Re: Where does "dpkg -l" reads package info?

[David Z Maze]

"Yildiz, Murat" <[EMAIL PROTECTED]> writes:
> where does "dpkg -l" or deselect|select reads package info?
> /var/lib/dpkg/available or status? or both?
> I tried to delete these two files

Ow.  Don't do that; you've now caused dselect to forget that anything
at all is installed on your machine, including things like libc6 and
dpkg.  You should have -old versions of both files; dpkg will probably
work if you copy the -old versions back.

> and run apt-setup and let read all 8 cd's
> but it seems it doesn't write available and status files again.

The canonical way to do this is to run 'apt-get update', hit 'u' in
aptitude, or select "update" from dselect's main menu.  What problem
are you really having?

> What to do when these two files are corrupt? (woody 3.0 stable)

("Corrupt" as in bad blocks?  Open the file with a text editor, delete
the damaged regions, re-run an APT update, and pray.)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Safety of Upgrading Unstable

[David Z Maze]

Mark <[EMAIL PROTECTED]> writes:
> I installed unstable about a month ago and have had nothing but good
> times. ... But, I'm curious to know how safe/dangerous it is to just
> say 'apt-get upgrade' presently.

Presently?  Both of my unstable machines work fine (though there is
that observation that some packages have dependencies on outdated C++
libraries, notably aptitude).

> I've been following debian-devel and debian-user looking for
> problems people have had with upgrading unstable and haven't seen that
> many (a few regarding kde / libfam issues).

This is probably the best way to find out if something is broken in
unstable.  debian-devel-announce is also a good list to read (and is
far lower traffic).  And of course, if you find something broken, file
a bug report!  (http://bugs.debian.org/)

> However, when I use aptitude to show me what will be
> upgraded/removed/installed during the upgrade, it is removing some
> packages that I would think shouldn't be removed.

I've generally found it possible to figure out aptitude's reasoning
when it wants to delete something.  Often this will be because a newer
package conflicts with the older one, and you have to pick one or the
other to get a consistent state.  Try looking under "packages which
depend on " on the per-package pages to get some hints on this.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: DocBook

[David Z Maze]

Mike M <[EMAIL PROTECTED]> writes:

> I want to restart my DocBook efforts.  I have SGML source and want
> to produce HTML and PDF docs
>
> Should I start by installing these packages: docbook, docbook-dsssl?

That's probably a good start; you also need a DSSSL processor (jade
and jadetex seem to be ~canonical).  Then you can run

  jade -t tex foo.sgml   # foo.sgml + foo.dsl -> foo.tex
  jadetex foo# foo.tex -> foo.dvi
  dvips foo -o   # foo.dvi -> foo.ps
  pdfjadetex foo # foo.tex -> foo.pdf

Alternatively, if you're in an XML world, you can use an XSLT
processor (like xsltproc) and libfop-java to get PDF out.  But that
has Java dependencies, is strictly XML, and the Debian dependencies
never seemed quite right to me.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Unidentified subject!

[David Z Maze]

debian_newbie <[EMAIL PROTECTED]> writes:
> How do I change the number of times I can boot up before fsck does a
> complete file system check? It does it on my Woody machine every 20
> times.

Other people have already mentioned tune2fs(8).

> Also, I'm using ext3 filesystem. isn't it a journalized fs? I
> thought journalized filesystems didn't have to be fscked?

The important difference is that ext2fs filesystems must be fsck'd if
the system is shut down uncleanly, ext3fs can recover without an
fsck.  That doesn't mean that the system never needs to be fsck'd;
there's always the off chance of a software bug or hardware failure
that causes minor corruption that you'd want to notice occasionally.

The suggestion I've heard is that you set the maximum mount count on
each filesystem separately, to something moderate and relatively
prime.  My laptop's root and /home partitions are set to 34 and 38
mounts, in some order.  (Yeah, so those aren't relatively prime.)  If
I boot once a day, each filesystem gets checked on average about once
a month.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: networking problem and philosophy

[David Z Maze]

Bret Comstock Waldow <[EMAIL PROTECTED]> writes:

> I'm trying to get Debian going on my Thinkpad T21, and synchronize with
> my Sony Clie PDA.

(This works fine for me, but I always build my own kernel.  I've had
better luck using coldsync than pilot-link, and the first sync always
fails.)

> I'm using the kernel-package and make-kpkg tools as shown in the Debian
> manual, and that part seems to work.  I've got a kernel that may have
> what I need to sync my PDA.
>
> But it won't give me my network access.
>
> I've looked a bit at the 'netenv' stuff, and it seems to have created
> hard definitions of my network settings in /etc/netenv.  I chose during
> install to use DHCP when it asked - the netenv settings don't seem to
> reflect this.

Hmm.  Is there anything informative in /etc/network/interfaces (the
"normal" place for network settings under Debian)?  You also might try
running your DHCP client by hand, and seeing if that works.  Also,
check that your kernel configuration includes a driver for your
network card.  I think the "Socket Filtering" option (CONFIG_FILTER)
is also required to use DHCP, even though that's not obvious from
anything in the kernel configuration.

> Clues, suggestions, explanations would be very welcome.  Should I be
> accepting 'netenv' during the install?  How can I get rid of it and use
> DHCP to configure my network on startup?  Or...?

If you're always using DHCP and never a static address, netenv is
probably unnecessary.  There's the minor question of whether your
network card is PCMCIA or not, but the basic Debian tools deal fine
with the case where you have a single network card that always gets
brought up using an address from DHCP.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Fonts in XFree86/fvwm changed during update on sid

[David Z Maze]

Lukas Ruf <[EMAIL PROTECTED]> writes:
> last Friday, I run an update of debian sid.
>
> Since then, the "graphical" fonts like used in mozilla or jpilot and
> xclock changed.

FWIW, I've noticed something similar: my laptop, having been
relatively off the network for a week, spontaneously changed its
default font in Galeon from Palatino to Times, even though the
selected font in the preferences is still Palatino.  Similarly, the
default font in LogJam (the only GNOME 2 application I ever really
use) is different.  xterm is fine, though, and xfontsel seems to pull
up the right font without a problem.  Working hypothesis: something in
the xft world is borked, though I don't know what right off hand, and
I haven't dug much further.  Quick poking around in the BTS doesn't
find anything obvious, but this isn't something I'm too familiar with
the guts of either.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Dumb question: How do you reboot?

[David Z Maze]

Leo Spalteholz <[EMAIL PROTECTED]> writes:

> On March 4, 2003 09:01 pm, Eric G. Miller wrote:
>> On Tue, Mar 04, 2003 at 09:17:39PM -0600, Kent West wrote:
>> > I believe the next version of Gnome's login screen (gdm)
>> > implements a menu allowing you to shutdown/reboot.
>>
>> Hmm, doesn't the version in Woody have that cabability?  GDM has
>> been able to do that for quite some time (I think SystemMenu is set
>> "false" by default).
>
> How come that works anyway?  If you are not logged in as root, in fact 
> you're not logged in at all, then why are you allowed to shut down 
> the machine?

The gdm process runs as root (it has to to be able to change user ID's
to yours when you log in), so it also has permission to run shutdown.

> Is this not a security risk?  Couldn't anyone just ssh in and then
> reboot the machine without logging in or what am I missing here?

It's hard to be able to click on the gdm window on the local console
from a remote ssh session.  :-)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Some kernel compile questions

[David Z Maze]

Sukrit <[EMAIL PROTECTED]> writes:

> 1. How do i decide which modules to load at boot time, which file is
> to be edited?

/etc/modules

> (i am thinking that i'll compile support for lots of devices -
> cd-writer, different network cards - as modules that way i won't
> have to recompile kernel when i add those devices. So i don't want
> all my modules to be loaded at boot time. Also, i don't want to
> dynamically and automatically load/unload modules)

...I was going to suggest using the built-in module autoloading
support, which works well for me, but I guess not.  :-)

> 3. i compiled the kernel, now do i need to get kernel-headers? What is
> the function of kernel-headers?

I'm not entirely clear.  In unstable, there is a kernel-build-*
package, which provides everything you need to build kernel modules
against the kernel in question; the kernel-headers-* package alone
frequently isn't enough.  But if you compiled the kernel, then you
already have kernel source, which means you already have a headers
tree without doing anything special.

> 4. Right now i am copying kernel bzImage into /boot renaming it, and
> editing lilo. Also for modules i 
> #make modules
> #make modules_install
>
> If i already have the same version of kernel running does old module
> tree get over-written or not?

Yes, it does overwrite the existing modules.

> ps please don't suggest i use make-kpkg :) i want to understand the
> process first before i choose automation.

I'm going to go ahead and suggest kernel-package anyways.  The
fundamental steps are the same (configure, compile, install) except
that you use Debian tools to install, which means you can later use
Debian tools to uninstall.  kernel-package also provides a much better
interface if you're going to try to add in modules that don't come
distributed with the kernel (OpenAFS, lm-sensors, ALSA, ...)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Get in Trouble Now

[David Z Maze]

arief_mulya <[EMAIL PROTECTED]> writes:

> But I do notice that gdm is back to the gnome1.4 version. And the font
> become very ugly.

The gdm is sid is the GNOME 1.4 gdm; I don't think the GNOME 2.x gdm
is there at all.  (Not entirely sure why, though.)  There are a few
bugs against the gdm package already requesting gdm2 (147637, 155638,
169226).

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Applying Debian diff to a newer upstream tar ball.

[David Z Maze]

Shaul Karl <[EMAIL PROTECTED]> writes:

>   I have just applied a debian .diff.gz to a newer upstream tar ball.
> Although the patch succeeded, any comments or pointers for discussion
> about this process would be appreciated. I do believe that my case was
> easy because the newer upstream tar ball was only slightly different
> from Debian's (old) .orig.tar.gz.
>   Anyway, what I did was:
>
>   1. gunzip -c nut_1.2.1-2.diff.gz > nut_1.2.1-2.diff
>   2. tar -xzf nut-1.2.2-pre2.tar.gz
>   3. (cd nut-1.2.2; ./configure)
>   3. patch --verbose -p1 -d nut-1.2.2 < nut_1.2.1-2.diff

I'd instead do

  1. tar xzf nut-1.2.2-pre2.tar.gz
  2. zcat nut_1.2.1-2.diff.gz | patch -p1 -d nut-1.2.2
  3. cd nut-1.2.2; dch -v 1.2.2.0pre2-0.1; debuild

with the devscripts package installed; that is, take the patch, apply
it to the updated source tree, change the package version number, and
build.  Besides this, if the patch applies, you're probably doing
pretty well for building an unofficial updated package.

(I guess the important thing is that I'm applying the patch to the
unconfigured package; it *shouldn't* make a difference, but you never
know...)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Bash scripting

[David Z Maze]

Jeff Elkins <[EMAIL PROTECTED]> writes:

> ARCH := $(shell uname -m | sed -e s/i.86/i386/ -e s/sun4u/sparc64/ -e 
> s/arm.*/arm/ -e s/sa110/arm/
>
> I'm working on polishing my meagre shell scripting skills and would
> appreciate some feedback on the line above, quoted from the kernel Makefile.
>
> 1. How would you use this in a straight bash script so it returned a value
> containing the boxes architecture?

The contents of the $(shell ...) Make fragment should be a valid shell
command.

> 2. As far as comparisons go, if I wanted to determine if a string contained
> ".xyz" or ".abc" would that be a variant of the fragment above?

Two somewhat common ways:

  uname -m | grep i.86 > /dev/null && echo x86

  case `uname -m` in
i?86) echo x86 ;;
  esac


-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Pkzip - Pkunzip

[David Z Maze]

Julio Diaz <[EMAIL PROTECTED]> writes:

> Can anyone recomend a good program to pkzip and pkunzip files in kde?

The 'unzip' package provides a program called 'unzip' which unzips
files.  This is a command-line tool, no particular dependence or lack
thereof on KDE or any other desktop environment.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: How can Iset an odd X resolutin?

[David Z Maze]

stan <[EMAIL PROTECTED]> writes:

> I've got a freshly installed Debian machien that I want to make function
> basicly as an X terminal for a Solaris box. The Solaris X resolutin is
> 1152x900 abt 8 bit color depth.
>
> How can I configure the Debian box to use this X setup?

Well, for the most part, X is X; any X setup should work fine.  (I do
remote connections between different sorts of machines all the time
and it works fine, even if one is an x86 machine runnign Debian and
another is a Sun running Solaris.)  You can set up XFree86 to use an
8bpp display by putting "DefaultDepth 8" into the Screen section in
/etc/X11/XF86Config-4, but you'll tend to be less happy here than if
you could get a higher bit depth at the same resolution.  You can also
specify display modes here (1152x900 is semi-standard); the minimal
section that accomplished this would look like

Section "Screen"
Identifier "Screen0"
Device "Card0"
Monitor "Monitor0"
DefaultDepth 8
SubSection "Display"
   Depth 8
   Modes "1152x900"
EndSubSection
EndSection

> And a 2nd question, is there an easy way to determint the fontpath being
> used on the Solaris machine?

