Request for review: keychain opt-in mechanism

[Ville Skyttä]

On Fri, 2005-08-05 at 03:02 +0200, Alexander Dalloz wrote:
> Am Do, den 04.08.2005 schrieb Ville Skyttä um 23:00:
>
> > The bad interaction with X login I mentioned earlier: I can no longer
> > reproduce it, things seem to work.  Still, two ssh-agents are started on
> > initial X login, but it might be tricky to avoid that, and probably
> > shouldn't be even attempted in the keychain package.  The culprit
> > appears to be "exec -l $SHELL -c "$SSH_AGENT ...""
> > in /etc/X11/xdm/Xsession, where the -l triggers loading of keychain
> > along with the shell profile, and $SSH_AGENT is then launched by the
> > shell later.  xinitrc-common doesn't set $SSH_AGENT if one is already
> > running, but that check is done too early wrt. keychain in the X
> > login/startup sequence.
> 
> Does that cause any harm on your system?

No, not any more.  I have no clue what's changed; certainly it's nothing
in the keychain package.  Probably some local configuration thing or a
plain human error.

> 1 ssh-agent for 1 root login, 2 ssh-agent processes as I am logged in as
> "adalloz" graphically and 1 ssh-agent for my "backup" user running
> automated backups through cron. Though "adalloz" has 2 ssh-agents I do
> not perceive any problem, while I share the opinion it could be prettier
> ;)

Yeah, it'd be cool if Xsession wouldn't start another one.  But as said
above, that's probably quite hard to avoid, and it doesn't seem to hurt,
so it's not a blocker.