Lukasz Majewski wrote:
> Paul, will you find any time soon to provide next version of mktime
> compatibility patch?

As far as mktime compatibility goes, I think I'd rather first cause mktime to be 
compatible with __time64_t, and worry about _TIME_BITS later. Proposed mktime 
patch attached; it's similar to my previous proposal except it leaves the 
__time64_t definition in posix/bits/types.h rather than moving it to 
time/mktime-internal.h.

This patch exposes __time64_t to user code if _LIBC is defined (because glibc 
needs __time64_t internally), or if __TIMESIZE != 64 (because of the compromise 
to expose __time64_t only on 32-bit time_t platforms). Once we add _TIME_BITS I 
expect that we should change the visibility rule to something that also involves 
_TIME_BITS, because __time64_t should not be exposed to user code if _TIME_BITS 
== 64. However, that can wait for the patches involving _TIME_BITS.