Re: [swbz 29035] mktime vs non-DST

[Carlos O'Donell <carlos@redhat.com>]

On 8/17/22 23:16, Paul Eggert wrote:
> On 8/17/22 19:37, Carlos O'Donell wrote:
>> The mktime() interface has been around for a long time and users do
>> not expect and have noticed the semantic changes we are seeing here
>> today.
> 
> The semantic changes that bother users boil down to the following:
> glibc mktime returns -1 when users specify impossible is_dst flags,
> whereas mktime used to succeed. That problem is fixed in Gnulib.

Right.

> The only disagreement between Gnulib and old (2.28) glibc in this
> area, is in relatively obscure cases where we have not seen any bug
> reports, and in these cases old glibc mostly gets it wrong (for
> example, see my previous email), whereas Gnulib mostly gets it right.
> Let's stick with the Gnulib behavior here.
 
What I'd like to see here as an outcome is a broader battery of tests in
glibc that cover the mktime() behaviour we want to support. We'll work on that
and review the recommended gnulib changes.

We have also seen a somewhat odd use of mktime() to detect DST transitions that
DJ and I need to audit and verify against the changes you have implemented in
gnulib.

Thanks for the quick comments. I think the next step is with us to do more
testing.

-- 
Cheers,
Carlos.