Linux-Setup Digest #441, Volume #19 Mon, 21 Aug 00 06:13:07 EDT
Contents:
Re: How to run WM on remote display? (Colin Watson)
Re: Fdisk or Cfdisk (Villy Kruse)
HELP: WVDIAL OK, but no browsers work (McManus Leo Root DSP Consultant)
The best Linux distribution for Server? ("Opps!")
adding new driver to kernel - not module (Zebee Johnstone)
Re: The best Linux distribution for Server? (Nicolas Iselin)
Internet Gateway with Coldera OpenLinux 2.2 (Simon)
Re: Netscape Bus error (Juergen Sauer)
linux and DTC16RX from HP (eric)
Re: netatalk printing multiple copies (David Martin)
----------------------------------------------------------------------------
From: [EMAIL PROTECTED] (Colin Watson)
Crossposted-To: comp.os.linux.x
Subject: Re: How to run WM on remote display?
Date: 21 Aug 2000 07:08:13 GMT
Jon Davis <[EMAIL PROTECTED]> wrote:
[Deleted comp.os.linux.help, as it doesn't exist except on some
misconfigured news servers, and rearranged the quoting below.]
>> In article <hf%n5.121356$[EMAIL PROTECTED]>, "Jon
>> Davis" <[EMAIL PROTECTED]> wrote:
>> > I've exported the DISPLAY property to the IP of my Windows 2000
>> > computer but when I run "startx" even from a telnet session it
>> > loads up the WM on the Linux computer's display.
[...]
>I'm assuming that's off-topic; the point is when I start Gnome remotely
>using "gmc" it fails to work within Exceed (console error message is
>"** WARNING **: Root window clicks will not work as no GNOME-compliant
>window manager could be found!") although something does show up in
>Exceed with a few icons and a file manager window but no desktop or
>anything I get in "startx", and if I start it remotely with "startx" it
>boots fine but it uses the local display instead of the remote display.
OK, this makes sense from what you say above. Firstly, they're called
environment variables, not properties :) Secondly, startx expects to be
the program that sets $DISPLAY, rather than reading from it, as this
makes sense for most cases (it means you can run startx within one X
session and have it start another, for instance). If you want to tell
startx to use a different display, say the first display on IP address
192.168.1.1, you should use:
startx -- 192.168.1.1:0
See 'man startx' and 'man xinit' for more information.
Note, though, that 'startx' normally starts up an X server itself, and
may get somewhat confused if it can't. You may find it more convenient
to set the $DISPLAY variable as before and then run ~/.xinitrc (or
~/.xsession, which is the equivalent used by display managers like xdm,
gdm, and wdm). If whichever file you run ends with something like:
window-manager &
exec gnome-session
(where window-manager is something like sawfish, fvwm, or
enlightenment), then a fairly GNOMEish environment should appear on your
eXceed display, assuming there's no window manager running there
already.
Hope that helps,
--
Colin Watson [[EMAIL PROTECTED]]
"Everyone, please welcome our new friend Stef. He's here with us
because he thinks he's a penguin." - http://www.userfriendly.org/
------------------------------
From: [EMAIL PROTECTED] (Villy Kruse)
Subject: Re: Fdisk or Cfdisk
Date: 21 Aug 2000 07:54:52 GMT
On 18 Aug 2000 23:42:05 GMT, Colin Watson <[EMAIL PROTECTED]> wrote:
>Dennis L <[EMAIL PROTECTED]> wrote:
>>Where do I get the Linux version? I have two distributions of Redhat
>>6.1 and neither has this file.
>
>If it's anything like Debian, they'll both be in the util-linux package.
>I'd *hope* that was installed by default ...
>
Same on Redhat.
Villy
------------------------------
From: McManus Leo Root DSP Consultant <[EMAIL PROTECTED]>
Subject: HELP: WVDIAL OK, but no browsers work
Date: Mon, 21 Aug 2000 09:56:55 +0200
Reply-To: [EMAIL PROTECTED]
I have my WVDIAL connecting to my ISP OK, but Netscape or Lynx do not tx
out to the modem. Any ideas? I have my ISP primary address.
Thanks
Leo
------------------------------
From: "Opps!" <[EMAIL PROTECTED]>
Subject: The best Linux distribution for Server?
Date: Mon, 21 Aug 2000 16:02:16 +0800
I am going to build a Web server for my school..there will be 1000 users
a/c with their own homepage,web mail (and access to MySQL, use of Server
scripting like PHP,PERL ),I am woundering which Linux Dist. can handle
this job the best, or I should use FreeBSD or something?