'xset q' will tell you this (among other things).  You'd need to be
logged in locally there, though.  Most of the X stuff on Solaris lives
in /usr/openwin and /usr/dt, you might try looking there (particularly
/usr/openwin/lib/X11/fonts).

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: wtf? (long and frustrated)

[David Z Maze]

Glenn English <[EMAIL PROTECTED]> writes:

> There ought to be a list for debian wannabes. I've tried several times
> to get woody going on a couple different boxen - most recently a Dell
> Latitude laptop.

You said Latitude C500/600?  Mine (a Latitude C600) has a very
nonstandard 1400x1050 display.

> Console is fine; X, of course, has been the problem (I haven't even
> looked at the PCMCIA Ethernet and wireless cards yet). 

PCMCIA has worked just fine for me; some minor issues with the
internal wireless card, but current versions of the kernel PCMCIA
drivers work well.

> When the installer says, "Have fun," and reboots, the screen blinks a
> couple times, and a curses dialog box comes up saying it can't run X,
> telling me why, and offering to run the X configuration program - that's
> cool. I say, "Yes," and a program starts - IN X!!! 
>
> This is not funny, folks; it's sadistic.

Uh, how are you configuring the X server?  You might try, in no
particular order:

-- Removing or disabling xdm/kdm/gdm so it doesn't try to go into X on
   startup;

-- Reading the logs ('startx -- -verbose >xlog 2>&1' can be
   informative, dumps a lot of data in xlog);

-- Running 'XFree86 -configure' to cause X to give you what it thinks
   is a sensible configuration file in ./XF86Config.new.

*reads /etc/X11/XF86Config-4* ...it appears that the version of X in
unstable actually understands the 1400x1050 display by default; I
don't have to do anything special for this to be recognized or run in
its default resolution.  So you also might try digging up an XFree86
4.2 backport, if you're using stable.

> The mouse doesn't work, but there's a window telling which keys on the
> numeric keypad to use instead. Laptops don't have numeric keypads, and
> the system knows this is a laptop (I installed "Support for Dell
> laptops" and I saw something flash by while it was booting).

Hold down the "Fn" key, and the jkluio789 keys will act as a keypad.

> When I pressed ctl-alt-backspace, the screen slowly faded to white.

I've seen that before.  From what I can tell it doesn't actually hurt
the display, but it is hard to get back to a useful display mode.

> I've installed Red Hat, Mandrake, and SuSE on this machine with no
> probs. I've used video and screen data from the XF86 config files from
> those installs, and from Dell's dox. 

Huh.  I'd think that'd be likely to work, then...weird.

> Is there some FM or FAQ I've missed? Is there a CI program on Debian to
> configure X? Or is vi /etc/X11/XF86Config it? 

Other people have mentioned using dpkg-reconfigure to update the
XF86Config file too.

BTW, there's also a debian-laptop list, you might try asking there for
help...

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Where to put local .debs?

[David Z Maze]

Abdul Latip <[EMAIL PROTECTED]> writes:

> I am just wondering where others put the local .debs
> (e.g. the kernel-image). Or is it usual just to keep it
> in /usr/src/ after "dpkg -i kernel..." ?

My desktop machine used to have a big pile of .debs in /usr/local/src,
and that was all fine.  These days I use mini-dinstall to keep a local
repository with current kernels in it (/usr/local/debian on the
desktop, $HOME/debian on the laptop).

The location of these things really doesn't matter at all.  I've
somewhat accepted the mantra that /usr/src is under /usr but not under
/usr/local, so it's the domain of dpkg to manage it.  And none of my
machines have /usr/src/linux at all; having it would just confuse things.

> - may I know what "char-major-10-135" is? I read it is
>   related to "RTC"; but I am not sure about how to get rid
>   of that message...

See Documentation/devices.txt in your kernel source, and look up
character major 10 minor 135.  You can get rid of the error message by
either building a kernel module that supports it (which one is
somewhat obvious from having looked at the devices file; it's under
"character devices" in the kernel config) or by aliasing
char-major-10-135 to 'off' with a file in /etc/modutils.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Setting up network

[David Z Maze]

"Sharninder" <[EMAIL PROTECTED]> writes:

>>
>> ifconfig eth0 my.ip.add.ress up
>> route add default gw my.gw.ip.address
>>
>
> put these two commands in a script under /etc/init.d and link it

Eew.  Debian already provides perfectly good infrastructure for
providing network settings.  Edit /etc/network/interfaces and add

auto eth0
iface eth0 inet static
address a.b.c.d
netmask 255.255.255.0
broadcast a.b.c.255
gateway a.b.c.1

(or whatever is appropriate).  'ifup eth0' will bring up and configure
the interface; this also happens automatically at boot time.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: .bash_profile and X

[David Z Maze]

Dai Yuwen <[EMAIL PROTECTED]> writes:

> I'm running xdm + fvwm. And my shell is /bin/bash.  I find any
> variable I set in ~/.bash_profile doesn't take effect after I've login
> X.

Well, sure: nothing in the "logging in via X sequence" runs bash as a
login shell (shell scripts might get run via /bin/sh, but IIRC that
definitely doesn't read .bash_profile).  I like to create a .variables
script that sets a bunch of environment variables, and then read it
from both .zshrc (or .bashrc) and .xsession; this approach would look
like

  # .variables: set environment variables
  PATH=$PATH:$HOME/bin
  CLASSPATH=.
  export CLASSPATH

  #!/bin/sh
  # .xsession: run user commands for X
  set +m
  . $HOME/.variables
  xrdb -cpp m4 $HOME/.Xresources.m4
  epist &
  xwrits typetime=6:30 breaktime=1:00 clock mouse &
  exec openbox

The other frequent suggestion is to just read your .bashrc (or
.bash_profile, etc.) from .xsession, instead of creating a separate
file for environment variable settings.  YMMV.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

The case of the missing Helvetica

[David Z Maze]

I have a current sid machine.  If I open a GNOME 2 application with a
font selector, looking for Helvetica fails.  As an exercise, for
example, one might open Gnumeric, and type some text into a cell.  I
can change the font to "Arial" since I copied a bunch of fonts from
the Windows side of my laptop, but "Helvetica" doesn't appear in the
list and manually selecting it gets me something slightly unfortunate
that's very visibly sans serif but not Helvetica.  It's clearly not
that X doesn't believe in the font; I can draw things in
-*-helvetica-medium-r-normal--0-110-123-124-*-0-iso8859-1 just fine.

I know little enough about how GNOME 2 font selection works to track
down where this is going on.  (The larger problem is that I get the
same ugly font in Galeon.)  Any hints as to why this is happening?  Is
it something on my machine, or a (presumably reported) bug in the font
system?

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: How can Iset an odd X resolutin?

[David Z Maze]

stan <[EMAIL PROTECTED]> writes:

> OK I'm partway here, but I'm still having troubles. I ran videogen and got
> a modeline as close to 1152x900 as it would generate
>
> Modeline "1152x896"  94.21 1152 1184 1440 1472 896 898 941 943
>
> I then put this line in /etc/X11/XF86Config-4. Here is the XF86Config-4
> file:

> Section "Screen"
>   Identifier "Screen0"
>   Device "Card0"
>   Monitor"Monitor0"
>   DefaultColorDepth 16
>   SubSection "Display"
>   Depth 16
>   Modes "1024x768" "800x600" "640x480"
>   EndSubSection
> EndSection

> Then I restarted teh X server. Unfortunately it did not use the new
> modeline.

Right; you'd need to include that mode in the Modes line for the
relevant bit depth.

I'm not entirely clear why you're doing this, though.  You had
originally said you wanted the new machine to essentially be an X
terminal for an extant Solaris machine.  You can just do that with any
X server, regardless of platform, resolution, or bit depth (hence my
suggestion earlier that you try to use a higher bit depth if your card
will support it).  You don't need to exactly match the display the
remote machine has; the remote machine doesn't even need to have a
display, and this still works fine.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Package system vs. source vs. both

[David Z Maze]

Radek Zajkowski <[EMAIL PROTECTED]> writes:

> As a result I downloaded the binaries of Xfree and it runs as a charm. I
> compiled Emacs and my crappy pentium200 is now a bit more friendly, or at
> least it offers the alternative to terminal.

> The problem I have created here is rather obvious, the package manager
> doesn't know I have X libraries on my system and therefore, anything
> requiring Xlibs will not install, since it forces the dependecies to be
> configured as well. You probably get the rest of the story.

So when you're doing this sort of thing what you almost always care
about is having a newer X server than Debian provides.  You can
install all of the X client stuff (libraries, xterm, twm, and so on)
using the Debian infrastructure.  You can install the Debian X server,
too, if it makes the packaging system happy, but it shouldn't be
necessary.  Meanwhile, put the XFree86 binary server tarball somewhere
like /usr/local, and repoint the /etc/X11/X symlink to point at
/usr/local/bin/XFree86.  Now as far as dpkg/APT are concerned, you
have X libraries (and those change very rarely), but you're using the
newer server that supports your hardware, which is what you really
want.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: default gateway

[David Z Maze]

Pavlos Parissis <[EMAIL PROTECTED]> writes:

> I am looking for the file that I have to modify in order to have static route
> enabled.
> I add the gateway manually with route add -net default gw 192.168.100.1 
> and I would like to find the config file.

You're almost certainly looking for /etc/network/interfaces; edit it
with your favorite text editor.  'man interfaces' has some details on it.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: how to map F12 to a string in XWindow

[David Z Maze]

Paulo Henrique Baptista de Oliveira <[EMAIL PROTECTED]> writes:

>   Something else people like to change are the bindings of the function
>   keys.  Suppose that you want to make F12 produce the string "emacs ".

>   Now I want to know how to the same thing in a xterm inside XWindow.
>   How can I do this?

You'd need to add an X resource.  From looking at xterm(1), it looks
like you should add

XTerm*VT100.Translations: #override F12: string("emacs ")

to your .Xresources file, then run 'xrdb -merge .Xresources'.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: how to apply diff file

[David Z Maze]

"nate" <[EMAIL PROTECTED]> writes:

>> I have downloaded the source file of an application (apt-get source appl).
>> Along with the .tar.gz file, I got and diff.gz file which I don't know how
>> to apply it to the .tar.gz file (after the extract of cource).
>
> last i checked apt-get source automatically extracted and patched
> the source with the diff, you can see if this is the case by checking
> to see if there is a debian/ directory within the extracted source tree
> of the app you apt-get source'd.
>
> in any case, the usual command to use with a diff is patch, be sure
> to decompress it first(gzip -d or gunzip)

...or, for completeness, if you have a .orig.tar.gz, a .diff.gz, and a
.dsc file all for the same package, you can use
'dpkg-source -x foo_1.0-1.dsc' to unpack the original source and apply
the Debian diff.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: 100dpi, 75dpi, unscaled ?

[David Z Maze]

dave selby <[EMAIL PROTECTED]> writes:

> Debian has a lot of fonts  what is the difference between 
>
> /usr/lib/X11/fonts/100dpi
> /usr/lib/X11/fonts/100dpi/:unscaled,
>
> I cant find any .. can someone enlighten me ? Do I need both ?

If I ask for

 -*-helvetica-medium-r-normal--*-110-100-100-*-*-iso8859-1

but there are only 10 and 12 point Helvetica fonts in
/usr/lib/X11/fonts/100dpi (I asked for 11.0 point), the :unscaled
version won't return a font at all, but the not-unscaled one will try
to scale one or the other to the requested size.  This is bad with
bitmapped fonts, since you don't wind up with smooth edges (or even
consistent-width lines, often).

> Also if I have the 100dpi versions, is there any advantage to having the 
> 75dpi versions, are these just less memory intensive versions of the same 
> thing ?

It's not memory, it's how big your screen is.  Try this exercise: take
the horizontal resolution of your display, divide by .8, and divide
again by the diagonal size of your monitor in inches.  If this number
is closer to 75, the 75dpi fonts are closer to accurate for you; if
it's closer to 100, the 100dpi fonts are better.  In any case you can
be happy with scalable (Type 1 or TrueType) fonts.  "Happiness" here
is relative, too, it's really how much you care about a "12 point"
font being exactly 1/6 inch tall.

(It turns out that X does have a concept of display pitch, and
'xdpyinfo' will tell you this, among many other things.  The last time
I looked at an X server source, though, X would only hand out 75 or
100 dpi fonts unless you asked really nicely [e.g., with explicit
resolutions in the font name]; modern XFree86 might be different.  I
think the newer font-rendering libraries try to use the reported
display resolution.  Getting this sort of thing right can be tricky,
but as I understand it, the Unix world has a lot more hope than
Windows...)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Lightweight xlock alternatives?

[David Z Maze]

Remo Inverardi <[EMAIL PROTECTED]> writes:

> Is there a lightweight xlock alternative? Actually, I just want to
> lock my X session, so I don't have to logoff when I leave my office to
> get coffee. Xlock with the "-lock blank" parameter is about what I'm
> looking for, but it's *way* to big.

