complaint from svlogd

[David Collier]

OK guys, I take the consensus, which is that if I'm doing

( fred |svlogd )

then any kind of kill, including SIGKILL, on fred should send svlogd an
EOF and it should shut down pronto.

I was convinced that my fred dropped out of the ps listing, but the
svlogd hung around for tens of seconds, but finger-trouble is easy to
cause when you're flapping about trying to understand a problem.

I will try to construct a test-case and see if it works as I'm told it
should, or does anything odd.

Thanks for all the help and patience.

D

In article <F4AC465B29B61A4FA792A4E6FEA8A2020164E596 at wamxm01.ciena.com>,
jcathey at ciena.com (Cathey, Jim) wrote:

> *From:* "Cathey, Jim" <jcathey at ciena.com>
> *To:* "Harald Becker" <ralda at gmx.de>, 
> <from_busybox_maillist at dexdyne.com>
> *CC:* <busybox at busybox.net>
> *Date:* Wed, 16 Feb 2011 08:48:39 -0800
> 
> >Anyone else may hold the writing
> >end of the pipe in an open state ... which shouldn't be the 
> intention
> on
> >a simple pipe like:
> >
> >"message_producer_process  | message_logger_process"
> 
> BTDT.
> 
> Yes, use fuser (or rummage in /proc/X/fd) to find out
> who's still got the writer side of the pipe open once
> the producer is dead and gone.  It's real easy to leak
> pipe fd's if you're not careful, but the shell (creater
> of all |'s) shouldn't have such a weakness in it.
> 
> -- Jim
> 
>