Re: [PATCH v3] Y2038: make __mktime_internal compatibles with __time64_t

[Paul Eggert <eggert at cs dot ucla dot edu>]

On 10/24/18 4:23 AM, Albert ARIBAUD wrote:
> In the case where neither _LIBC nor NEED_MKTIME_WORKING are defined,
> "another" mktime() is called. Do we assume this other mktime will set
> errno to EOVERFLOW, or do we do it too in that case?

I would, yes. If the other mktime is broken Gnulib can deal with it in a 
different way; you don't need to worry about it.

> Also, I assume tests would be needed for both cases where
> __mktime_internal() returns -1? If so, I would need struct tm values
> which trigger either one, or at least indications on how to find ones.

You want outlandish values, where (for example) tm_year is INT_MIN or 
INT_MAX. On typical 64-bit platforms there are no outlandish values, as 
EOVERFLOW isn't possible when int is 32 bits and time_t is 64 bits.