On Thu, 15.07.10 11:01, Tim Waugh (twaugh(a)redhat.com) wrote:
On Wed, 2010-07-14 at 21:41 -0400, Bill Nottingham wrote:
> Re: cups, if the entire point is to reserve the sockets early with
> systemd, why would portrelease still be required?
Also, re: this comment:
# This is evil stuff. CUPS should use proper enumeration instead of
# retriggering these devices. CUPS folks, please fix this, otherwise Kay will
# come after you!
ExecStartPost=/sbin/udevadm trigger --subsystem-match=usb --attr-match=bInterfaceClass=07
--attr-match=bInterfaceSubClass=01
ExecStartPost=/sbin/udevadm trigger --subsystem-match=usb
--property-match=DEVNAME=/dev/usb/lp*
Kay actually suggested doing this in the first place IIRC. We have udev
triggers for when printers are connected/disconnected, but the triggers
require cupsd to be running. The situation the above lines are to deal
with is that a new printer is connected at boot, so the (earlier) udev
trigger had to abort because cupsd wasn't running.
Nah. I talked to him just 1 minute ago.
The right approach here is to enumerate existing devices when CUPS
starts up. All programs that care about devices should do that:
1) subscribe to device events
2) enumerate what's already there
If everybody would retrigger the events for their devcies then the
number of events would be drastically increased.
Kay pointed me to this:
http://article.gmane.org/gmane.linux.hotplug.devel/13925
Which is his usual reply to issues like this.
Lennart
--
Lennart Poettering - Red Hat, Inc.