"Thomas G. Lockhart" <lockhart@alumni.caltech.edu> writes:
> ... How about if we continue to test for the other too
> and then the code can choose which one to enable. I can flip around the
> source blocks to get the right sense. I think this may help the Alpha
> port also, since I vaguely recall that Dec/Compaq (*sigh*) ran afoul of
> our tests too. And although the BSD-style TZ-in-tm seemed to be old
> fashioned, with reentrant/threaded code it is the wave of the future so
> could/should be the first choice for the code to consider.
I had some other fixes to make in configure.in today, so I went ahead
and added a call to AC_STRUCT_TIMEZONE. So the configuration symbol
HAVE_TM_ZONE is now available for use. You probably know better than
I what parts of the code have to be touched to prefer using the tm_zone
field of struct tm over the global timezone variable, so I'll leave that
part of the repair to you...
regards, tom lane