer (long) */
#define mp_get_slong(a)        (sizeof (long) == 8 ? (long)mp_get_sint64(a) : (long)mp_get_sint(a))
#define mp_set_slong(a, b)     (sizeof (long) == 8 ? mp_set_sint64((a), (b)) : mp_set_sint((a), (int32_t)(b)))

/* get integer, set integer (unsigned long) */
#define mp_get_ulong(a)       (sizeof (long) == 8 ? (unsigned long)mp_get_uint64(a) : (unsigned long)mp_get_ulong(a))
#define mp_get_magl(a)     (sizeof (long) == 8 ? (unsigned long)mp_get_mag64(a) : (unsigned long)mp_get_mag(a))
#define mp_set_ulong(a, b)    (sizeof (long) == 8 ? mp_set_uint64((a), (b)) : mp_set_uint((a), (uint32_t)(b)))

/* set to single unsigned digit, only 8 bit guaranteed */
void mp_set(mp_int *a, mp_digit b);
mp_err mp_init_set(mp_int *a, mp_digit b) MP_WUR;

/* get integer, set integer and init with integer (deprecated) */
MP_DEPRECATED(mp_get_mag/mp_get_uint) unsigned long mp_get_int(const mp_int *a) MP_WUR;
MP_DEPRECATED(mp_get_magl/mp_get_ulong) unsigned long mp_get_long(const mp_int *a) MP_WUR;
MP_DEPRECATED(mp_get_mag64/mp_get_uint64) unsigned long long mp_get_long_long(const mp_int *a) MP_WUR;
MP_DEPRECATED(mp_set_uint) mp_err mp_set_int(mp_int *a, unsigned long b);
MP_DEPRECATED(mp_set_ulong) mp_err mp_set_long(mp_int *a, unsigned long b);
MP_DEPRECATED(mp_set_uint64) mp_err mp_set_long_long(mp_int *a, unsigned long long b);
MP_DEPRECATED(mp_init_uint) mp_err mp_init_set_int(mp_int *a, unsigned long b) MP_WUR;

[czurnieden]

You did know that I would insist on documentation, didn't you? ;-)
But otherwise: yepp!

[minad]

Tests are also missing. Maybe I have time later today to fix those things.

EDIT: Btw. I changed the names again to be more in line with what tommath looks now. @czurnieden are you ok with it?

[minad]

@czurnieden Updated docs + tests

[minad]

@czurnieden Ok now?

[czurnieden]

Ok now?

Yepp.
At least I couldn't find more ;-)

[minad]

@sjaeckel There you go

[minad]

@sjaeckel I made another proposal, _l and _ul suffix for the long macros.

Maybe also use i32/u32 then for shorter names?

[sjaeckel]

that looks very good now! I like it

[minad]

Thanks

[minad]

If you have time to merge other stuff, I consider #273, #288, #290 and #291 as ready