> I've found the following clever ssh-agent stanza in Richard Crowley's
> `.profile`[https://raw.github.com/rcrowley/home/master/.profile]:
>
> which ssh-agent >/dev/null && {
> : ${SSH_AUTH_SOCK:=$(echo /tmp/ssh-*/agent.* | cut -d" " -f1)}
> [ -S "$SSH_AUTH_SOCK" ] && {
> export SSH_AUTH_SOCK
> } || {
> eval $(ssh-agent)
> ssh-add
> }
> }
>
> Wondering why it doesn't work properly in OpenBSD ksh as I am not so
> good with POSIX sh.
Can't really tell which part doesn't work in OpenBSD ksh (I got used to
mksh and it got a lot of bashism lately). Anyway, I personally would
rather write it like this (not tested, I personally use
~/.ssh-agent-info to store ssh-agent's socket and PID):
if which ssh-agent >/dev/null
then
test -z "$SSH_AUTH_SOCK" && \
SSH_AUTH_SOCK=`echo /tmp/ssh-*/agent.* | cut -d\ -f1`
if test -S "$SSH_AUTH_SOCK"
then
export SSH_AUTH_SOCK
else
eval `ssh-agent`
ssh-add
fi
fi
Hmm, just glanced over OpenBSD's manual for ksh and it seems all
constructs used are supported by OpenBSD's ksh - so the original code
should work there too... strange... or is TMPDIR set to something other
then /tmp? Does ssh-agent place its socket into /tmp/ssh-*/agent.*? (The
manual only states the socket lives in $TMPDIR/ssh-*/agent._ppid_.)
--
Eckehard Berns
