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

[C de-Avillez]

address@hidden:~/Downloads$ gcc -o inv-year inv-year.c
address@hidden:~/Downloads$ ./inv-year
time() = 1555368087
localtime() = 2019-04-15 17:41:27
             (mday=15 wday=1, isdst=1)
struct tm (after adjustment) = 0009-04-15 17:41:27
                              (mday=15 wday=1, isdst=1)
inv-year: mktime() failed: Value too large for defined data type
1 address@hidden:~/Downloads$

On Mon, Apr 15, 2019 at 3:55 PM Assaf Gordon <address@hidden> wrote:
>
> Thanks Bernhard,
>
> On 2019-04-15 2:14 p.m., Bernhard Voelker wrote:
> > I can easily reproduce here on my regular openSUSE:Tumbleweed from latest 
> > git:
> >
> >    $ src/date --debug '+%-Y' -d '- 2010 years'
> [....]
> >    date: error: adding relative date resulted in an invalid date: '(Y-M-D) 
> > 0009-04-15 22:10:37'
>
> This makes it easy to pinpoint (hooray for "--debug" :) ).
>
> This error is given if gnulib's "mktime_z" fails
> to convert the adjusted "struct tm" to "time_t"
> (adjusted because its tm_year was decremented by 2010).
>
> https://opengrok.housegordon.com/source/xref/gnulib/lib/parse-datetime.y#2177
>
> To see if this is glibc issue, or perhaps an gnulib/mktime_z wrapper
> issue, can you (and/or others) try the attached C program?
>
> It calls time(2)+localtime(3)+mktime(3) to emulate the date adjustment.
>
> Because the adjustment is to year 9 (about 1961 years before epoch),
> the time_t value is negative. perhaps that's the issue? or perhaps
> combined with a specific timezone it becomes problematic?
>
> On my system it gives:
> ----
>    $ gcc -o inv-year inv-year.c
>
>    $ ./inv-year
>    time() = 1555361050
>    localtime() = 2019-04-15 14:44:10
>                  (mday=15 wday=1, isdst=1)
>    struct tm (after adjustment) = 0009-04-15 14:44:10
>                                   (mday=15 wday=1, isdst=1)
>    mktime() after date adjustment = -61874070118
> ----
>
>
> regards,
>   - assaf
>
>
>
>
>
>


-- 
..hggdh..