In what way is it heavyweight?  I generally use xscreensaver, which
runs as a daemon; 'xscreensaver-command -lock' locks the screen.
Pretty graphicy things run as separate processes, so I think starting
it up should be pretty cheap.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Gnucash errors in unstable

[David Z Maze]

David Goodenough <[EMAIL PROTECTED]> writes:

> It does seem odd that a -dev package is needed to get a program to run.
> Surely -dev packages are for compile time things, not run time.
>
> This suggests that there are two bugs, one in the depency list for
> gnucash and the other in the packaging of libgwrapguile (not dev).

Have you looked at the Debian Bug Tracking System?  There are, in
fact, bugs against both gnucash (184683) and libgwrapguile1 (184681)
about this.  (http://bugs.debian.org/)  Poking around suggests that
the actual bug involves (use-modules (g-wrap gw-wct)) in
/usr/share/gnucash/guile-modules/gnucash/main.scm, which probably
causes Guile to try to load libgw-wct.so as a dynamically loaded
module; thus, libgw-wct.so doesn't show up as a direct dependency of
gnucash, but libgwrapguile1 doesn't provide the .so link since this
generally isn't done for real library packages.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: tune gcc

[David Z Maze]

Jerome BENOIT <[EMAIL PROTECTED]> writes:

> I have a very naive question:
> Can Debian tune the GNU C Compiler ?
>
> [e.g., put `-cpu=pentium' automaticly on Pentium box]

There's a pentium-builder package that tries to do this, or you can
try to set CFLAGS for the program in question.  With a limited number
of exceptions, though, conventional wisdom is that this buys you
little in performance at the cost of portability of binaries to other
x86-based machines.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Kernel compile

[David Z Maze]

"GBV" <[EMAIL PROTECTED]> writes:

> Anyone knows the best web reference to
>
> source to download the newest kernel
> compile a new kernel, in debian with dpkg
> configure lilo
> booting, and restoring if a sinister occurs..
> managing several kernel versions

I don't know of a Web reference (somebody else probably does), but:

(1) Kernel sources are frequently available in Debian
(kernel-source-...); if not, you can look on
http://www.kernel.org/.

(2) 'apt-get install kernel-package', and read its documentation
(/usr/share/doc/kernel-package/README.gz, make-kpkg(8)).

(5) Essentially, put all of the kernel versions you care about in your
boot loader configuration.  If you decide you don't want one any
more, you can just remove it using dpkg like any other package.

Short summary, assuming you have everything (kernel modules, gcc,
libncurses5-dev, bin86, fakeroot, kernel-package) installed:

  for i in /usr/src/modules; do tar xzf $i; done
  export MODULE_LOC=$PWD/modules
  wget http://www.kernel.org/pub/linux/kernel/v2.4/linux-2.4.20.tar.gz
  tar xzf linux-2.4.20.tar.gz
  mv linux linux-2.4.20
  cd linux-2.4.20
  cp /boot/config-`uname -r` .config
  make menuconfig
  make-kpkg --rootcmd=fakeroot --revision=custom.1 \
--append-to-version=-dzm kernel-image
  fakeroot make-kpkg --append-to-version=-dzm modules-image
  sudo dpkg -i ../*.deb

Some of this is unnecessary if you don't have any add-on kernel
modules (ALSA, lm-sensors, OpenAFS, etc.).  You can run this from
anywhere.  There's nothing particularly special about /usr/src/linux,
and none of my machines have a link or directory there.  If you have a
previously built kernel, you might try starting with that kernel's
.config rather than what you have installed now.  Finally, all of this
will produce kernel-image-2.4.20-dzm and related packages; I'm
starting to think the --append-to-version is nice (and will typically
use the name of the machine I'm building the kernel for) since then
I'm guaranteed to not conflict with the official Debian kernel
packages.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: GTK/debconf/Gnome warnings with apt-get install

[David Z Maze]

Charlie Zender <[EMAIL PROTECTED]> writes:

> I am a newbie using debian unstable. A volatile combination, no doubt.
> Whenever I 'apt-get install foo' or 'apt-get dist-upgrade', the
> procedure works but prints lots of GTK and debconf/Gnome warning
> messages as follows:
>
> Setting up toshset (1.58-3) ...
> Xlib: connection to ":0.0" refused by server
> Xlib: No protocol specified
>
> Gtk-WARNING **: cannot open display: :0
> debconf: unable to initialize frontend: Gnome
> debconf: (DISPLAY problem?)
> debconf: falling back to frontend: Dialog

Hmm, that sounds like you have Debconf set up to use GNOME to ask
questions at configuration time, which can be troublesome.  You might
change the default by running

  dpkg-reconfigure -fdialog debconf

> Any hints on how to get rid of these warning messages?
> Is there a permission I need to set somewhere? xhost?

Eew.  You should probably forget xhost exists.  :-)  The Remote-X-Apps
Mini-HOWTO is what I usually point people at when they ask this sort
of question; one easy workaround if you're just worried about X access
from root on the local machine is to set the XAUTHORITY environment
variable to point to $HOME/.Xauthority (as your mortal user).

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: bug apt-get dist-upgrade

[David Z Maze]

hina <[EMAIL PROTECTED]> writes:

> Do you want to continue? [Y/n] y
> Get:1 http://okki666.nerim.net ./ xchat-common 2.0.2.CVS20030317-1
> [320kB]

You might read the description for xchat-common; on my system, it says
that it is "useless" without certain other packages installed.  Even
better, use a package front-end (like aptitude) that will show you the
descriptions and make it easier to install recommended packages.

> zeus:/avi/mldonkey/mldonkey-distrib-2.02-0# xchat

You're trying to run an IRC client, as root?  That sounds like a
terrible idea...

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Memory leak - somewhere :)

[David Z Maze]

Jeetu Golani <[EMAIL PROTECTED]> writes:

> For sometime, I've been feeling that there's a memory leak in one of
> the apps because after quite a few days of use KDE becomes
> slow. Apps I typically use are KMail,Konqueror,KMerlin,Konsole,K3B.
>
> I've noticed that memory seems to be consumed at an alarmingly huge
> rate. Sometimes about 300MB+ gets used up overnight inspite of no
> additional programs being run. For e.g. if with a clean boot into
> KDE I have 300MB of free Physical memory left (as displayed by free)
> and the whole swap free then the next morning I may see just 5MB of
> Physical memory free and the Swap used to a degree.

You could post the output of 'free' that causes you to believe that.
Or read a Linux FAQ, any one, it doesn't matter: Linux considers it
better to use memory to improve performance (e.g. by doing disk
caching) than just let it sit idle, so you'll typically have very
little physical memory "free" without the system performance being
degraded at all.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: algorithm for correct sources.list lines?

[David Z Maze]

Paul Scott <[EMAIL PROTECTED]> writes:

> 1. Is the algorithm for creating lines for sources.list defined
>somewhere?

sources.list(5) has most of the information you need; you can also
poke around with a Web browser to look for things.

> 2. Is this line correct?
>
> deb
> ftp://metalab.unc.edu/pub/linux/devel/lang/java/blackdown.org/debian
> unstable main non-free
>
> 3. Is there an http version of this line?

Syntactically, it's correct, sure; whether that particular resource
happens to exist or not is outside of APT's control.
/etc/apt/sources.list on my laptop has

deb http://www.ibiblio.org/pub/linux/devel/lang/java/blackdown.org/debian woody 
non-free main

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Make-kpkg

[David Z Maze]

"Stephen J. Thompson" <[EMAIL PROTECTED]> writes:

> I am trying to find documentation to create modules with
> make-kpkg. I am trying to find out how I unpack a tar file and
> manipulate it so that I can issue a make-kpkg module_image ...

Like, a tar file in /usr/src that a *-source package installed?  You
can unpack it in /usr/src, or unpack it somewhere else and set the
MODULE_LOC environment variable to point to the resulting modules/
directory.  You also need a kernel source tree that was built using
kernel-package.

(In other words, I think you can't do this without full kernel source,
since make-kpkg wants to get information about things like the
revision of the kernel out of the files in the debian/ directory.  The
kernel-headers/kernel-build packages don't seem to provide this data.)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Root login to Windows X

[David Z Maze]

Francisco Castellon <[EMAIL PROTECTED]> writes:

> 1.  (*) text/plain  ( ) text/html   

(Plain-text mail only, please!)

> I just finished installing my first Debian system and I seem to not be
> able to log on to Windows X using the root account, I can easily log in
> with any other account though.

...oh, so you're set: you can just use su or sudo to become root.

> I am trying to do some administrative tasks but I want to do them
> through the desktop since they are far simpler that way.

Like what, specifically?  Most of the things I actually want root for
involve either (a) manipulating installed packages [dpkg, aptitude],
(b) adding users [adduser], or (c) changing individual file
permissions [chown, chmod, &c.]; none of these really have good GUI
equivalents.

In general you want to do the least possible with the root account.
Imagine, for example, that somewhere hidden deep within one of the
GNOME libraries is some code that might randomly scribble over
/dev/hda.  If you're logged in as a mortal user, your program dies
with an obscure "permission denied" error, which is odd, but no damage
done.  If you're logged in as root, the safety checks go away, and
your hard drive gets scribbled over because of someone else's bug.
Doing everything-but-what-you-need-to as a normal user also minimizes
(but doesn't eliminate) the damage you can do by typing 'rm * ~' with
one more space than you really intended, among other things.

> Is there a way that I can get to the desktop with the root account?

That having been said, if you *really* want to, there's generally a
frob in your display manager configuration that lets you change it.
In the particular case of gdm, there's an 'AllowRoot' setting in
/etc/gdm/gdm.conf you can change; I don't know about other display
managers.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Potato to Woody upgrade

[David Z Maze]

Larry <[EMAIL PROTECTED]> writes:

> If I currently have Potato installed, and a stack of
> Woody CD's, how do I upgrade from Potato to Woody?
>
> Do I have to just install over Potato?
>
> Or, does apt-get or dpkg allow a more graceful upgrade
> in some automatic fashion?

You should be able to drop each of the Woody CDs into your CD-ROM
drive in turn and run 'apt-cdrom add'.  Having done that, you should
be able to do 'apt-get dist-upgrade', or do the upgrade within a
package tool like dselect or aptitude.  You don't need to reinstall at
all.

> If there's an upgrade method, will it goof up things
> like my lan connections, or remote printer access,
> mutt, etc?

It shouldn't; in general, packages try to be good about preserving
user settings on upgrade.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: joysticks and xfree 4.x

[David Z Maze]

[EMAIL PROTECTED] writes:

> I've got all of the linux drivers installed and everything works fine
> from the OS point of view (thanks to the joystick package).
>
> There seems to be a total lack of information as to how to get the
> joystick to work under X.
>
> Can anyone point me to some docs on how to do this or share their
> xfconfig file ?

What do you want to do with it?  My desktop machine has a joystick,
and while the joystick and ALSA don't play along nicely, if I unload
the sound drivers and load the joystick kernel driver, if works just
fine with things like flightgear without any particular XF86Config
hacking.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: PDF presentation from LaTeX: TexPower? Prosper?

[David Z Maze]

Glen Snyder <[EMAIL PROTECTED]> writes:

> 2. Can anyone strongly recommend another latex package (Prosper?)
> supplied by debian that would allow me to insert a bunch of EPS figures
> and math formulas, and produce an electronic presentation that can run
> on Acrobat Reader. 

I taught a short Java class in January and made slides using Prosper.
It worked great (and *much* better than trying to use Magicpoint last
year); I used 'xpdf -fullscreen -papercolor black' to actually run the
slides, since xpdf is (a) free and (b) much smaller and faster than
acroread.

(Things that didn't work: last year's slides were done with
m4-generated MagicPoint, but the presentation looks much different
from the printed slides, and the syntax sucks and it's hard to get
things to look nice.  I had a somewhat brief foray into the world of
DocBook, but I didn't want to do DSSSL/XSLT hacking to try to get
something even a little presentable.  I finally realized that if what
I really wanted was PS/PDF, then I could generate it very
straightforwardly from *TeX.  I tried TeXPower but disliked it for
reasons I don't recall before settling on Prosper.

The one caveat with Prosper is that the slides look terrible in xdvi.
This is fine; you need to go ahead and run dvips over it and then
preview things in gv.  The same may actually be true for TeXPower,
before you abandon that.)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Scrolling Apps = High CPU Utilization

[David Z Maze]

"M. Kirchhoff" <[EMAIL PROTECTED]> writes:

> First, let me describe the problem: When using the wheel on my
> logitech wheel mouse (PS/2, ImPS) to scroll up and down in various
> apps (Mozilla, AbiWord), I see very high CPU activity as reported by
> "top".
...
> nVidia GeForce2 MX 200, 64MB vRAM
> Debian Woody, stock bf2.4 kernel, stock "nv" video drivers, *no*
> kernel patches
>
> Any ideas why CPU util. is shooting so high? Is that normal? I
> understand that I'm just using the generic stock "nv" drivers for my
> video card, but would using the proprietary nVidia drivers make much
> difference for simple 2D use? Would recompiling my kernel to better
> suit my platform make a significant difference?

