[PATCH] Fix multiple minor tzset glitches [BZ #24004]

[Florian Weimer]

* Carlos O'Donell:

> On 4/12/19 2:41 PM, Paul Eggert wrote:
>> On 4/12/19 10:07 AM, Carlos O'Donell wrote:
>>>> One possible compromise would be for localtime to poll, and for
>>>> localtime_r to not poll...
>>>   If users have to make code changes it's better to just add a new
>>> api
>>> for "Check for update?" and "Do the update!".
>>
>> If we add that API, will we also change localtime() to stop checking for
>> and doing updates in the (undocumented) cases where localtime checks for
>> and does them now? And if not, how will programs arrange for localtime
>> to be faster? This would seem to be the key question.

> Sure! I would be on board with changing localtime() to stop checking
> for and doing updates (unless someone has a convincing argument not to
> do this).

I want to make localtime the canonical interface, using thread-local
storage.

Thanks,
Florian