On 05/17/2018 11:06 PM, Joseph Myers wrote:
> Does this use of 64-bit time_t fix bug 19738 ("__tzfile_default 32-bit
> time_t overflow")?  If it does, we should add a testcase for that bug then
> resolve it as FIXED in 2.28.

Looks like it.  Here's a test case which fails before on i386 (assuming 
that the system posixrules file as the required contents) and passes 
afterwards.

I would call the bug fixed because dates where the overflow occurs with 
a 64-bit (internal) time_t are after the expected life-time of the sun, 
where daylight savings time is certainly meaningless.

Thanks,
Florian