[solved] Strange posts in /var/log/xdm.log

Tony0945 · Posted: Tue Nov 10, 2020 2:07 am Post subject:

OK. But I get a black vt7, no xdm chooser.

GDH-gentoo · Posted: Tue Nov 10, 2020 3:10 am Post subject:

I get the greeter and the small xconsole window. Remind me what version of OpenRC is the one you have installed.

Tony0945 · Posted: Tue Nov 10, 2020 3:30 am Post subject:

GDH-gentoo · Posted: Tue Nov 10, 2020 4:56 pm Post subject:

/sbin/openrc-run is a symlink to openrc in that version, right?

It looks like 13.7 didn't support error_log yet, so the service script needs a little adjustment, but the service should have started nonetheless.

/etc/init.d/xdm-debug

Tony0945 · Posted: Tue Nov 10, 2020 5:39 pm Post subject:

Thanks, GDH-gentoo. I'll try it.

I tried to switch back from openbox to mate and found that XSession is still Openbox and indeed openbox is still running.

Tony0945 · Posted: Tue Nov 10, 2020 6:17 pm Post subject:

xdm-debug on boot: htps://dpaste.com/782JB4UW6
xdm-debug after starting browser: https://dpaste.com/4B626LQY8
Should I try a different level?

Anon-E-moose · Posted: Tue Nov 10, 2020 6:31 pm Post subject:

GDH-gentoo · Posted: Tue Nov 10, 2020 6:58 pm Post subject:

Tony0945 · Posted: Tue Nov 10, 2020 7:12 pm Post subject:

Tony0945 · Posted: Tue Nov 10, 2020 7:17 pm Post subject:

Tony0945 · Posted: Tue Nov 10, 2020 7:30 pm Post subject:

Presumably chooser is configured by Xaccess which is bone stock.

Going to try "pattern followed by hostnames"

Tony0945 · Posted: Tue Nov 10, 2020 9:30 pm Post subject:

n desperation I was going to emerge consolekit to see if that fixed things and found that sys-auth/consolekit no longer exist!

GDH-gentoo · Posted: Tue Nov 10, 2020 9:39 pm Post subject:

Tony0945 · Posted: Tue Nov 10, 2020 10:40 pm Post subject:

Anon-E-moose · Posted: Wed Nov 11, 2020 12:36 am Post subject:

/usr/lib64/X11/xdm/Xsession

This guy does a few things

sets up for .xsession-errors

check if $1 arg (kdeism)

has these set

Tony0945 · Posted: Wed Nov 11, 2020 12:54 am Post subject:

Anon-E-moose · Posted: Wed Nov 11, 2020 1:00 am Post subject:

He expects to see this

startup=$HOME/.xsession

and with it not existing he

if [ -x "$startup" ]; then
exec "$startup"
else
exec /bin/sh "$startup"

tries to "exec /bin/sh" on an non-existent file, not sure what the system does in cases like that.

At least that's how I read the script.

Edit to add:

Tony0945 · Posted: Wed Nov 11, 2020 1:41 am Post subject:

Anon-E-moose · Posted: Wed Nov 11, 2020 1:51 am Post subject:

Tony0945 · Posted: Wed Nov 11, 2020 3:37 am Post subject:

Tony0945 · Posted: Wed Nov 11, 2020 3:14 pm Post subject:

Anon-E-moose, I instrumented both /usr/lib64/X11/xdm/Xsession and /etc/X11/chooser.sh
When I start xdm with openrc-run, the first doesnt' even run. I think if /etc/X11/xdm/Xsession exists, it supersedes the one in /usr/lib.
Anyway, executing /usr/lib64/X11/xdm/Xsessio manually resuklts in

GDH-gentoo · Posted: Wed Nov 11, 2020 5:54 pm Post subject:

In my experience with XDM, not too many processes write to /var/log/xdm.log. One of the first things that Xsession does is redirect output to ~/.xsession-errors, so if the "-2" messages were printed by something that is being run by that script, I would expect them to appear in that file. On my own computer, with Gentoo's default /etc/X11/xdm/xdm-config, processes that write to xdm.log are:

The xdm process launched by the init system.
The X server spawned by xdm ("Server Started <PID>" in the debug output) as instructed by /etc/X11/xdm/Xservers, long-lived.
A child xdm process associated with the server's display, that displays the login screen and then spawns Xsession ("pid: <PID>" in the debug output), long-lived. The setproctitle() invocation makes it appear as "xdm: <DISPLAY>" in the output of ps (with any option that displays the "COMMAND" column).
For display :0, an xconsole process spawned by Xsetup_0 ("attempting to execve() /usr/lib64/X11/xdm/Xsetup_0" in the debug output), long-lived.
Xsession (attempting to execve() /usr/lib64/X11/xdm/Xsession" in the debug output), long-lived, but then it redirects its output.
An xrdb process for reading /etc/X11/xdm/Xresources ("attempting to execve() /usr/bin/xrdb" in the debug output), short-lived.
The startup and reset scripts specified in xdm-config ("attempting to execve() <script pathname>" in the debug output), all short-lived. These are /usr/lib64/X11/xdm/{GiveConsole,TakeConsole} for display :0, and /usr/lib64/X11/xdm/{Xstartup,Xreset} otherwise. None of them print any output (directly).
/usr/lib64/X11/xdm/chooser processes that display the "XDMCP Host Menu" window. I can't tell exactly when they are spawned, but these occurrences should be announced by a "RunChooser <display>" message in the debug output, and I don't see any messages like that in Tony0945's logs, or mine.

I'm very skeptic about the ConsoleKit patch having anything to do with the "-2" messages, because this:

Tony0945 · Posted: Wed Nov 11, 2020 6:20 pm Post subject:

That's a very good roundup, GDH-gentoo.

It seems like I can suppress them by running the browsers in an xterm with "palemoon 2>/dev/null" or "firefox-bin 2>/dev/null".
In the documentation, it says that child processes inherit the log file. openbox is a child of xdm, and xterm is a child of openbox and palemoon would be a child of xterm.

But why has it stopped on the Ryzen system and still going on the Phenom II? And why does the redirection mostly work?
The two setups are not exactly the same. On the Ryzen system, instead of openbox being called directly, a fake seesion "menu" is presented and openbox, mate, or fluxbox are chosen from menu.

Tony0945 · Posted: Sun Mar 28, 2021 7:25 pm Post subject:

SOLVED:

As to why it stopped on the Ryzen system and continues on the Phenom, the reason is that the ryzen system switched from nouveau to radeon.

The first clue cam when a portage update changed the -2 to a -2 with a message in front. The message referred to being unable to switch to some razy mode like 32x78 depth 8.

The final solution was to add parameters.