On 09/11/15 19:20, Pádraig Brady wrote: > On 09/11/15 16:58, Pádraig Brady wrote: >> On 09/11/15 16:48, Jim Meyering wrote: >>> On Mon, Nov 9, 2015 at 8:07 AM, Pádraig Brady <p...@draigbrady.com> wrote: >>>> On 09/11/15 16:02, Bernhard Voelker wrote: >>>>> On 11/09/2015 04:27 PM, Pádraig Brady wrote: >>>>>> I see on most GNU/Linux distros that kill(1) is >>>>>> provided by the shell or util-linux. >>>>>> Should we just remove it from coreutils? >>>>> >>>>> What about non-Linux systems, i.e., where util-linux is not >>>>> available? I personally don't have such a system, but I think >>>>> GNU software should not forget about such platforms. >>>> >>>> Absolutely. Though in this case the shells cater >>>> for kill(1), or the platform already provides it. >>> >>> Note that POSIX requires an 'exec'able kill program. >>> I.e., the shell-provided one is insufficient when you want to >>> invoke it via a program like xargs. >> >> OK I'll assume for now that platforms >> might not have a separate kill program, >> and move it to disabled_by_default_progs.
On second thoughts we might continue to build kill, since there seems to be quite a bit of divergence on how -l is handled at least: GNU> /bin/kill -l HUP 1 solaris> /bin/kill -l HUP 1 freebsd> /bin/kill -l HUP error, usage is... util-linux> /bin/kill -l HUP HUP debian-procps> /bin/kill -l HUP HUP INT QUIT .. debian-procps> /bin/kill -l 1 HUP INT QUIT .. The procps handling of `kill -l 1` seems to contravene POSIX. I also notice that GNU kill is used with kfreebsd on debian. For these reasons I might just keep the test cleanups from my previous patch but continue to build kill by default? cheers, Pádraig
