| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
In the event that there is more than one session that matches
a particular tty when the VT changes we use the following rule:
1. take the oldest session where display-device == active VT device
2. take the oldest session where x11-display-device == active VT device
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
In certain cases it is necessary to distinguish the display-device
(tty of session leader) from the x11-display-device (tty of xorg
process).
|
|
Use an async helper job to collect session info for the
OpenSession() method.
|
|
So in Fedora's Rawhide /bin/login is now using libck-connector via the
PAM module and CK will track such sessions. However, it doesn't work
when root logs in and I think that's a bug. Attached patch should fix it
(is already in Rawhide). Thanks.
David
|
|
These tools will be used to generate and verify the parameters
used to open a session.
|
|
|
|
Basically this is a mini login. It authenticates, opens a session,
sleeps 20 seconds, and then closes the session.
|
|
|
|
Add a simple test app for the connector. Adapted from
the example David put in the header.
|
|
I modified the API of the new connector a bit.
* include errors
* make names a bit more consistent with what they do
* have symmetry between open and close
* use refcounting
* do parameter checking on inputs
I also made some changes to whitespace and style. And converted
descriptions to gtkdoc style comments.
|
|
Here are some patches to add one library and one PAM module so other
programs (apart from gdm) can easily interact with ConsoleKit. It's
mainly going to be used, I think, by login(1) and xinit(1). I'm not
super happy about doing a PAM module but the Fedora util-linux package
maintainer seems to be insist on this instead of just patching login(1).
The PAM module isn't built by default. Both pieces of code are licensed
under the MIT license as god knows what might want to use them. Sounds
good to you?
|
|
System dependent tools will be built in subdirectories under tools.
|
|
Restructure things a bit so that we can be completely async
about creating sessions. We want to do this so we can use
callouts to gather session info or to verify info. Also
make the OpenSession create a parameter list and then create
the session rather than creating the session and changing
attributes. This is less racy and the session will be complete
and correct before any signals are emitted.
|
|
|
|
|
|
|
|
Also fix a distcheck error.
|
|
Make the proc.h API more coherent. Convert some warnings
to debug statements. Fix two leaks. Add more error checking.
|
|
Copy the iso8601 function from glib to support versions less than 2.12.
|
|
|
|
|
|
The system idle hint is TRUE when there are no busy (ie non-idle) sessions.
So the system is idle when there are no sessions or all sessions are idle.
|
|
This makes it easier to use from C code. Also fix a bug where
we weren't emitting the argument in the idle-hint-changed signal.
|
|
Rename GetIdle to GetIdleHint (etc) to make it clear that these
are not authoritative properties.
|
|
Try to limit the use of errno in case it can change.
|
|
Don't spew warnings when getfd() can't find a suitable console fd. Also
add a comment to note where getfd.c comes from.
|
|
Change to using an async queue for events from VT watching threads. Add
mutex protection to shared data. Compress the event queue when possible by
discarding all but the most recent VT activation event.
|
|
It was getting impossible to read all on one line.
|
|
Added GetCreationTime and GetIdleSince methods. And added the
output to the list-sessions tool.
|
|
Add the GetIdle and SetIdle methods, and the IdleChanged signal.
SetIdle is restricted to the uid of the session in question.
|
|
Only allow the uid and pid that opened the session to close it.
|
|
|
|
|
|
|
|
While try to play with ConsoleKit i realize it cannot create its pid file, so
if needed attached patch adds that with copy/paste from PolicyKit :).
|
|
|
|
|
|
|
|
Just the use vt number given by the vt-manager instead of
asking for the current vt again. A little bit less racy this way.
|
|
The minimum stack size 16k seems to crash glibc
when we do printf in debug.
|
|
Instead of exiting on disconnect this attempts to reconnect.
Based on patch from: Timo Hoenig <thoenig@suse.de>
|
