bug#35289: closed (Re: bug#35289: date+%-Y -d "- N years" errors when N > 111)

[Assaf Gordon]

Hello,

On 2019-04-15 5:10 p.m., O. Emmerson wrote:
> For me it gives:
>
> $ ./inv-year
> time() = 1555369320
> localtime() = 2019-04-16 00:02:00
>                                (mday=16 wday=2, isdst=1)
> struct tm (after adjustment) = 0009-04-16 00:02:00
>                                (mday=16 wday=2, isdst=1)
> inv-year: mktime() failed: Value too large for defined data type
>

On 2019-04-15 6:50 p.m., C de-Avillez wrote:
[...]
> address@hidden:~# gcc -o inv-year inv-year.c
> address@hidden:~# ./inv-year
> time() = 1555375408
> localtime() = 2019-04-16 00:43:28
>                (mday=16 wday=2, isdst=0)
> struct tm (after adjustment) = 0009-04-16 00:43:28
>                                 (mday=16 wday=2, isdst=0)
> mktime() after date adjustment = -61874061392
>
> So: a pristine 19.04 runs it. My laptop (which is my work machine,
> full of other packages & programs), does not.

Thank you both for testing.

So, to summarize:
whenever "inv-year" fails - it is a problem with glibc on your
setup, *not* a problem in coreutils' date(1) program.

If there is a setup where "inv-year" succeeds but date(1) still fails,
then it is a problem in coreutils.

I'm glad to hear latest Ubuntu 19.04 is working fine
(though the reason for the earlier failure is still a mystery).

As Paul suggested, trying 'strace' on the failing system
might reveal more details.

regards,
 - assaf