------------------------------
From: [EMAIL PROTECTED] (Zebee Johnstone)
Subject: adding new driver to kernel - not module
Date: 21 Aug 2000 05:19:01 GMT
Reply-To: [EMAIL PROTECTED]
I've got a driver for a satellite card
that is currently working as a module.
I want to compile it into the kernel, not load it as
a module.
There's nothing obvious about how to do this. There are various
makefiles about but they all intertwine.
I have a working .o file, can I just tell it to include that
into the kernel somehow?
If I have to recompile the driver code, where do I add the
makefile that the driver needs to compile?
Zebee
------------------------------
From: Nicolas Iselin <[EMAIL PROTECTED]>
Subject: Re: The best Linux distribution for Server?
Date: Mon, 21 Aug 2000 10:36:22 +0200
Opps! wrote:
>
> I am going to build a Web server for my school..there will be 1000 users
> a/c with their own homepage,web mail (and access to MySQL, use of Server
> scripting like PHP,PERL ),I am woundering which Linux Dist. can handle
> this job the best, or I should use FreeBSD or something?
Take Debian. It might be a bit harder to get used to it, but once you
have read the few essential readme's and man-pages (especially apt-get),
it is much easier to install and remove packages than on other systems.
And, BTW, Debian has just stepped to version 2.2, is now using a 2.2.x
kernel etc...
Nicolas
------------------------------
From: Simon <[EMAIL PROTECTED]>
Subject: Internet Gateway with Coldera OpenLinux 2.2
Date: Mon, 21 Aug 2000 10:12:05 +0100
I'm trying to setup Internet connection sharing with OpenLinux 2.2
Ipchains 1.3.8-1 is installed.
I've followed Caldera's Doc. Ref: 991112-0009 for setting up a
gateway....
file ..../etc/rc.d/rc.local
modprob i_masq_ftp
modprob i_masq_raudio
modprob i_masq_vdolive
ipchains -P forward DENY
ipchains -A forward -j MASQ -s 192.168.1.0/24 -d 0.0.0.0/0 # (my
network is 192.168.1.*)
file.../etc/sysconfig/network
IPFORWARDING=yes
file..../etc/hosts.allow
ALL:192.168.1.11
ALL:192.168.1.12
ALL:192.168.1.13
ALL:192.168.1.14
I've set up the windows NT machine with Default gateway 192.168.1.1 -
the linux box ....
what am I doing wrong ???
------------------------------
From: Juergen Sauer <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.questions
Subject: Re: Netscape Bus error
Date: 21 Aug 2000 06:12:16 GMT
Don Thompson <[EMAIL PROTECTED]> schrieb
am Sat, 12 Aug 2000 15:16:25 GMT in comp.os.linux.questions:
DT> hi,
DT> Since I've upgraded to SuSE 6.4, I can't get netscape to run except when
DT> logged in as root. I have tried setting a user with all the same groups
DT> as root, and get the same results: clicking the icon does nothing,
DT> typing netscape from a terminal window responds with Bus error.
DT> What the scoop?
Perhaps no problem with netscape itself, perhaps it's the
crappy multimedia plugin "plugger", which crashes and this
crash ist fatal for the netscape main also.
You may check this, if there exists a /tmp/ndebug file, which is
a junk file (debug) from plugger. If you use netscape from/with
more than 1 user the ndebug file will crash the plugger with
"permission denied".
Try "rpm -e plugger". Then it *will* work.
There is a plugger update with "disabled" debugging feature.
mfG
Jojo
--
- Professionelle Linux Server, Professioneller Support und Dienstleistungen
- AutomatiX GmbH - Vollautomatische Kransteuerungen & SAP f�higes Lagerger�t
- J�rgen Sauer Neue Str. 11 28790 Schwanewede mailto:[EMAIL PROTECTED]
- +49-4209-4699 +49-172-5466499 FAX +49 4209 4644 http://www.automatix.de
- Hinweis: Nach �28 Abs.3 Bundesdatenschutzgesetz WIDERSPRECHE
- ich der Nutzung meiner Daten fuer Werbezwecke!
------------------------------
From: eric <[EMAIL PROTECTED]>
Subject: linux and DTC16RX from HP
Date: 21 Aug 2000 09:45:09 GMT
Has somebody already use DTC16RX from HP on a Linux box? I try to install such
DTC on RedHat version
I don not known what to install on these machine because there is a manager from
HP ux only .
======
User of http://www.foorum.com/. The best tools for usenet searching.
