Re: X11 behavior since Tiger and "BadAtom" datapoint
Re: X11 behavior since Tiger and "BadAtom" datapoint
- Subject: Re: X11 behavior since Tiger and "BadAtom" datapoint
- From: Mike Zuhl <email@hidden>
- Date: Wed, 14 Jun 2006 14:35:26 -0700
Rich Cook wrote:
I assume you are running ssh from your Mac to the 4.2 boxen?
To start the session, yes, but then I log out. Tunneling X protocol
thru ssh is what causes the problems, so I connect directly to the
TCP port of the Mac.
Have you tried not tunneling the X connection? I.e., ***as an
experiment*** -- not permanent, try setting your DISPLAY to yourmac:
0 and using xhost to allow the 4.2 box.
Yup, this is exactly how I run stuff. Since I'm on a secure LAN, I
run "xhost +" on X11
to turn off authentication, then I login via ssh to the various
servers and start
the sessions with environment variable "DISPLAY=mymac:0.0". (BTW,
you have to run
"defaults write com.apple.x11 nolisten_tcp -boolean false" to get X11
to talk TCP.)
So no ssh tunneling and it works fine.
To reiterate:
1) Using ssh tunneling gets "BadAtom" on FreeBSD 4.2.
2) Direct TCP connection w/o ssh in the loop works fine.
If the error goes away then it's SSH to blame.
Yep. Pretty weird! You would think that if ssh had a problem it
would show up elsewhere,
too.
Other datapoints:
3) The "BadAtom" happens on commands to the window manager, e.g,.
"xwininfo"
- Also on TclTk, which uses X11 server atoms for IPC
- Regular screen traffic seems unaffected
(i.e., "xwininfo" would be a good tool for testing this bug)
4) FreeBSD 5.3 likes ssh tunneling just fine.
5) Sun users have reported a similar problem. There is a decent
chance that
their version of ssh is somewhat old, like FreeBSD 4.2
6) Pre-Tiger, FreeBSD 4.2 liked ssh tunneling just fine, too.
So it seems that there was a change, perhaps inadvertent, to ssh that
was not entirely
compatible with older ssh servers. New ssh clients talk to new
servers just fine,
old ssh clients talked to old servers just fine, but new clients
talking to old servers
cause X protocol screw-ups.
Since Mac OS, FreeBSD, and probably Sun all use ssh from the same
open-source code
base, I think this is an issue with that code base and has nothing to
do with Apple.
If so, it should show up elsewhere, e.g., old Sun app talking to a
fresh Linux
console. Anyone have the setup to test this assumption?
Have you tried the -Y option to ssh?
Nope. My read is that's just authorization which isn't the issue
here. Also, as soon
as I got my tools to work, I got on to paying work :-).
--Mike Zuhl
On Jun 13, 2006, at 3:44 PM, Mike Zuhl wrote:
I two FreeBSD servers I use: one running FreeBSD 4.2 and one
running FreeBSD 5.3. I also connect to a farm of various Linux
machines running recent releases. Since Tiger I've been getting
the "BadAtom" error on the FreeBSD 4.2 X clients connecting over
ssh. These happen when the client tries to get info from the
window manager using the xwininfo command or running a TclTk app,
which I presumes talks to the window manager for its (somewhat
unusual) IPC.
Neither FreeBSD 5.3 nor the recent copies of Linux suffer the same
problem. And, strangely, neither does FreeBSD 4.2 when I turn on
TCP sockets in X11 and connect directly instead of using ssh.
I suspect that the problem is not X11 but in differences in
Tiger's ssh and how it operates with older versions of ssh, both
in Sun's and FreeBSD.
--Mike Zuhl
_______________________________________________
Do not post admin requests to the list. They will be ignored.
X11-users mailing list (email@hidden)
This email sent to email@hidden