"Bela Lubkin" <filbo@armory.com> wrote in message
news:200703261557.aa22645@deepthought.armory.com.. .
> John DuBois wrote:
>
>> In article <5ywMh.417$f56.2@bgtnsc05-news.ops.worldnet.att.net>,
>> Richard <aapex@att.net> wrote:
>> >"Steve M. Fabac, Jr." <smfabac@att.net> wrote in message
>> >news:46014B54.6FF1A9A@att.net...
>> >> Richard,
>> >>
>> >> What fix have you developed to overcome this problem?
>> >>
>> >> As a guess, I would add an explicit line in the called shell script
>> >> that reads the TZ value from /etc/TIMEZONE to overwrite the FoxBASE
>> >> exported value.
>> >
>> >Steve,
>> >
>> >Yes, that's exactly what I did. I cannot find any indication in the
>> >documentation of why it would change the TZ format.
>>
>> The exec system call converts the TZ value to XENIX format
>> when exec'ing a XENIX binary.
>
> Is that documented anywhere? (both the fact that it does; and what is
> meant by "XENIX format" of a TZ value vs. OSR5's "modern" ancient
> format?)
>
> I vaguely remember when this was implemented...
>
> Anyway, shouldn't the exec() system call from a XENIX binary, when
> invoking a UNIX binary, mangle $TZ in the opposite direction?
>
>>Bela<

One would think so, but it doesn't. Displaying the TZ variable within the
called script shows the "XENIX" format, and running just the "date" binary
confirms the behavior. This, of course, was never a problem when TZ was in
the simple "CST5CDT" format. Software geriatrics certainly presents some
interesting problems.

--
Richard Seeder