Y2038: add function __difftime64aaribaud/y2038-rfc-3-wip

Note:

1. The implementation expects __time64_t arguments, and could,
   in theory, require 64 bits to express the difference accurately;
   but it returns a double, which only provides about 55 significant
   bits.

   We could make it return a long double, which would be more than
   enough for 64 bits accuracy. But then, existing source code which
   uses difftime, and therefore stores difftime results in doubles,
   would need to change to long doubles. However, we want 64-bit time
   support to work without any application source code change.

   Besides, 55 bits allow for counting seconds accurately over 417
   billion years, which is arguably enough for most actual uses of
   difftime.

2. The 64-bit time implementation was obtained by duplicating the
   original 32-bit code then simplifying the source code based on
   the knowledge that __time64_t is a 64-bit signed integer for
   all architectures. This led to the following removals:
   - in the subtract64 function, removal of code which handled
     unsigned time handling (as __time64_t is signed);
   - in the difftime64 function, removal of code which handled
     time bitsize smaller than or equal to that of a double matissa
     (as __time64_t has more bits than a double mantissa can hold)

3. The 32-bit implementation is now a wrapper around the 64-bit
   one.

0 files changed, 0 insertions, 0 deletions