------------------------------
From: David Martin <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.powerpc
Subject: Re: netatalk printing multiple copies
Date: Mon, 21 Aug 2000 12:58:52 +0300
Reply-To: [EMAIL PROTECTED]
==============EF671B8B597AD797501AC5CA
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
"Sheldon D. Stokes" wrote:
> I'm currently running an x86 based linux box and a macintosh at home.
> I've got an older laserwriter (4/600) which only talks appletalk. I've
> added a localtalk to ethernet bridge. I am trying to set up my x86
> linux box (running mandrake 7.1) to print to the laserwriter. I've got
> it to print, but I'm having a weird problem where it prints anything I
> send it 5 times. It's somehow related to a permission problem I think.
> I've got world read and write permissions on the spool directory. The
> lpr.log file looks like this:
>
> Aug 19 16:37:41 peecee ifpap[874]: starting for stokes
> Aug 19 16:37:41 peecee ifpap[874]: accounting with psa[875]
> Aug 19 16:37:41 peecee ifpap[874]: sending to pap[876]
> Aug 19 16:37:41 peecee ifpap[874]: PostScript
> Aug 19 16:38:19 peecee ifpap[874]: 876 done
> Aug 19 16:38:19 peecee ifpap[874]: 875 died with 2
> Aug 19 16:38:19 peecee lpd[859]: lp: job could not be printed
> (cfA028peecee)
>
> It says it can't be printed, but it DOES print, and prints again and
> again.
>
> And the atalk log file (specified in /etc/printcap) looks like this:
>
> status: Permission denied
> status: Permission denied
> *4255
> status: Permission denied
> status: Permission denied
> *4256
> status: Permission denied
> status: Permission denied
> *4257
> status: Permission denied
> status: Permission denied
> *4258
> status: Permission denied
> status: Permission denied
> *4259
> sendmail: warning: sendmail is set-uid root, or is run from a set-uid
> root process
> sendmail: fatal: My hostname peecee is not a fully qualified name - set
> myhostname or mydomain in /etc/postfix/main.cf
> status: Permission denied
> status: Permission denied
> *4260
>
> It prints fine, but it appears that there is a problem with it removing
> the printed file. Why it stops after 5 tries in that case is beyond my
> understanding.
>
> Any help would really be appreciated.
>
> Sheldon
This looks like a problem I have been experiencing under both YDL1.2 on
PPC and Redhat 6.2 - nobody on any of the groups or lists I frequent had
apparently heard of it. After a lot of playing around I have isolated the
problem (last week) and have a work around for it, but it is not
straightforward to fix and it is not related to permissions but the line
printer accounting mechanism.
The basic problem is an incompatibility between the Redhat 6 derived
version of the old BSD lpd server that most recent distributions seem to
be shipping and the netatalk postscript filter psf. What is happening is
that lpd is passing additional, undocumented arguments (against the
original BSD specs) to filters that use the "if" printcap mechanism (like
netatalk psf), which is officially for filters that do print accounting.
psf expects any of the RFC 1179 job control characters as arguments from
its caller (lpd in this case) - which is actually overkill for the BSD lpd
spec but I think has been put in for Unix System V based printing systems
that vendors like Sun use. If it gets an extra argument that is not in
its list of expected ones it assumed that it is an accounting filename.
That is why you see psa being forked (psa is the netatalk accounting
package). psa is expected to get a page number count for the postscript
job and write it in to the accounting file for that printer. psa should
do this after the job has printed. The problem is that psa is called with
nothing to do and can terminate almost the instant it is forked by psf.
psf is designed to wait until psa is finished before it returns its
completion status to lpd (that is the "died with 2" message) - the psa
process has already disappeared from the process list. This causes psf to
return an abnormal completion code to lpd. lpd thinks there was and error
printing the job and starts again. And so lpd just keeps printing and
printing until the job is removed from the job queue (or not with your
second problem). The offending lpd argument is actually the job name. In
RFC 1179 is says that the job control character for the job name is 'J',
and that is what netatalk psf expects. The version of lpd I have (and I
suspect you also have) send the job name to psf as '-j jobname'. This
case mismatch causes the problem because psf is not expecting a lowercase
j - it does not know what to do with it, and its internal arithmetic for
counting arguments gets fooled into thinking that an accounting files name
has been passed (the accounting filename is expected to be passed without
a preceding argument identifier, unlike all of the others that psf can
accept).
