innreport's ConvDate

[Russ Allbery]

Alexander Bartolich <alexander.bartolich at gmx.at> writes:

> The date parsing code in innreport checks neither leap years nor
> daylight saving.  Incorrect handling of 29th of February loses the
> report of a whole day.  This must be fixed and is fortunately easy to
> fix.  However, I'm not so sure about consequences and cure of the second
> issue.  Probably the 25th hour can be recognized by checking
> non-sequential time stamps.

Thanks, filed as ticket #90.

For daylight savings time, yeah, we'd probably have to do something like
that, and that's a mess.  The best solution would be to switch to a
timestamp format that either included time zone information or always used
UTC, but that's of course a big change.

> Anyway, for documentation purposes I post a benchmark comparing
> innreport's current implementation (ConvDate), my version using a hash
> (ConvHash), and Perl's Time::Local::timelocal.

Perl's version is painfully slow since it stats /etc/localtime on each
conversion.  Looking at it in an strace is kind of frightening.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>

    Please send questions to the list rather than mailing me directly.
     <http://www.eyrie.org/~eagle/faqs/questions.html> explains why.