Re: [PATCH]: Fix for several getdate.y issues (amended patch #3)

[Ondřej Vašík]

Hello,
Jim Meyering wrote:
> I'm a little leery of this patch, because it makes interpretation
> of UTC+dd and UTC-dd dependent on the value of dd.  For dd <= 14, dd
> represents hours.  Otherwise, it represents minutes.  That would have
> to be documented, but seems odd enough that my reflex is to avoid the
> change so I don't have to document something so strange.

You are right, it has to be documented or changed. Opengroup
specification for offset in ENVVAR TZ states that first two numbers
should be considered as hours. Therefore one possible solution could be
to replace my change for dd <= 14 with dd<=99 (or other way to separate
two digit specification) and documenting it - it would be easier to
understand and hopefully will not break anything. Current situation
causes behaviour that commonly used UTC+5 format is misinterpreted as
UTC+0:05. 
I used the dd <= 14 approach to allow TZ UTC+00:20 to be written UTC
+0020 (or UTC+20)... there are no TZ in use between UTC and UTC+/-1 now,
but there used to be several in the past. None of them was shifted by
less than 20 minutes(+0:20, -0:25, +0:30 , -0:44 were used, latest
stopped in 1972). Anyway POSIX specification for TZ ENVVAR could be good
excuse to not support fully that unlikely TZ offset - UTC-0:44 will
still work.
Where should I document current the changes? In gnulib getdate.texi
only? Or in both gnulib and coreutils documentation?

Greetings,
         Ondrej Vasik

signature.asc
Description: Toto je digitálně podepsaná část zprávy