Re: "date", "uptime" core dumps on huge time stamps (+ "stat" fix)

[Jim Meyering]

Paul Eggert <address@hidden> wrote:
> I just notice that coreutils sometimes calls 'localtime' without
> checking that the result is not NULL, which can lead to core dumps on
> hosts with 64-bit time_t and 32-bit int (e.g., 64-bit Solaris).  Here
> are some fixes.
>
> The "stat" patch doesn't fix a core dump on any host that I know
> about, but it's possible on far-future hosts and anyway the fix is
> easy.  The stat.c patch also changes the behavior of stat to be nicer
> if stat is given time stamps that localtime can't convert.
>
> 2004-02-02  Paul Eggert  <address@hidden>
>
>       Don't dump core if localtime returns NULL (possible on
>       hosts with 64-bit time_t and 32-bit int).
>       * src/date.c: Include "inttostr.h".
>       (batch_convert, main):
>       If time conversion fails, exit with nonzero status.
>       (show_date): Return int to report conversion failure.
>       Print the time as an int if localtime fails.
>       * src/uptime.c: Print "??" if the current clock can't
>       be converted by localtime.  This won't happen until the year
>       2*31 + 1900, but we don't want to dump core even if the current
>       clock has the wrong value.
>
>       * src/stat.c: Include "inttostr.h".
>       (human_time): Print the date/time as a number of seconds since the
>       epoch if it can't be converted by localtime.  This is better than
>       just saying "invalid", and is consistent with what "ls" does.
>       Don't dump core if the year has more than 48 digits; this isn't
>       possible on any contemporary host, but we might as well do it right.

Thank you!
I've merged those changes.