The reason why some jobs print OK first try or print several times and
then finish normally appears to be simple luck. psa appears to get forked
at the same priority as its parent (psf). If psf gets going on the
process queue before psa then it can finish first, subsequently psa gets
some cpu time and finishes - the correct sequence for a normal completion
is made and everything works the way it should. Sometimes this can happen
first go, on the other hand it can go on ad infinitum - I had a job print
17 times on me before I realised what was happening and stopped it. It
seems to depend on system load and job size to some extent, but there are
no hard and fast rules.
My work-around involves hacking about the source code for psf to make it
handle the arguments that lpd is sending. This stops psf spuriously
forking the account process and setting up this race between psa and psf
that leads to endless printing. I can send you the diff for my patch to
psf if you want - you will need to get a netatalk1.4b2+asun2.1.3 source
distribution to compile it into. Another alternative that may work (I
have not got around to trying it) is to scrap using lpd and try either
lprng or cups as a printer server daemon. These may not send spurious
arguments to "if" filters that causes your problem. If you can find the
source for a readhat 5.x era lpd and get it to build, that will also cure
the problem - I had an older Hurricane based PPC port (Linux PPC R4.2)
that was fine. The only difficulty might be libc incompatibility
problems, but I am speculating because I have not tried it.
Chopping about psf is not really the ideal solution, after all it is the
line printer daemon that is really causing the problem, psf is actually
just doing its job. The main problem I have had is trying to work out who
actually maintains the lpd that everyone seems to be shipping - I still
don't know who looks after that code so I can try and get it sorted out
probably. I have rather idealistically posted a bug report with Redhat
about it - but I don't expect to hear from them. I understand that they
are scrapping lpd in favour of lprng in their next release, which may or
may not eliminate the problem. And because they are scrapping it there
does not seem to be a lot of incentive on their behalf to work out a fix
now.
I hope all of this helps illuminate at least the first part of your
problem.
--
David C Martin ([EMAIL PROTECTED])
Materials Engineering Laboratory,
Mechanical Engineering Department,
University of Oulu
PO Box 4200
Oulun Yliopisto 90014 FINLAND
==============EF671B8B597AD797501AC5CA
Content-Type: text/html; charset=us-ascii
Content-Transfer-Encoding: 7bit
<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<html>
"Sheldon D. Stokes" wrote:
<blockquote TYPE=CITE>I'm currently running an x86 based linux box and
a macintosh at home.
<br>I've got an older laserwriter (4/600) which only talks appletalk.
I've
<br>added a localtalk to ethernet bridge. I am trying to set up my
x86
<br>linux box (running mandrake 7.1) to print to the laserwriter.
I've got
<br>it to print, but I'm having a weird problem where it prints anything
I
<br>send it 5 times. It's somehow related to a permission problem
I think.
<br>I've got world read and write permissions on the spool directory.
The
<br>lpr.log file looks like this:
<p>Aug 19 16:37:41 peecee ifpap[874]: starting for stokes
<br>Aug 19 16:37:41 peecee ifpap[874]: accounting with psa[875]
<br>Aug 19 16:37:41 peecee ifpap[874]: sending to pap[876]
<br>Aug 19 16:37:41 peecee ifpap[874]: PostScript
<br>Aug 19 16:38:19 peecee ifpap[874]: 876 done
<br>Aug 19 16:38:19 peecee ifpap[874]: 875 died with 2
<br>Aug 19 16:38:19 peecee lpd[859]: lp: job could not be printed
<br>(cfA028peecee)
<p>It says it can't be printed, but it DOES print, and prints again and
<br>again.
<p>And the atalk log file (specified in /etc/printcap) looks like this:
<p>status: Permission denied
<br>status: Permission denied
<br>*4255
<br>status: Permission denied
<br>status: Permission denied
<br>*4256
<br>status: Permission denied
<br>status: Permission denied
<br>*4257
<br>status: Permission denied
<br>status: Permission denied
<br>*4258
<br>status: Permission denied
<br>status: Permission denied
<br>*4259
<br>sendmail: warning: sendmail is set-uid root, or is run from a set-uid
<br>root process
<br>sendmail: fatal: My hostname peecee is not a fully qualified name -
set
<br>myhostname or mydomain in /etc/postfix/main.cf
<br>status: Permission denied
<br>status: Permission denied
<br>*4260
<p>It prints fine, but it appears that there is a problem with it removing
<br>the printed file. Why it stops after 5 tries in that case is
beyond my
<br>understanding.
<p>Any help would really be appreciated.
<p>Sheldon</blockquote>
This looks like a problem I have been experiencing under both YDL1.2 on
PPC and Redhat 6.2 - nobody on any of the groups or lists I frequent had
apparently heard of it. After a lot of playing around I have isolated
the problem (last week) and have a work around for it, but it is not straightforward
to fix and it is not related to permissions but the line printer accounting
mechanism.
