Re: [swbz 29035] mktime vs non-DST

[DJ Delorie <dj@redhat.com>]

Paul Eggert <eggert@cs.ucla.edu> writes:
> First, mktime is not "broken" if

Yeah, leftovers from the first iteration, and it just means that the
results don't match what the test expected.  We get to argue over what a
valid test is later ;-)

> illustrate why the differences between old glibc and Gnulib (a)
> largely don't matter,

Sadly, when someone has an application that's been expecting those
results for ages, and suddenly their code breaks, they don't care if the
new version is "better".  It's just "different".  You've given a few
examples of how mktime() might be abused, but we have no idea what kind
of abuses someone else's code might be relying on.

So while I agree that gnulib's code is more consistent, and thus better,
I'm still going to argue the other side until we've covered all the
options - for all those users who will still be affected by the "right"
change.