My guess is that you wind up doing the scrolling operation all in
software with the nv driver, which is why you get that much CPU
usage.  I'd guess that switching to the (non-free) NVidia driver would
in fact help you get around this problem.  There's not much you could
change in the kernel configuration to improve it, I don't think.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Using rexec command

[David Z Maze]

"Victory" <[EMAIL PROTECTED]> writes:

> 1.  (*) text/plain  ( ) text/html   

(Please post to the list in plain text only, and set your mailer to
wrap lines at 72 columns.)

> Anyone know how to set the flag for the "exec" entry in
> "/etc/inetd.conf" file so that it allow me to run " rexec -l root -p
> passwd hostname uname -a" from the remote workstation without adding
> remote workstation in to server /etc/hosts and without using
> /etc/resolv.conf file ?.

You almost certainly don't want to use the rsh/rexec family of
commands; ssh can do the same things, and is much more resistant to
network sniffing.  I'd create a DSA keypair, add the public key to
root's authorized_keys file on the remote host, and then do
'ssh [EMAIL PROTECTED] uname -a'.  Unless you're doing something more
involved than this, and really do have a clue about what you're doing,
you should absolutely have a password on your private key, though you
might find it convenient to use ssh-agent to save the key in memory.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: XFS over EXT3

[David Z Maze]

Alvin Oga <[EMAIL PROTECTED]> writes:

> On Wed, 15 Oct 2003 [EMAIL PROTECTED] wrote:
>
>> Greetings to all:
>> 
>> I'm planning to install debian and planning to use XFS instead of Ext3,
>> does anybody know how to do ti, or know of any advantage of one file
>> system over the other, any recomendation will be appretiated.
>
> for a benchmark comparison
>   http://aurora.zemris.fer.hr/filesystems/

This doesn't strike me as really rigorous.  It also doesn't strike me
as really current; their filesystems are patches vs. kernel 2.4.5,
which is from May of 2001, and they express some concerns over the
"testing" nature of ext3, where it seems to be quite stable for most
people on this list.

> use xfs instead of ext3 ... 
> ( if ext2 gets seriously messed up, ext3 is dead )

"If the bits get corrupted, your filesystem doesn't work."  This
applies to any filesystem.  I've never heard of people having issues
with ext3, and it being a layer on top of ext2 is generally seen as an
asset (since older tools can still use the underlying filesystem).

Common wisdom seems to be that ext2 is just fine for most users, but
it has the annoyance of long fsck times if the machine goes down
unexpectedly.  ext3 is also fine for most users and gets around this.
reiserfs is supposedly good for situations where you have lots and
lots of small files (e.g., news servers), but there's some FUD about
tool support for it.  I haven't really heard anything compelling about
anything else, including XFS.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: The following packages have been kept back

[David Z Maze]

"D." <[EMAIL PROTECTED]> writes:

> Haven't you ever done a dselect update and then a apt-get -u upgrade
> and found that you have 30 or some large number of packages that are
> not going to be installed?

Not really; apt-get isn't intended to be used that way.  See the first
paragraph of apt-get(8).  I'm reasonably happy using aptitude for most
things, especially since it gives me at least a little control over
upgrade conflict handling ("no, don't remove these seven packages I
care about, put perl on hold instead").

> When that happens, I have done the apt-get dist-upgrade and it will
> install a lot of packages that were not going to be installed during
> a normal upgrade... I'm running testing in case your curious.

...and if you *are* in that world, you probably want to be using
dist-upgrade all the time anyways, since sometimes packages change
names, split up, and so on.  'apt-get upgrade' is mostly useful for
incremental changes if you're tracking stable and you really really
really don't want your installed package list to change at all.  (And
even then, I still use aptitude on my woody machine.)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: modutils problem

[David Z Maze]

LeVA <[EMAIL PROTECTED]> writes:

> I have noticed that when I load a module the
> /etc/modutils/ files, which conatins the "post-install"
> lines, doesn't run.
> So when I load my emu10k1 for example, which has a
> /etc/modutils/emu10k1 file, which contains this line:
> post-install emu10k1 /usr/local/etc/emu-script
> doesn't run.

Did you run update-modules as root after creating the /etc/modutils
file?  (Nothing actually runs the commands in /etc/modutils/*
directly, but update-modules combines those files together into
/etc/modules.conf, which is what's actually read by modprobe.)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Where can I get lame for "testing"

[David Z Maze]

stan <[EMAIL PROTECTED]> writes:

> lame appears in dselect, and I've tried to install it that way, but
> it does not install. I've tried apt-get install lame, and I get "No
> installation caidate"
>
> What can I do to fix this?

Use (DFSG-free, not patent-encumbered, royalty-free) oggenc, from the
vorbis-tools package, to create Ogg Vorbis files for compressed music.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Epiphany.

[David Z Maze]

David Palmer <[EMAIL PROTECTED]> writes:

> What risk do I run with apt-get remove epiphany (the game), without
> disturbing epiphany the browser (with bookmarks).

What I'd suggest you do:

(0) Install aptitude, if you haven't yet.

(1) Start aptitude.

(2) Press '/', type "ephiphany" in the box that comes up.  This will
search for packages with "epiphany" in their name; pressing '\'
will find the next one.

(3) Check that you've found the right package.  With the default
setup, the package description will be displayed in the bottom
half of the screen; you can also press enter to see the full
description and dependencies.

(4) Press '-' to remove (or '_' to purge) the package.

(5) Press 'g'; aptitude will display what it's going to do.  (Press
'q' if it's wrong.)

(6) Press 'g' to make it go.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: big troubles in little libc

[David Z Maze]

iain d broadfoot <[EMAIL PROTECTED]> writes:

>   I'm having a few problems with programs dying:
>
> liferea:0x407196c9 in free () from /lib/libc.so.6
> gaim: 0x407466c9 in free () from /lib/libc.so.6
>
>   I can't see a bugreport about this on libc6, and it doesn't feel
>   like the individual apps are doing anything in particular
>   wrong...

Segfaults in free() are a fairly typical symptom of the application
screwing up memory management, actually.  freeing a block that was
never malloced, freeing a block multiple times, and writing outside of
the allocated boundaries could all cause this.  If you're up to
debugging it, electric-fence is a very useful tool; the way I use it
is generally like this:

  myprog$ ./configure
  myprog$ make CFLAGS=-g
  myprog$ gdb myprog
  (gdb) set environment LD_PRELOAD /usr/lib/libefence.so
  (gdb) run

and you'll generally get a segfault where the program actually messes
up, rather than when the symptoms are seen later.  There's a cost in
both memory and runtime, but using LD_PRELOAD inside gdb makes this a
lot less permanent.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: moving boot drive around and lilo

[David Z Maze]

james terris <[EMAIL PROTECTED]> writes:

> Then I enter the command:
>
> LD_LIBRARY_PATH=/mnt/lib /mnt/sbin/lilo -r /mnt
>
> And i get the error:
>
> sh: /lib/ld_linux.so.2: version 'GLIBC_PRIVATE' not found (required by
> /mnt/lib/libc.so.6)

Not having any idea what you're booting off of, does something like
this work:

  LD_LIBRARY_PATH=/mnt/lib:/mnt/usr/lib /mnt/lib/ld-linux.so.2 \
/mnt/sbin/lilo -r /mnt

(In other words, explicitly specify the dynamic loader to use, in case
the one on your boot floppy doesn't do the right thing.)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: update in sid has killed gnome-terminal

[David Z Maze]

TR <[EMAIL PROTECTED]> writes:

> I just did an upgrade in a machine running sid and after that can't star
> a gnome terminal anymore.

Congratulations.  I take it you've tracked down the nature of the
error, looked on http://bugs.debian.org/gnome-terminal, and reported
the bug if it hasn't already been, right?  Because you get to do that
sort of thing when you run unstable; it does break sometimes, and the
Debian bug-tracking system is a far better way to communicate with
developers about broken packages than complaining without details on
debian-user.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Configuring three nics on a gateway/web server

[David Z Maze]

Lucio <[EMAIL PROTECTED]> writes:

> I've successfully installed and configured Apache on a already
> working gateway box.
>
> I know this can play a little unusual (web and gateway server in the
> same box) but unfortunately at the moment I just have this hardware
> at my disposal.
>
> However, I have inserted a third nic (eth2) on which the httpd
> daemon should have listen.

So what's the third network card actually connected to?  "Apache
should listen on it" isn't actually informative.

> Some information on the net :
> eth0: gateway
> eth1: local net
> eth2: httpd
>
> network: 101.102.103.112
> broadcast: 101.102.103.119
>
> router: 101.102.103.113
> eth0: 101.102.103.114   netmask 255.255.255.248
> eth1: 192.168.0.1   netmask 255.255.255.0
> eth2: 101.102.103.115   netmask 255.255.255.255  <-- is this netmask correct 
> ???

So what that's saying is that eth0 is connected to a network
containing 101.102.103.113 through .118 only; .119 is the broadcast
address.  eth1 is connected to the private network on 192.168.0.x.
And eth2 is a single-host network; there are no other IP addresses on
the network eth2 is connected to, but its subnet is a subset of the
network eth0 is connected to.  This gets confusing.

Is your end goal just to have a separate IP address for the "Web
server"?  You could do that by assigning a second address to eth0
(it'd be called eth0:1).  Or if you really really wanted to use eth2,
you could connected it to the same thing eth0 is connected to, and
give it the same network parameters but a different IP address.  But
then your firewall rules might cause issues, as you showed before, and
it's not obvious whether packets going to "the router" in your
application go out via eth0 or eth2.  In normal situations where you
have "the gateway" you only have a single external connection; that's
where your cable modem or DSL or T1 or whatever plugs in.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: how to set up a mix from stable/testing/etc

[David Z Maze]

Christian Schnobrich <[EMAIL PROTECTED]> writes:

> Woody's spamassassin (2.20) provides me with a hit rate of about 60%;
> partly to increase this, partly because I'm curious about that fancy
> Bayes filter, I'd like to have a more recent spamassassin.
>
> But how?
> I'm a little shy of installing tarballs -- I'm afraid this would give me
> trouble when eventually I want to install the .deb package.

If all you want is a single package, installing from source isn't a
bad way to go.  You can install in $HOME or /usr/local and the Debian
packaging system will stay away.

> Leaves me with getting it from testing; I regularly see posts on this
> list where people claim to have a mixed setup from stable / testing /
> unstable, but I found no clue on how I could do this myself.
>
> Could someone please provide me with a few keywords for a google search,
> or point me directly to the appropriate documentation?

Google for 'apt pinning'.  The old-fashioned way is to put a testing
(or unstable) source in /etc/apt/sources.list, run 'apt-get update',
'apt-get install spamassassin', and then take the testing source out,
but then you don't track updates.

The other thing to be aware of doing this is that you'll get more than
just spamassassin.  Of note, unstable's spamassassin (2.60-1) depends
on spamc (>= 2.30), and spamc depends on libc6 (>= 2.3.2-1), so you
wind up pulling in the unstable libc6.  (You do at least avoid the
unstable perl.)  Sometimes this can cascade uncomfortably, leaving you
with a mostly-unstable system with a couple of stable packages
floating around.  YMMV.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: how to figure out the gateway number?

[David Z Maze]

Andrew Kasza <[EMAIL PROTECTED]> writes:

> I have debian 3.0. I have to change the configuration of network (I
> mean I have to change IP, netmask and so on).
>
> I know my IP address, netmask, broadcast.
>
> Is there a way to figure out the network and gateway number?

The network address is your IP address bitwise-AND your netmask.

  IP address 18.208.0.22, netmask 255.255.0.0 ==> network 18.208.0.0
  IP address 18.101.2.57, netmask 255.255.255.240 ==> network 18.101.2.48
  IP address 192.168.7.34, netmask 255.255.255.0 ==> network 192.168.7.0

The gateway address is specific to your network.  I'm used to seeing
it being the "first" address in the network (e.g., 18.208.0.1), but
I've also seen mention of a convention where it's the "last" address
(192.168.7.254).  In any case, whoever gave you the IP address should
be able to tell you what the default gateway should be.

> Is it enough to do the following three steps?:
> 1step 'ifdown --all'
> 2step ' to change the file /etc/network/interfaces'
> 3step 'ifup --all'

That should be sufficient, yeah.  If I were doing it I'd just down/up
the one interface I was changing (likely eth0).

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: why is debian the only distribution that won't let me run X?

[David Z Maze]

[EMAIL PROTECTED] (Bruno Boettcher) writes:

>   seems there's no way to get X running under debian on my new Medion
>   laptop

It'd help if you told us exactly what was wrong; looking at the log
files you sent pointers to, it sounds like "the X server doesn't start".