<p>The basic problem is an incompatibility between the Redhat 6 derived
version of the old BSD lpd server that most recent distributions seem to
be shipping and the netatalk postscript filter psf. What is happening
is that lpd is passing additional, undocumented arguments (against the
original BSD specs) to filters that use the "if" printcap mechanism (like
netatalk psf), which is officially for filters that do print accounting.
psf expects any of the RFC 1179 job control characters as arguments from
its caller (lpd in this case) - which is actually overkill for the BSD
lpd spec but I think has been put in for Unix System V based printing systems
that vendors like Sun use. If it gets an extra argument that is not
in its list of expected ones it assumed that it is an accounting
filename. That is why you see psa being forked (psa is the netatalk
accounting package). psa is expected to get a page number count for
the postscript job and write it in to the accounting file for that printer.
psa should do this after the job has printed. The problem is that
psa is called with nothing to do and can terminate almost the instant it
is forked by psf. psf is designed to wait until psa is finished before
it returns its completion status to lpd (that is the "died with 2" message)
- the psa process has already disappeared from the process list.
This causes psf to return an abnormal completion code to lpd. lpd
thinks there was and error printing the job and starts again. And
so lpd just keeps printing and printing until the job is removed from the
job queue (or not with your second problem). The offending lpd argument
is actually the job name. In RFC 1179 is says that the job control
character for the job name is 'J', and that is what netatalk psf expects.
The version of lpd I have (and I suspect you also have) send the job name
to psf as '-j jobname'. This case mismatch causes the problem because
psf is not expecting a lowercase j - it does not know what to do with it,
and its internal arithmetic for counting arguments gets fooled into thinking
that an accounting files name has been passed (the accounting filename
is expected to be passed without a preceding argument identifier, unlike
all of the others that psf can accept).
<p>The reason why some jobs print OK first try or print several times and
then finish normally appears to be simple luck. psa appears to get
forked at the same priority as its parent (psf). If psf gets going
on the process queue before psa then it can finish first, subsequently
psa gets some cpu time and finishes - the correct sequence
for a normal completion is made and everything works the way it should.
Sometimes this can happen first go, on the other hand it can go on <i>ad
infinitum</i> - I had a job print 17 times on me before I realised what
was happening and stopped it. It seems to depend on system load and
job size to some extent, but there are no hard and fast rules.
<p>My work-around involves hacking about the source code for psf to make
it handle the arguments that lpd is sending. This stops psf spuriously
forking the account process and setting up this race between psa and psf
that leads to endless printing. I can send you the diff for my patch
to psf if you want - you will need to get a netatalk1.4b2+asun2.1.3 source
distribution to compile it into. Another alternative that may work
(I have not got around to trying it) is to scrap using lpd and try either
lprng or cups as a printer server daemon. These <i>may</i> not send
spurious arguments to "if" filters that causes your problem. If you
can find the source for a readhat 5.x era lpd and get it to build, that
will also cure the problem - I had an older Hurricane based PPC port (Linux
PPC R4.2) that was fine. The only difficulty might be libc incompatibility
problems, but I am speculating because I have not tried it.
<p>Chopping about psf is not really the ideal solution, after all it is
the line printer daemon that is really causing the problem, psf is actually
just doing its job. The main problem I have had is trying to work
out who actually maintains the lpd that everyone seems to be shipping -
I still don't know who looks after that code so I can try and get it sorted
out probably. I have rather idealistically posted a bug report with
Redhat about it - but I don't expect to hear from them. I understand
that they are scrapping lpd in favour of lprng in their next release, which
may or may not eliminate the problem. And because they are scrapping
it there does not seem to be a lot of incentive on their behalf to work
out a fix now.
<p>I hope all of this helps illuminate at least the first part of your
problem.
<pre>--
David C Martin ([EMAIL PROTECTED])
Materials Engineering Laboratory,
Mechanical Engineering Department,
University of Oulu
PO Box 4200
Oulun Yliopisto 90014 FINLAND</pre>
</html>
==============EF671B8B597AD797501AC5CA==
------------------------------
** FOR YOUR REFERENCE **
The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:
Internet: [EMAIL PROTECTED]
You can send mail to the entire list (and comp.os.linux.setup) via:
Internet: [EMAIL PROTECTED]
Linux may be obtained via one of these FTP sites:
ftp.funet.fi pub/Linux
tsx-11.mit.edu pub/linux
sunsite.unc.edu pub/Linux
End of Linux-Setup Digest
******************************
