[tz] proposed changes for Win32 and a improved mktime() algorithm

[enh]

On Tue, May 9, 2017 at 9:48 PM, Paul Eggert <eggert at cs.ucla.edu> wrote:
> enh wrote:
>>
>> we wouldn't use LGPL code in Android's libc.
>
>
> What does Android use for mktime? Something slow? (If so, it's in good
> company. :-)

afaik we're the only people shipping a mostly unmolested copy of the
tzcode implementation :-)

(the only behavioral difference i can remember is:

  // http://b/31339449: POSIX says localtime(3) acts as if it called
tzset(3), but upstream
  // and glibc both think it's okay for localtime_r(3) to not do so
(presumably because of
  // the "not required to set tzname" clause). It's unclear that POSIX
actually intended this,
  // the BSDs disagree with glibc, and it's confusing to developers to
have localtime_r(3)
  // behave differently than other time zone-sensitive functions in <time.h>.

which i think i mentioned on the list at the beginning of the year.)