>   chipset is a GeforceFX from nvidia, Proc [EMAIL PROTECTED]
>   the log files are here:
>   http://www.inforezo.org/~bboett/xfree.log
>   http://www.inforezo.org/~bboett/XFree86.0.log
>   http://www.inforezo.org/~bboett/XF86Config-4

Is there a reason you're trying to use the "vesa" driver, as opposed
to the "nv" driver?  You might have a little more luck with the
latter.  (This is the free unaccelerated NVidia driver that comes with
XFree86.)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: ldconfig and hosing my system

[David Z Maze]

Bill Benedetto <[EMAIL PROTECTED]> writes:

>  I ran "ldconfig" when I had an LD_LIBRARY_PATH variable set to
>  point to old old libraries.  Once "ldconfig" finished, then no
>  commands would work.  Nothing.

Oops.  :-(  If you happen to have a statically linked shell around,
then you could try adding 'init=/bin/sash' or something equivalent to
your kernel command line at boot time.  (In LILO, type 'linux
init=/bin/sash' or something equivalent; in GRUB, press 'e' at the
boot menu screen to edit what gets run.)

>  At this point I *BELIEVE* that I hosed /etc/ld.so.cache .  (I
>  could certainly be wrong, of course.)  But since I can't bring
>  the system up I don't know how to recreate /etc/ld.so.cache .

Right, that sounds consistent with what you're seeing.

>  I tried booting knoppix and using the /etc/ld.so.cache there but
>  the kernel still paniced upon reboot.  I don't know if the
>  kernel paniced the same way or not.

A kernel panic seems a bit weird, though maybe you're interpreting the
"can't load init" message as a kernel panic.  But if you can boot from
a Knoppix CD and mount your root partition (say, on /mnt), then either
of

  /mnt/sbin/ldconfig -r /mnt
  chroot /mnt /sbin/ldconfig

have a hope of working ("pretend that /mnt is the root directory, then
run ldconfig, which is statically linked").

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Xserver-Xfree86

[David Z Maze]

"David R Hovland" <[EMAIL PROTECTED]> writes:

> Hi guys, I'm back.  Thanks for the advise, it worked, part way.
> I ran apt-get install xserver-xfree86.
> Now the problem seems to start at:
> (--) Assigning device section with no busID to primary device
> (EE) No device detected.

What video card do you have?  What driver are you telling X to use?
What version of X?

> I have run apt-reconfigure xserver-xfree86 and set the busID to
> PCI:1:0:0 (as seen on a google search) also PCI:01:00:00 with above 
> results.  the lspci gives me 01:0.0 for this card.

That's probably unnecessary.  It'd be important if you happened to
have two cards that looked the same to the X server.

> PS- is this post ok as far as word wrap, no html?

Yup.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Auth problem connecting to my ftp server

[David Z Maze]

Micha Feigin <[EMAIL PROTECTED]> writes:

> 220 litshi.luna.local FTP server (Version 6.4/OpenBSD/Linux-ftpd-0.17)
> ready.
> 500 'AUTH GSSAPI': command not understood.
> 500 'AUTH KERBEROS_V4': command not understood.
> KERBEROS_V4 rejected as an authentication type
>
> What does it meen, should I be worried and how do I stop it?

It means your default 'ftp' is the ftp program from the Kerberos
distribution, possibly from the krb5-clients.  If you're not using
Kerberos (unless you're doing something strange on your local network,
you're almost certainly not; Kerberos can be pretty unhappy around
NATs, too) you could just uninstall the offending ftp.  "If you were
using Kerberos, you'd probably know."  Otherwise, the message is
harmless.

If you did want to have a Kerberized ftp around but not use it most of
the time, you could also change the 'ftp' alternative to point to
something else.  On my system (sid), ftp is an alternative link which
could point to either netkit-ftp or krb5-ftp.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: failing to upgrade sysvinit from Knoppix version

[David Z Maze]

Simon Tod <[EMAIL PROTECTED]> writes:

> After a hdd install from knoppix 3.2 I tried to dist-upgrade to
> Debian unstable. It all seems to work fine but my debian_version is
> still reported as testing/unstable.

Isn't that what it's supposed to be?  That's certainly what it is on
this sid machine.

> I've tracked this down to the fact that a number of packages from
> the hdd install are not at the latest versions available in unstable
> but are unoffical versions with different dependencies. The most
> significant being sysvinit.
>
> If I try 'apt-get install sysvinit', I'm told I have the latest
> version installed. Not true. The knoppix version is 2.84 something
> or other, whereas 2.85 is available in unstable.

Knowing what exactly the version is might be helpful.  If the Knoppix
people have added an epoch to their version number, APT would be
entirely correct in concluding that 1:2.84-mumble is newer than
2.85-7.  You might also check that your APT sources.list is correct,
and that you haven't set up funny APT pinning rules.

ObAreYouSureYouWantToRunUnstable: booting the laptop was, um, fun this
morning, because I didn't have an /sbin/modprobe.  Funky interactions
between modutils and module-init-tools I haven't finished tracking
down yet.  Nothing in the BTS looks obvious, though.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Debian Newbie Question on Network Config

[David Z Maze]

"Alberto Tobias" <[EMAIL PROTECTED]> writes:

> 1.  (*) text/plain  ( ) text/html   

ObFormatting: please set your mailer to send plain text only, and wrap
lines at 72 characters.

> I have however one question. I have troubles with my network card. I
> can get it up and running ok, using the tulip drivers from
> scyld.org. I can configure it, add the default routes etc etc. But I
> have to do this everytime I boot the system, because I don't know
> where to add the configuration so it is set up right when it boots.
>
> Where do I do this? Which startup scripts do I need to change?

Edit /etc/network/interfaces.  There's a man page (interfaces(5)), and
a sample file in /usr/share/doc/ifupdown/examples/network-interfaces.gz.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: please help, lost my partition

[David Z Maze]

LeVA <[EMAIL PROTECTED]> writes:

> I lost my partition. One of my most important ones...  It is a 100gb
> partition with all my personal datas, emails, documents etc...

You might be better off, given what you describe, getting a new hard
drive and then restoring the data from backups.

> I had a crash in kde, and I had to reset the computer. After the boot, the
> boot process said that it can not mount /dev/hda2. Unfortunatelly, I can not
> write you exactly the output, because I can't copy/paste, so I have just my
> brain to remember the lines.
> It says Can't find ext3 filesystem on dev ide0(3,2).
> mount: wrong fs type, bad option, bad superblock on /dev/hda2, 
> or too many mounted file systems.

"It broke."  This "shouldn't" happen if you're using ext3; if the
system spontaneously crashed, the filesystem should be more-or-less okay.

> But then probably I made a huge mistake: typed fsck /dev/hda1. It
> asked me a lot (realy lot), so I ctrl+c'd, and typed fsck -y
> /dev/hda1. It worked a lot (20-30mins), and wrote a lot of things,
> like bad imagic number (or like that), and wrong inode etc... but
> the question was always the same: Clear?. I couldn't choose. Yes,
> clear... After fsck finished, I had a still unusable partition,
> could not mount it, and couldn't fix it, because after that, every
> time I typed fsck /dev/hda2, it said can not find superblock,
> specify another superblock with -B option. I tried a lot of number
> but neither of them worked. So I lost that partition.

When this happened on my laptop, I really strongly suspected bad
hardware.  I eventually came to the conclusion that, sometimes, bit 2
on the IDE channel was getting set to zero, regardless of what it
should have been, and this was somewhat unsurprisingly nuking
important filesystem data.  Given that you've had similar problems
with this hard drive before, I'd really strongly suspect hardware
issues.

> When I type cfdisk /dev/hda it can not detect that /dev/hda2 is an
> ext3 fs, so under the FS Type column it writes Linux, instead of
> Linux ext3.

Linux pretty much ignores the filesystem-type word in the partition
table.  Also, ext3 *is* ext2, just with an added journal file.  So
this isn't really indicative of anything; you'd get the same results
on a healthy system.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: failing to upgrade sysvinit from Knoppix version

[David Z Maze]

Simon Tod <[EMAIL PROTECTED]> writes:

> David Z Maze <[EMAIL PROTECTED]> writes:
>
>> Knowing what exactly the version is might be helpful.  If the
>> Knoppix people have added an epoch to their version number, APT
>> would be entirely correct in concluding that 1:2.84-mumble is newer
>> than 2.85-7.  You might also check that your APT sources.list is
>> correct, and that you haven't set up funny APT pinning rules.
>> 
>
> There's nothing wrong with my sources.list, it points
> to sid only. I'm not using any apt pinning rules.
>
> The problem is, as you guessed, the epoch number.
> Version 2:2.84-161.
>
> So, any ideas? How do I "downgrade" to the more up to
> date version in sid?

I'm not sure that I can think of a better way than downloading the
packages by hand and installing them directly using dpkg.  :-/  Can
you readily identify the packages that are different versions than
what's in unstable?  You can directly download the packages from your
favorite Debian mirror; this is easier if there are a small number of
them, of course.  You also might be able to set up your APT
preferences file; apt_preferences(5) says that priorities over 1000
will causes packages to be downgraded.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Q: Why is linux-wlan-ng not an ordinary part of the kernel?

[David Z Maze]

Mariano Kamp <[EMAIL PROTECTED]> writes:

> I believe to have a prism wlan card in my notebook and while digging
> around what to do I came to the conclusion that I need a module,
> which is in linx-wlan-ng?!

Not necessarily.  I believe there are also working drivers in the
kernel, and in the separate PCMCIA driver pack; look for "Hermes"
driver support.

> It is a kernel module and part of the official debian distro, isn't
> it? My expectation had been to see this turn up in menuconfig, but
> it doesn't. There also isn't a precompiled module for the current
> 2.4.22 kernel in the repository.

"Part of the distribution" doesn't necessarily mean "part of the
official kernel source".  For linux-wlan-ng, you actually need to get
the source to the package, put it in /usr/src/modules, and use
kernel-package to build the modules (if you're building your own
kernel, which it sounds like you are).  See
http://newbiedoc.sourceforge.net/system/kernel-pkg.html for more
information on building kernels "the Debian way".

If you're using Debian stable, there isn't anything official for
kernel 2.4.22 at all.  If you're using unstable, it looks like the
maintainer hasn't built modules for that kernel yet; you might look
through the list of relevant bugs in the BTS and file a wishlist bug
if there isn't one yet.

> Is there any particular reason for this or is it just, that the
> module somehow is prevented from becoming a *first class* citizen of
> the kernel?

"The upstream maintainer hasn't convinced the upstream kernel
maintainers to include it."

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Driver installation

[David Z Maze]

"Hoyt Bailey" <[EMAIL PROTECTED]> writes:

> I have the driver and kernel headers on a CD.  I assume installation
> is via apt-get install xx Ok what should xx be.  Where
> should I put the files to install them.

Having no other details, if you just have a pile of .deb files
somewhere, you can install them directly using 'dpkg --install'.  They
don't need to live anywhere in particular or have any particular
directory structure.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: debian-installer-demo not working

[David Z Maze]

Andrea Tasso <[EMAIL PROTECTED]> writes:

> hi all, I tried debian-installer-demo, but it does not work, it
> stops just after the language selection screen, I mean the next one
> opens, but when you ask to load modules, that is the only way to go
> on, it does not.

The Debian bug-tracking system is a far better place to complain about
specific somewhat obscure packages that have just made it into
unstable, especially where the package description especially says
it's good to "test or debug" things.  See http://bugs.debian.org/ if
you haven't yet.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Restarting killed processes

[David Z Maze]

"BruceG" <[EMAIL PROTECTED]> writes:

>> Ainsi parla BruceG le 304ème jour de l'an 2003:
>>
>>>I recently installed sendmail / ipopd / apache /squirrelmail to
>>>make a SMTP/POP mail server with a Web interface. I'm running
>>>Debian Stable. My PC is kind of clunky and old (100 Mhz, 16Meg RAM,
>>>1 Gig disk space). I've noticed that some processes stop overnight.
>>>Apache, mandb, klogd,... - mainly it's Apache that stops.
>>
>> Do they _really_ stop, or only logging ? if it's this one, man
>> logrotate may help you
>
> Yeah, they stop. I was on a different PC and tried to connect to my web
> server. Couldn't connect. I went to the server and saw the message "VM:
> killing process Apache" on the console. I tried wading through the logs,
> but they're a little difficult for a newbie to understand.

That sounds like your system is running out of memory; 16 MB of RAM is
pretty tight.  How much swap do you have?  (Can you scavenge more
physical RAM?)  A kernel message starting with "VM" probably pertains
to the virtual memory subsystem, which is consistent with you running
out of memory.  Running 'free' might also help you look at what the
memory situation is.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: print command

[David Z Maze]

Vivek Kumar <[EMAIL PROTECTED]> writes:

> Is there any other command to print any character say "*" 80 times..
>
> like echo "**"
> (In bsh  or ksh)
>
> Is there any short command ??

Depending on what you're actually trying to do; Perl is the big hammer
you can throw at anything, of course.

perl -e 'print "*" x 80 . "\n"'

If you're just writing a script, I'd just hard-code the separator (or
maybe do without entirely).

