Re: Fwd: Re: trap / kill
Raphael Geissert <atomo64@gmail.com> writes:
> With Adam agreeing to make this private conversation public, please read
> below and comment. If necessary I don't mind a report to be filled
> against the package.
Please open a bug against debian-policy for this issue so that we can
track the discussion. I think we should at least talk about whether to
add signal numbers for kill/trap given how widely they're used in portable
code like Autoconf and Libtool.
Thanks!
> ---------- Forwarded Message ----------
>
> Subject: Re: trap / kill
> Date: Wednesday 26 March 2008
> From: "Adam D. Barratt" <adam@adam-barratt.org.uk>
> To: Raphael Geissert <atomo64+debian@gmail.com>
>
> On Wed, 2008-03-26 at 13:32 -0600, Raphael Geissert wrote:
>> On 25/03/2008, Adam D. Barratt <adam@adam-barratt.org.uk> wrote:
> [...]
>> > As you mentioned <87odap7ihc.fsf@windlord.stanford.edu>, if you follow
>> > the thread from that message you may notice that a) I replied to that
>> > mail at the time :) and b) Clint replied pointing out that posh
>> > *doesn't* actually support signal numbers... (with the exception of 0,
>> > which is required by POSIX and shouldn't be caught by the tests).
>> >
>> > There's also #436438, which led to Clint removing support for kill -9
>> > and trap 2 from posh.
>>
>> According to the last archive scan there are 179 packages affected out
> there.
>> I've strong doubts about how to solve the problem, so we either
>> request policy to require kill|trap -SIGN and kill|trap -[0-9] or we
>> annoy the maintainers of those 179 packages.
>
> I'm tempted to raise the issue with Russ (with his lintian and policy
> hats on) to gauge his opinions on requesting support for the XSI
> versions of kill and trap.
>
>> Btw, kill -s SIGNNAME pid does work in Debian's lenny bash.
>
> I should hope it does - that's required by POSIX. ;-)
>
> dash otoh does appear to happily support kill -9 and kill -KILL (tested
> in sid, but the package version is the same in lenny).
>
> dash(1) also features the use of a signal number in combination with
> trap, although it's SIGUSR1/30, which isn't even on the XSI list.
>
> [...]
>
> Cheers,
>
> Adam
>
> -------------------------------------------------------
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: