Date: Thu, 25 Aug 94 11:49:51 CDT
From: "Cave Newt" <roe2 AT midway DOT uchicago DOT edu>
To: dj AT ctron DOT com, djgpp AT sun DOT soe DOT clarkson DOT edu, storner AT olicom DOT dk
Subject: Re: 1.12.maint1, second beta

Henrik Storner writes:

> DJ writes:
> 
> > As for timezones, all works properly if you use the minimal setting -
> > no DST fields, like this:
> > 
> > 	set TZ=est5
> 
> Sorry, but - no. stat() is still broken. I used the following bit of
> code to try this out:

And, in addition, "using the minimal setting" is unacceptable.  A TZ
setting of the form "EST5EDT" is *correct*, and I can't require my
users to change a perfectly valid setting because of bugs in djgpp's
libraries.  (The long form may be even more correct, but I'm 95%
certain it will break older programs which were compiled before Posix
got popular.)  I'm not even asking that the "EDT" part be honored cor-
rectly; as my initial bug report noted, emx ignores that, too.  What
it *doesn't* do (but djgpp does) is ignore the "5" when the EDT is 
present.  And, of course, emx's time/ftime values are consistent with 
its stat() values regardless of what the TZ setting is (or isn't).
DJ:  try testing in some other timezones besides Eastern.  The current 
code has that hardwired in at least two places.

Btw, thanks to Henrik for helping to debug all of this.  I grabbed
the libc sources and started poking around, but I got stuck when 
turbo_assist went off into an unknown (private?) int 21h function.
Compiling with any version of Turbo C isn't an option for me, but I
did check the MSC license and didn't find any funny no-competition
clauses, FWIW.

Greg Roelofs