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: print command

[David Z Maze]

[EMAIL PROTECTED] writes:

> Quoting Bijan Soleymani <[EMAIL PROTECTED]>:
>
>> In perl you could do:
>> perl -e 'for(1..80){print "*";}print "\n";'
>
> Can you just explain how to use this for ? It seems far away the ones I know
> (C,C++,basic,Java,php,etc.)

That invocation happens to do the perlish thing of setting $_ to each
of 1, 2, 3, ..., 80 in turn.  Perl also supports the C-like for, but
that's not what you'd use in a one-liner.

For gratuitousness, in a completely different style:

(define (star-string n)
  (define (star-helper n s)
(if (= n 0) s
  (star-helper (- n 1) (concat "*" s
  (star-helper n ""))

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: stability of libc6-i686?

[David Z Maze]

csj <[EMAIL PROTECTED]> writes:

> On Fri, 31 Oct 2003 15:33:34 -0700,
> Monique Y. Herman wrote:
>> 
>> Unstable has a new package available: libc6-i686.  Apparently
>> libc6 optimized for the 686 architecture.  Now, this sounds
>> attractive to me, but the package warns of commercial apps
>> potentially blowing chunks.  IBM's jdk is specifically called
>> out for this.
>> 
>> Does it seem fairly safe to go ahead and install this?  If the
>> problem is only with commercial apps, it shouldn't crash my
>> system or anything, right?
>
> To add to the question: Is it safe to build this (and still be
> able to build third-party free or almost free software like
> MPlayer or the DRI XF86 fork)?  I see a massive build-dependency
> on something called linux-kernel-headers.

You might see the discussion on debian-devel about this.  (Quick
summary: the "dependency" has always been there, but now the
kernel-headers-used-by-libc are in a separate package.)  Reading
debian-devel (or at least skimming it) isn't a bad idea if you're
tracking unstable.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Howto remove grub?

[David Z Maze]

"W. Borgert" <[EMAIL PROTECTED]> writes:

> I have a machine where I want to cleanly install Debian, but the
> MBR has remains of grub.  I searched the net, but all hits I got
> involved a DOS diskette or similar.  I don't have such a disk,
> but I can boot the machine using Knoppix or other Debian rescue
> CDs.  What is the magic command line?  Thanks in advance!

What you want to do is not so much "remove grub" as "install some
other MBR".  You could install LILO, GRUB, or the MBR from the 'mbr'
package to get a new MBR.  I think the Debian installer will also
offer to install LILO for you at the end of its install sequence,
which will overwrite your GRUB bits.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: setting hardware clock from NIST

[David Z Maze]

[EMAIL PROTECTED] (Haines Brown) writes:

> I have an executable script, "time.rc" which has: 
>
>   #! /bin/bash
>   rdate -s time-b.nist.gov 
>   clock -w

It's almost certainly better to find a local time server and not
hammer on the NIST's; I'd also use ntp (ntp-simple package) to keep
your clock up-to-date while the system is running.  See
http://www.ntp.org/ for more information, and contact your ISP to see
if they have their own time server.

At any rate, on my system, the hardware clock is automatically updated
from the system clock at boot and shutdown time, by
/etc/init.d/hwclock.sh in the util-linux package.  So if you installed
and configured NTP, you'd get the same effect as this script.

> Second, where to put it? I placed a copy of my time.rc into
> /etc/init.d, and then created a symlink to it in /etc/rc2.d so that
> the hardware clock is reset on boot, and also in /etc/cron.daily, so
> that the clocks are reset daily according to NIST. Will this work; is
> there a better arrangement?

That setup is probably fine, though I'd do either a cron script or an
init.d script, not both (if your machine spends a lot of time shut
down, anacron can run delayed cron jobs at boot time).  If you do want
an init script, I'd also make it more policy-compliant; try working
from /etc/init.d/skeleton.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: kaffe and/or sablevm in mozilla?

[David Z Maze]

jjluza <[EMAIL PROTECTED]> writes:

> like it is said in the doc, these packages (free implementation), make
> you to be able to compile and run java program.
> But they don't work to browse Internet and its java applets.
> You need a closed source one to do that (sun, ibm or blackdown one)

Wow, that's pretty impressive FUD.

There are two technical obstacles I can think of to using a free JVM
as a Mozilla plugin.  One is actually writing the plugin, which is
probably the easier part.  The second is an implementation of the Java
class libraries that actually supports things like the Java AWT ("GUI
stuff"); my impression is that the only Java standard library
implementation out there is GNU Classpath, and they're strictly
text-only.  (So you could, in theory, write a DFSG-JVM Mozilla plugin,
but it'd be useless without the non-free class libraries.)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: new user question about stable branch

[David Z Maze]

"Chris Ochs" <[EMAIL PROTECTED]> writes:

> Is the stable branch frozen in place except for security/bug fixes from the
> time it was released?

Yes.

> I installed woody and then upgraded to kernel 2.4.18, which made me
> think what other packages are update from time to time.

In the particular case of the kernel, the 2.4.18 kernel was
distributed with Debian 3.0 ("woody"), but it was still a little fresh
to be considered for the default kernel.

> Also, I'm assuming that running woody is the best bet for mission critical
> stuff?

This is probably the use case that stable is most intended for, yeah.
People running desktop machines seem to be frequently frustrated that
stable has "old" packages, but if you don't want to be running the
GNOME-of-the-week because that server really really needs to be up,
woody is a pretty good call.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: setting hardware clock from NIST

[David Z Maze]

[EMAIL PROTECTED] (Haines Brown) writes:

>> From: David Z Maze <[EMAIL PROTECTED]>
>
>> It's almost certainly better to find a local time server and not
>> hammer on the NIST's; I'd also use ntp (ntp-simple package) to keep
>> your clock up-to-date while the system is running.  
...
> Thanks for the advice. I normally don't reboot for months at a time,
> and so need to sync clocks with cron. I'll kill the symlink in
> /etc/rc2.d. I assume that /usr/sbin/ntpd is the executable that I
> should symlink in /etc/cron.daily.

If you install ntp-simple it will start a daemon that will
periodically poll the time servers and gently keep your clock in
sync.  (If you're five seconds off, that time will be made up
gradually, rather than abruptly shifting the clock.)  No need to set
up a cron job.

> In reading the doc, I see that ntpq, run without argument, is a way
> to do a simple test of ntp functionality. However, that does not
> seem to be part of the ntp-simple package, nor is it itself a
> package. Have you used ntpq, and if so, how?

It's in the base ntp package, which is suggested by ntp-simple.  For
looking at my own ntp daemon, 'ntpdc -s' will give you a short summary
of what other time daemons you're talking to, and 'ntptrace' will show
the synchronization chain from yourself to a stratum-1 time server.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Can't build X (unstable and experimental)

[David Z Maze]

csj <[EMAIL PROTECTED]> writes:

> I can't build either the unstable or experimental versions of
> Debian's xfree86 packages (4.2 and 4.3).  The build ends with the
> following error messages:

Why are you building X?  Which X?  And how?

My general recommendation, if you need XFree86 4.3 for hardware
support, is to download the binary .tar.gz files for the X server only
from xfree86.org, and unpack them somewhere like /usr/local where the
Debian package management system won't step on them.  Configure this X
and set the /etc/X11/X symlink to point to /usr/local/bin/XFree86.
IIRC, the two tarballs you want are Xxserv.tgz and Xmod.tgz.

If you don't want to do that, there are several backports of XFree86
4.3 out there to your favorite (stable-or-newer) Debian distribution;
see http://www.apt-get.org/ for details.

> #BEGIN STDERR
> lnx_io.c: In function `KDKBDREP_ioctl_ok':
> lnx_io.c:90: error: structure has no member named `rate'
> lnx_io.c:98: error: structure has no member named `rate'
> lnx_io.c:100: error: structure has no member named `rate'
> lnx_io.c:101: error: structure has no member named `rate'
> lnx_io.c:102: error: structure has no member named `rate'

...but if you really did want to track this down, look at the
referenced line number, find the type of the structure that's being
referenced, and figure out where that structure comes from.  If it
randomly started losing on unstable in the past week, it's possible
that the source is directly depending on a  header,
but those headers changed from a 2.4.mumble kernel to a 2.6.0.mumble
kernel recently.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Java on linux

[David Z Maze]

"Hoai Nguyen" <[EMAIL PROTECTED]> writes:

> Do you know any java API
> for linux that handles mount and umount cdrom.

Please set your mailer to send in plain text only, no HTML, and wrap
lines at 72 columns...

...but if I needed to do this, I'd use Runtime.exec() to shell out to
the normal Unix 'mount' and 'umount' commands.  Which is starting to
require special permissions (either root permissions or 'user' mount
permission on the device in /etc/fstab), and you need to know the path
anyways.  What are you really doing, and why are you doing it in Java?

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: apache-ssl or libapache-mod-ssl ?

[David Z Maze]

"Lynn W." <[EMAIL PROTECTED]> writes:

> My question is, given my httpd.conf settings for PHP4 and ColdFusion,
> should I apt-get apache-ssl or libapache-mod-ssl?

Yes, one of those is probably what you want.  :-)  You might look at
the respective upstream Web pages (http://www.apache-ssl.org/,
http://www.modssl.org/), which mention some of the differences, but
they're essentially both different ways of doing the same thing.

My personal opinion is that the separate module (e.g., mod_ssl) is a
little cleaner than patching the Apache source to get SSL support.
I've also found mod_ssl to be very slightly more flexible in one
corner case (in particular, I can get either certificate or password
authentication under mod_ssl, but nobody in the real world actually
uses personal certificates).  My work actually uses apache-ssl, and it
works fine.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Installing packages from source with apt

[David Z Maze]

Vincent Lefevre <[EMAIL PROTECTED]> writes:

> What is the best way to compile and install packages from source with
> apt?
>
> I don't want to use "dpkg -i" once the packages are built (as suggested
> in the how-to) since dpkg doesn't check dependencies and may break the
> system (it did in the past...).

dpkg does *check* dependencies, it just doesn't go out of its way to
*correct* them.  That is, dpkg shouldn't let you install a package if
its dependencies aren't already installed.  If you never use a --force
option and your packages work, it should be pretty hard to break your
system using 'dpkg --install'.

One useful alternative might be to install the package using APT (so
you pick up the dependencies), then recompile it from source and
install it using 'dpkg --install'.

> Vincent Lefèvre <[EMAIL PROTECTED]> - Web:  - 100%
> validated (X)HTML - Acorn Risc PC, Yellow Pig 17, Championnat International

(I was at HCSSiM in 17*117+4...  :-)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Using dselect and apt - urgent

[David Z Maze]

"Mike Simpson" <[EMAIL PROTECTED]> writes:

> 1.  (*) text/plain  ( ) text/html   

(Please set your mailer to post in plain text only, and wrap lines at
72 characters.  And don't include the word "urgent" in your subject
line; everyone's question is "urgent" in some form or another.)

> We have our website (aidu.mod.uk) now on our own webserver for the
> past 8 months.  I use dselect to get updates for our system, however
> in all the months we have not had an update.

What APT sources are you using?  Do you have the security.debian.org
line in it?  (See http://www.debian.org/security/.)  Otherwise, you'll
only get updates when a new point release of Debian stable is
released, which is fairly infrequent.

> I downloaded from the debian site info on dselect and noticed the
> bit on proxy servers, thinking that this is preventing the update
> from working I entered the following command to set the environment
> variables.

Is your site actually using a proxy server?  If dselect was working
before (you could update and see a list of packages without getting
errors) then you're probably not, and you don't need to set this.  If
you are, you should be able to get that information from your network
administrator.

> I now think I have made a mistake by setting this variable and
> probable need to reset it.
>
> Q1 what do I reset it to and how?

'unset http_proxy' should do it, as would logging out and logging in
again.

> I use the default address to get the updates from
> http://http.us.debian/org/debian stable main contrib non-free

Are you actually in the US?  Your email address suggests not; you
might get better performance by using a nearer mirror, like

  deb http://www.uk.debian.org/debian stable main

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: a few Qs about debian's apt

[David Z Maze]

"Woon Wai Keen @ doubleukay.com" <[EMAIL PROTECTED]> writes:

> a friend of mine has some questions regarding debian. hope you guys could
> help me answer them :)

I notice you're asking a lot of questions about apt-get.  It often can
be a little difficult to figure out what apt-get is doing; a
higher-level package manager, such as aptitude, can often be more
informative.

> 1) does 'apt-get upgrade' upgrades:
> i) the kernel,
> ii) base apps
> iii) local apps (/usr/local)

(i) Only in very limited ways, for certain minor updates, but often
not within a particular kernel revision, and never between kernel
versions.  (ii) Yes, always "base apps", however you define those.
(iii) No, Debian packages always leave /usr/local untouched (with the
exception of possibly creating empty directories for user extension
packages).

> 2) where does apt-get saves all it package information?

In a combination of /var/lib/dpkg (used by dpkg), /var/cache/apt (for
downloaded packages), and /var/lib/apt.  But you should (almost) never
need to look at these directories directly.

> 3) is there a way to just upgrade the local apps instead of all
> local/kernel/base at the same time? if so what is the apt-get
> argument?

APT and dpkg never touch things the administrator has installed in
/usr/local, or things users have compiled in their home directories.
In other words, it never "upgrades the local apps", as I understand
your terminology.

> 4) what do we do if we need to synchronize the package information
> manually, say for some reason apt-get fails to include version
> information on newly installed package; it's still using the old
> version although the package has been overwritten by the latest
> version?

Um, what do you mean by "overwritten by the latest version"?  Why do
you think a newer version is installed than what APT or dpkg think is
installed?

> how do debian define non-base apps? in the bsds, non-base apps which is
> called local apps are those not part of the vendor-approved base
> distribution.

Aah.  Debian doesn't have that distinction.  Everything that's "part
of Debian" and included on the official CD is built to install in
/usr.  Some packages are flagged "essential" or "section: base", but
these are mostly artificial distinctions that have minor importance in
the packaging system.

> getting back to debian, say for an application that is not part of
> debian base distribution, how do we go about getting apt-get to
> upgrade them, or does debian does not segregate the definitions of
> base/local apps?

If something's not on any of the Debian CDs, it might be in a newer
version of Debian ("testing" or "unstable"), or there might be
unofficial packages of it.  Otherwise, you get to compile it from
source, and APT will have nothing to do with it.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: nice

[David Z Maze]

"Hoyt Bailey" <[EMAIL PROTECTED]> writes:

> There was some discussion about nice -10 a few days ago and the question was
> how to set nice.  The command is nice -10 .
> Reference debian "reference".
> In Linux:
> nice:  Range -20 (Not nice) to 19 (Very nice)
> In English:
> Priority level: 1 to 39.

That's only sort of correct.  The default niceness is 0, and you can't
get negative niceness without being root.  The "priority" that top,
ps, etc. report is only sort of related to niceness; it's a number
that comes out of the kernel process scheduler.  But otherwise, yes,
lower (even negative) niceness implies higher priority.  You can also
use renice(8) to change the niceness of an existing process.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: update-rc.d

[David Z Maze]

Miguel Alvarez Blanco <[EMAIL PROTECTED]> writes:

> a) if the sysadmin wants to fiddle with the links, he may do so
> (presumably by hand?) and the packaging system will not touch them
> *provided the sysadmin leaves at least one of the links*. Now, I can
> see how to use this to remove all but an unreachable link (i.e., a
> K99ipmasq link in an unused runlevel) so that the system does not
> touch them, but I do think that it is very ugly, tricking the system
> and not fixing it.

You probably want the service to stop on shutdown if it's running,
right?  Then leaving /etc/rc0.d/K99ipmasq and /etc/rc6.d/K99/ipmasq is
consistent with what you want, and still lets update-rc.d believe the
package is "installed".

IMHO the Debian system makes a lot of sense here, though it's not
perfect.  Everybody knows how to use 'rm' and 'ln -s', and publishing
those as The Official Way To Tweak Runlevels makes things easy for
sysadmins; I don't know how to use update-rc.d (or, on RH, chkconfig)
without reading the man page.  OTOH, it does kind of seem like it'd be
nice to be do things like easily configure a package to never start a
service ('rm /etc/rc?.d/S??service') or recreate the symlinks at the
right sequence number; I know RH's chkconfig can do the last one.
Probably a good opportunity to write code to fix the UI problem and
submit it as a wishlist bug against sysvinit.  :-)

> b) update-rc.d will remove its links if requested and the script is
> not there (well, or if -f is used as I do), if its first argument is
> purge, so that the user has requested the configuration to be
> removed.

That invocation is intended for the *package* to request that the
configuration be removed.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Portable shell scripts

[David Z Maze]

csj <[EMAIL PROTECTED]> writes:

> I tend to write scripts which are tcsh-compatible.  So
> "#!/bin/tcsh".  But its somewhat a waste of effort to write one
> set of scripts for bash and another for tcsh.  My main problem is
> handling the variables.  Is there a shell-portable way to specify
> variables?

...to what end?  I'm confused here: if you set variables inside the
script, they won't be visible to the parent of the script, but they
generally will be visible to children.

  #!/bin/tcsh
  setenv TESTVAR hi
  /bin/sh -c 'echo $TESTVAR'

  #!/bin/sh
  TESTVAR=hi
  export TESTVAR
  /bin/tcsh -c 'echo $TESTVAR'

will both print "hi" because that's the string being passed through
the environment; it doesn't care what shells are (or aren't) involved.

> I write mostly convenience scripts that are generally less than a
> console screen in length.

So do I, but that doesn't stop me from wanting to do

  make 2>/dev/null

occasionally.  :-)  Using a Bourne-like interactive shell is also
useful, since I wind up typing a lot of one-liners that might include
shell loops.  (I personally use zsh, but I think bash has about the
same feature set these days; at least, it has programmable completion,
which is the reason I originally switched to zsh.)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Debian version

[David Z Maze]

Drew <[EMAIL PROTECTED]> writes:

> I am trying to figure out what version of debian I am
> running.  /etc/issue* and /etc/debian_version all
> state unstable/testing, but I don't know when this box
> was installed.  This box could be potato, from when
> potato was unstable, but how to tell? 
>
> I am attempting to determine what release this box is
> slink, potato, woody, sarge? 

One thing you might try doing is looking at the version of the libc6
package, which has pretty reliably changed between releases, and
compare it to the version in the various releases.  I don't entirely
remember when testing appeared but I'm pretty sure that if your
/etc/debian_version mentions testing then you're post-slink.  :-)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: wireless LAN in place of existing cabled one

[David Z Maze]

"Benedict Verheyen" <[EMAIL PROTECTED]> writes:

> my current LAN looks like this:
>
> cable - eth0 (public ip) -server
> modemeth1 (192.168.0.1)
>   |
>hub
>   |
>   |
>pc 1
>
> My server runs dhcp, apache, exim, fetchmail, webmail and so on.
> Now the wife is fed up with the cable running through our living room
> up the stairs to my room where the server, the hub and pc1 are.
> Now we (she) wants to go wireless. I asked a local dealer and he
> works with D-Link equipment more specifically the Di-714P+ or
> the Di-614+. This would be the future setup:
>
> cablemodem --- router --wireless-- server -- hub -- pc 1
> |
> -wireless-- clients

It looks like your local dealer is selling you the wrong product.  :-)
You don't want a "wireless router"; in your situation, you want a
bare-bones "wireless access point".  So my home network looks like:

cable --- server --- hub --> pc 1
  +  --> pc 2
  \  --> wireless
   \  --> laptop

I can't really recommend any specific products in this space, but they
definitely do exist.  If you've already bought it, you might see
whether you can turn off ~all of its features (it's probably set up as
a NAT box that does DHCP on the internal side) and deploy it in this
position on your network.

> 1. The server acts as a gateway now where eth0 is an ip from my
> isp and eth1 is a fixed internal ip where a DHCP daemon is listening
> to distribute ip's to the clients (currently pc1 but 1 other pc will
> follow and will be placed downstairs). Now i think i can still use
> the server as gateway with the new setup but i will not be able to
> secure the LAN with the firewall script that runs on it, correct?

In the diagram you have, "clients" gateway would be "router"; "hub"
wouldn't be used at all, except by "pc 1".

> 3. Is  the network traffic encrypted by default?

Probably not.  You can almost certainly enable WEP on the AP and on
the client machines, and with a sufficiently reputable product you
could probably also enable a MAC ACL ("limit which hardware addresses
can talk to the wireless network").

[My impression is that these are both useless steps, though: WEP is
almost trivially cracked, and many network cards let you change their
MAC address in software.  So you still want to make sure that machines
inside the network are secure, and use end-to-end security if you're
transmitting sensitive data, even within the local net.]

> 4. What kernel options do i have to activate to be able to use a
> wireless usb card (DWL-120+) . Usb is already compiled in. I'm
> not even sure these will function under Linux. Any place i can
> find out?

I thought "USB network interface" was actually a somewhat standard
thing, so you might not need special drivers.  In any case, looking on
http://www.linux-usb.org/ does seem to be informative, and support
does seem to exist for that particular card.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: madwifi .deb package

[David Z Maze]

"Jeffrey L. Taylor" <[EMAIL PROTECTED]> writes:

> Anyone know where I can download a .deb package for the madwifi
> drivers?  I don't really have room on this laptop (540MB HDD) to keep
> the kernel sources that it wants around.

I've thought about building a package, but the source is non-free and
so it would never be able to be a part of Debian proper.  You'd also
need a package specific to your kernel; if I built one, it'd only
support the kernel-image-2.4.22-1 in unstable (or possibly have no
prebuilt modules at all; many kernel module packages work that way).
You might try building the driver on another machine that has the
source for the kernel on (presumably) your laptop, or seeing if you
could get away from building the modules against only a kernel-headers
package.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Stock 2.4.16 kernel, initrd and ext3

[David Z Maze]

ScruLoose <[EMAIL PROTECTED]> writes:

> I've been wanting to switch from ext2 to ext3 on this machine, including
> the root filesystem.
>
> I'm using the stock Debian 2.4.16-i686 kernel, which (according to
> /boot/config-2.4.16-686) has ext3 support _as a module_...

2.4.*16*?  That's really old; even woody shipped with a 2.4.18 kernel
(though not as the default kernel).

> Now, my understanding is that this kernel uses initrd, and thus it'll be
> okay for me to switch my root fs to ext3 (it'll load initrd first,
> notice that it needs ext3, ins the mod, mount the root partition, then
> carry on booting)...
>
> 1)  confirm/deny?

That sounds correct.  (An initrd-based kernel will probably have
almost nothing built in as a driver, and everything will be as
modules, which get loaded from the initrd.)

> 2)  Do I need to modconf the ext3 module in order for this to work?

You shouldn't need to.

> 3)  Once I've done tune2fs -j and tweaked my fstab, are there any other
> steps I'm missing before I reboot?

No, I think that's about it.  ext3 is convenient that way.

> 4)  How do I find out for myself whether my current kernel is using
> initrd?

Probably look at your bootloader configuration; there should be a
mention of /initrd.img or /boot/initrd-$KVERS.img.

> Normally I'd just try it and see if it worked, but leaving the system
> un-bootable and having to dig out the rescue disk is such a pain... 

With ext3 you also have the bonus that the filesystem is still a
mostly-valid ext2 filesystem.  And all of the modern rescue media I've
seen (not the woody installer, but things like SysRescueCD and
Knoppix) support ext3.  So you shouldn't be able to hose yourself too
badly doing this.  :-)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: kernel-headers & compiling source

[David Z Maze]

Jeff Penn <[EMAIL PROTECTED]> writes:
> I have read through the kernel-header docs, & am still not sure I 
> understand what they are for.  I assumed that they enable source to be 
> compiled when using a kernel-image.
>
> If this is correct, what is the procedure for compiling i2c-source or 
> lm-sensors-source?.

If you're compiling your own kernel, you should read those packages'
respective README.Debian files, which give instructions.
(Essentially, after you 'make-kpkg kernel-image', 'make-kpkg
modules-image'.)  If you're using one of the stock kernels, you may be
comparatively out of luck; I'm trying to get packages into unstable
that contain precompiled modules for the provided 2.4.20 kernels.
*glances at ftpmaster*

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: kernel upgrade to 2.4.20-k7

[David Z Maze]

Fraser Campbell <[EMAIL PROTECTED]> writes:

> No need to wait.  I'm running kernel-image-2.4.20-k7 on my Athlon.  You'll 
> only need kernel-headers if you plan on compiling software.

...where "software" specifically means "kernel modules".  I suspect
most people will never have a reason to install a Debian
kernel-headers package.  If you built your kernel from source, you can
use the unpacked source tree for its headers.  If you're not trying to
compile a kernel module that's not packaged for Debian already and
don't have a kernel source tree sitting around, you don't need this
package.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: default login manager

[David Z Maze]

Aryan Ameri <[EMAIL PROTECTED]> writes:
> I have installed Ralph Nolden's KDE 3.1 packages on woody stable.
> However after installing kdm, when starting the computer, xdm is still the
> default login manager. Howshall I change the default login manager to
> kdm?

The most straightforward thing to do would be to remove xdm.  You
could also change the links in /etc/rc?.d to cause one display manager
or the other to start without removing the other one.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: OT: functional languages

[David Z Maze]

[EMAIL PROTECTED] (Robert Land) writes:
> On Fri, Dec 13, 2002 at 11:23:43AM -0500, Derrick 'dman' Hudson wrote:
>> "imperative" and "procedural" are the same thing, and C is a prime
>> example.  It is such because the structure of a C program is a
>> collection of procedures which start with "main".  Each procedure is a
>> linear list of statements to be executed in order.
>
> Could you specify a "linear list" more clearly? - the 
> contrary would be a "nonlinear list" which on the first
> view seems to be self-contradictory.

For example, in C:

int fib(int n)
{
  if (n < 2)
return n;
  return fib(n-2) + fib(n-1);
}

the rules require that fib(n-2) is called before fib(n-1).  In Scheme:

(define (fib n)
  (if (< n 2)
n
(+ (fib (- n 2)) (fib (- n 1)

the recursive function calls can happen in either order, and I believe
it's even semantically correct for e.g. both function call parameters
to be evaluated before either function is called (again, C and ilk
strictly specify this order).

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: default login manager

[David Z Maze]

Oliver Fuchs <[EMAIL PROTECTED]> writes:
> 3) update-rc.d -f xdm remove

This is arguably bad advice; if xdm ever gets updated (as in a
security release), update-rc.d will notice that there are no links for
xdm, conclude that the package was never installed, and recreate the
links, leaving you back where you started.  You need to make sure that
at least one link is left behind (like the K links in runlevels 0 and
6) so this doesn't happen.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: ALSA (and kernel?) problem

[David Z Maze]

George Georgalis <[EMAIL PROTECTED]> writes:
> I haven't used alsa in a while but when I did, I used these scripts (RedHat)
> to compile/install.
>
> http://galis.org/scripts/alsa-INSTALL.sh
> http://galis.org/scripts/alsa-driver-0.5.11.sh
> http://galis.org/scripts/alsa-lib-0.5.10b.sh
> http://galis.org/scripts/alsa-utils-0.5.10.sh

I'd ignore these scripts in a Debian world.  First, I'd use APT to
install the ALSA userspace.  If you're using a stock kernel, there are
probably precompiled ALSA modules to go with it; if not, install
alsa-source, unpack /usr/src/alsa-driver.tar.gz, and then go to the
top directory of your kernel tree and run 'make-kpkg modules-image
--added-modules=alsa-driver'.  (Assuming, of course, that you used
kernel-package to compile your kernel in the first place.)  This
produces an alsa-modules-*.deb in the parent directory of your source
tree; install it using 'dpkg -i'.

(Comment: the same pretty much applies for any add-on kernel module.
Caveat: I only know for sure that modules exist in unstable for 2.4.19
kernels, modules don't necessarily exist for any particular module for
any particular kernel.  But you're golden if you're building your own
kernel.  And you might even be able to use the 0.9.0rc6 ALSA source
out of unstable, with some luck.)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Installing Debian on an iPAQ

[David Z Maze]

"Darryl L. Pierce" <[EMAIL PROTECTED]> writes:
> Anybody have experience doing this? Can anybody point me to some
> information on doing so?

Back when I was playing with one, general advice was that you only
wanted to try to put Debian on it if you had a lot of persistent
storage.  The Familiar distribution was sort of like Debian, and its
package manager could install ARM .deb packages if need be.  At any
rate, the place to go is probably http://www.handhelds.org/.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: OT: functional languages

[David Z Maze]

Eric E Moore <[EMAIL PROTECTED]> writes:
> "Eric G. Miller" <[EMAIL PROTECTED]> writes:
>> In C, statements are executed in order.  I'm not too up on
>> functional languages, but I seem to recall they need special syntax
>> to execute statements sequentially.
>
> Not really.  top level forms in a scheme program are executed
> sequentially, and there's a number of forms that execute their
> statements sequentially.  I'm not up on haskell, or other "pure"
> functional languages though :)

Right, in Scheme it's easy enough to do

  (begin
   (foo)
   (bar)
   (baz))

to cause things to be executed in order.  Haskell doesn't have such a
construct intrinsically.  But what it does have is a concept called a
monad, which allows you to attach state to execution.  Probably the
most common monad is "IO", which reflects the state of input/output
operations (like printing to the display).  The 'do' keyword then
executes statements in order within a monad (doing ugly magic with
lambdas, I think).

I've found that monads are kind of hard to wrap your head around.  I
wrote some monad code once upon a time but don't really remember how
it works; what it's supposed to do is pretty obvious.  :-)  (In
particular, http://web.mit.edu/dmaze/Public/hdc/ has source code for a
non-optimizing compiler for a simple Java-like language to MIPS in
Haskell.  Because you all care, I'm sure, and because it seemed like
more fun than my thesis at the time.  :-)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: IDE for java

[David Z Maze]

Eduardo Gargiulo <[EMAIL PROTECTED]> writes:
> Is there any IDE for java (debianized if possible). Which? Where should
> I point my sources to install it?

I tend to be perfectly happy with Emacs (and in particular I generally
use XEmacs 21).  What features do you want out of it?

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: AudioCD

[David Z Maze]

"Sergey A. Ovchar" <[EMAIL PROTECTED]> writes:
> How can I play the Audio-CD's ?

Use an audio CD player.  gtcd is buried somewhere in the GNOME stuff
and has always worked adequately for me; you might also install the
'cdtool' package, which gives you command-line programs like 'cdplay'
and 'cdstop'.

If it doesn't work, typical things to check are your sound card mixer
settings and whether there's an audio cable from your CD-ROM drive to
your sound card.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Lame,cdparanoia,AidioCD

[David Z Maze]

"Sergey A. Ovchar" <[EMAIL PROTECTED]> writes:
> How can I convert several *.wav to *.mp3, by the _one_ command, using lame. I'm 
>interesting about batch mode.
> Reading This F.. Manual didn't take desired effect :(.
>
> And how can I redirect output trom "cdparanoia -B" to the lame ?

It sounds like you really want abcde, which is a nifty little
console-mode program that rips, encodes, and tags everything off of a
CD.  Also, the standard Ogg Vorbis encoder, oggenc, will take multiple
files on its command line, so 'oggenc *.wav' will produce a directory
full of .ogg files for you.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: IDE for java

[David Z Maze]

"Benedict Verheyen" <[EMAIL PROTECTED]> writes:
> When you say "make", is that the same make program most c++ programs use?
> If so, i didn't know one could use it with other languages. I'll
> have to start learning it then.

make, in general, is good for describing ways of turning one sort of
file into another sort of file.  It happens to be very widely used for
C and C++ programs, but many other languages don't have very nice
properties about what files correspond to.  (Even Java has issues: a
single Java source file can result in multiple class files, and since
there's no source-level separation of interface and implementation,
you might have to simultaneously compile multiple source files with
circular dependencies.)  But there's absolutely nothing
language-specific in make.

> I know there is a java specific make too: jmake.
> Any experience with that one?

None at all; the project I'm working on at work for a long while just
used

  jikes `find $SOURCE_ROOT -name '*.java' -print`

as its build system, and this worked just fine.

I will also mention, as a resource, a Java-based IDE called Eclipse
(http://www.eclipse.org/); I've now used it for under a day, and can
report that it doesn't deal well with local CVS repositories but it
does have an Emacs-like keybinding set and has some nice features like
displaying the first couple of lines of a class's javadoc if you hold
the mouse over a class name in the source, and fairly rapid automatic
recompilation.

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Openafs woes

[David Z Maze]

Andrew Perrin <[EMAIL PROTECTED]> writes:
> I'm running debian woody on a home machine that's behind an NAT
> masquerader (also woody). The home machine runs the OpenAFS client
> to connect to the UNC campus's AFS shared directory space. Generally
> this works fine, but there's one situation that consistently causes
> a problem.

What I've been told around here is that AFS over a NAT will maybe
probably work, if you don't have too much load and don't let it sit
idle too long, and that you can vaguely expect to lose randomly.

> The scenario is this:
> 1.) Cable modem service dies while a file in the AFS space is open
> (usually a perl or latex file in emacs).
> 2.) Cable modem service returns, and IP connectivity is fine (including to
> the AFS server).. BUT
> 3.) attempting to access files and directories in the AFS space results in
> "no such file or directory."

At this point, you might try running 'fs checks' to see if your local
AFS client believes the world exists.  Variations on 'fs flush .'
might help too.

> 4.) tokens reports appropriate kerberos tokens for the user
> 5.) klog lets me create new tokens seamlessly, but none of this allows for
> actually accessing the AFS space.

Being nitpicky, Kerberos tickets, AFS tokens.  :-)  If your site
doesn't use kaserver (UNC it appears does) but you do use krb5, you
need to remember to get addressless tickets with 'kinit -A' before
getting tokens using aklog.  But this doesn't actually apply to you.

> 6.) I can umount /afs (as root) but can't remount it. If I try
> /etc/init.d/openafs-client start, I get "I/O error."  If I try restart,
> the system hangs completely, requiring a cold reboot (power cycle).

Stopping and restarting the AFS subsystem has never worked well for
me.  I'd never try to just unmount /afs, always run
/etc/init.d/openafs-client stop, but there's no guarantee that it'll
happily restart without a reboot.  :-(

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: dhclient-script, debug-enter, debug-exit

[David Z Maze]

Mark Yobb <[EMAIL PROTECTED]> writes:
> 1.  (*) text/plain  ( ) text/html   

(Can you configure your mailer to not send HTML mail to mailing lists,
please?)

> Help somebody?  I'm using the most recent stable version of woody.
>
> /etc/dhcp3/dhclients-script
> /etc/dhcp3/dhclient.conf
> /etc/dhcp3/dhclient-enter-hooks.d/debug-enter
> /etc/dhcp3/dhclient-exit-hooks.d/debug-exit
>
> Are not being installed when I install dhcp3-client.  I have
> uninstalled this package and the tried to reinstall it and the files
> are not showing up.

dpkg tries to make sure that the right things happen with
configuration files; if the administrator deleted a file, dpkg tries
to keep it deleted.  If you're removing and reinstalling the package
anyways, probably the easiest way is to tell dpkg to throw away all of
the configuration files:

  dpkg --purge dhcp3-client
  dpkg --install dhcp3-client_*.deb

Alternatively, you could probably use the --force-confmiss option to
dpkg, but suggesting that someone use a --force option to dpkg feels
dirty.  :-)  (It *is* necessary if you've lost the conffile and the
package isn't one you can remove without breaking dependencies/your
system; it might be nice if dpkg made this option a little less
hidden, or even if APT exposed it.)

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: sed

[David Z Maze]

(Please send a new mail to debian-user if you're talking about
something new, rather than replying to a message you're not actually
replying to...)

"behapy" <[EMAIL PROTECTED]> writes:
> kdjwiskjkdf+-www.kde.org+-333.kjkd.html
> kdjfwiwji+-kbs.co.kr+-cgi-bin+-kkk.cgi
> kdjfwiwji+-kbs.co.kr+-cgi-bin+-kk2.cgi
>
> =>
>
> www.kde.org+-333.kjkd.html
> kbs.co.kr+-cgi-bin+-kkk.cgi
> kbs.co.kr+-cgi-bin+-kk2.cgi
>
> I'd like to remove the first words everyliles ~+-
>
> sed -e 's/^*+-//g' SOMEFILE.TXT don't work.
> How do I do?

If I were sed, I'd parse that as ^ (start-of-line) *+ (one or more
asterisks) - (hyphen), which isn't what you want.  You really want
start-of-line, any number of characters which aren't +, +, -; this
translates in sed-ese to

  sed -e 's/^[^+]*+-//g' file.txt

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: make-kpkg install question

[David Z Maze]

"Ramsay D. Seielstad" <[EMAIL PROTECTED]> writes:
> Hello all, between following the list and reading the various man
> pages and documentation I'm about ready to try my hand at rolling my
> own kernel.  I'm planning on using 'make-kpkg' to do this and the 
> one item I have not seen (or found an answer for) is an option for
> installation into the directories using the version revision.

I'm a little unclear what you're asking...

> For example, I'm currently running 2.2.19, the kernel and related 
> files are located in /boot/* (vmlinuz has a link to 
> /boot/vmlinuz-2.2.19) and installed modules are located in
> /lib/modules/2.2.19/* 
>
> Is there a specific environment variable or command line option
> to set for this, or is that the default action?

...but this is the default you get if you use make-kpkg normally.  You
can change the version number of the kernel using the
--append-to-version option if you want to;

  make-kpkg --revision=1 --append-to-version=-watertown kernel-image

will generate a kernel-image-$KVERS-watertown package where 'uname -r'
returns $KVERS-watertown, the kernel is in
/boot/vmlinuz-$KVERS-watertown, and modules are in
/lib/modules/$KVERS-watertown (where $KVERS is the version of the
kernel, e.g. 2.4.20).

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Mounting partitions HELP

[David Z Maze]

Joris Huizer <[EMAIL PROTECTED]> writes:
> 1. How can I mount the /dev/hda ( the bootloader says
> /dev/hda1 ) (as root and/or as normal user) ?

See mount(8) and fstab(5).

> 2. Is it possible to clean ("format", initialize) the
> /dev/hdb3 and set a new partition ( /tmp) there ?

Create a new filesystem using e.g. mke2fs(8), and mount it there as
before.  Be aware that mounting the new filesystem will mask the
preexisting contents of it; this isn't a problem if you're putting the
new filesystem on /tmp and remounting it by rebooting, but you need to
go through a little more caution if you're trying to create a new
partition for something that already exists and has persistent state
(e.g. /var).  Also, you get into big trouble if your /tmp partition is
too small (well, and /var too, but "APT stops working" is less bad
than the trouble you get if /tmp fills).

-- 
David Maze [EMAIL PROTECTED]  http://people.debian.org/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
-- Abra Mitchell


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED] 
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]