Re: 1.2 installation, more notes
> o I'm having problems recompiling the kernel. Compilation and module > installation proceeds smoothly (using kernel-sources-2.0.27_1.00 ), but > at boot time the slhc, slip and ppp modules I selected (and I suspect the > rest in that category) are not recognized/loaded. Even if I tell > make config to include slip/ppp/slhc in the kernel instead of as modules, > I cannot use slip or ppp. I noticed these modules were advertised as > being (NEW!) on the first compile. > The kludge I used was to compile the kernel with my specifications, but > use the /lib/modules/2.0.27-old directory instead of the one installed by > the make modules_install. This might be the same one I noticed some time ago (older kernel). When you compile a new kernel with the same number (2.0.27 in this case), the new modules go to the same modules directory. However, the installation does not remove the old modules from the same directory. If you compile some module into the new kernel, the same module is both in kernel and in modules. It seems that the kernel gets confused in this case. Esa --------- Esa Turtiainen http://www.turtiainen.dna.fi [EMAIL PROTECTED] fax: +358-0-801 3008 -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED]
findutils_4.1-12 makes /var/lib/locate owned by root.root
It seems that in new Debian installations /var/lib/locate is owned bu root.root instead of nobody.nogroup and that is why locate does not work. Cron routine can not update the database because it is run as nobody. I tried to trace the bug and it seems that in the source package the directory is installed OK but some later makefile changed the permissions again. Esa - Esa Turtiainen http://www.turtiainen.dna.fi [EMAIL PROTECTED] fax: +358-0-801 3008 -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED]
Repost: some experiences of a Rex installation
Did this ever made to the list. I have got some confusing automatic messages. Esa I have installed two machines from the current Rex distribution. The distribution was aligned on Monday to my mirror FTP.FUNET.FI that is usually quite up-to-date. There has not been modifications in the mirror this week any more and so I hope to have the final one. The base set of the system was installed last week with previous boot set (28-11). I had some problems but I uderstood that they should be solved by now. I found the following dependency problems: - gs-aladdin should depend on libpng1 (shouldn't the non-free be frozen with rex?) - kbd recommends too strongly on svgatextmode there are all kind of problems with svgatextmode especially in internationalization and I do not want to use it yet (forcing kbd without svgatextmode works fine) - mc should depend on libgpm1 And some likely bugs: svgatextmode: 8-bit international ISO Latin 1 characters are displayed incorrectly in all the fonts I tried. X: The keyboard maps of XKB extension of XFree 3.2 are not working really well. If you select Finnish from menu you are not going to get the right map. It seems that if you - run new xf86config (keyboard do not work with old) - but do *not* select keyboard extension in the configuration then you'll get the right keyboard map from the kbd package. (In fact, the resulting keymap is the best I have ever managed to get to work in X with *all* the right dead keys.) I do not have any xmodmap settings. msql: it tries to send messages to the msql user which it has correctly created but forgotten to create /var/spool/mail/msql which means that smail stores the message in 'error' subdirectory. The error message email repeats at 15 second interval. The message says "Can't start server : UNIX bind : Permission denied". The reason may be the permission problems of the '/tmp' being root.root rwxr-xr-x after the base installation. This generates an interesting amount of trash mail files. Updatedb is never run. The likely reson is that cron seems to die. One interesting note is that I must *not* have bind installed to get NFS partitions mounted in the boot. If I have local DNS it is not up yet and the server names are not resolved at all (in a small installation they are in /etc/hosts but I have decided not to do that). If I do not have local bind, it correctly asks tha names from configured external nameservers. BTW: I have tried to use network masks in /etc/export. I have got them to work in *the last line* of the file. If I have three lines in the file, just the third export allows mounts according the mask. BTW2: Are you sure that nameservers should be separated with comma in /etc/resolv.conf like suggested in the boot set dialog. I have used just blank and it seems to work. Never tried comma. I would be happy with CD drivers in the kernel if there would be a way to disable autodetection of the devices that are not there. Is there a boot option to do that? Space is not such a big issue anymore. I have not found xload yet. fvwm95 is missing it. The old configuration file is missing xterm_color as well but it seems to be obsolete now(?) (I prefer this, the green cursor is awful.) Esa --------- Esa Turtiainen http://www.turtiainen.dna.fi [EMAIL PROTECTED] fax: +358-0-801 3008 -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED]
Some experiences of a Rex installation
I have installed two machines from the current Rex distribution. The distribution was aligned on Monday to my mirror FTP.FUNET.FI that is usually quite up-to-date. There has not been modifications in the mirror this week any more and so I hope to have the final one. The base set of the system was installed last week with previous boot set (28-11). I had some problems but I uderstood that they should be solved by now. I found the following dependency problems: - gs-aladdin should depend on libpng1 (shouldn't the non-free be frozen with rex?) - kbd recommends too strongly on svgatextmode there are all kind of problems with svgatextmode especially in internationalization and I do not want to use it yet (forcing kbd without svgatextmode works fine) - mc should depend on libgpm1 And some likely bugs: svgatextmode: 8-bit international ISO Latin 1 characters are displayed incorrectly in all the fonts I tried. X: The keyboard maps of XKB extension of XFree 3.2 are not working really well. If you select Finnish from menu you are not going to get the right map. It seems that if you - run new xf86config (keyboard do not work with old) - but do *not* select keyboard extension in the configuration then you'll get the right keyboard map from the kbd package. (In fact, the resulting keymap is the best I have ever managed to get to work in X with *all* the right dead keys.) I do not have any xmodmap settings. msql: it tries to send messages to the msql user which it has correctly created but forgotten to create /var/spool/mail/msql which means that smail stores the message in 'error' subdirectory. The error message email repeats at 15 second interval. The message says "Can't start server : UNIX bind : Permission denied". The reason may be the permission problems of the '/tmp' being root.root rwxr-xr-x after the base installation. This generates an interesting amount of trash mail files. Updatedb is never run. The likely reson is that cron seems to die. One interesting note is that I must *not* have bind installed to get NFS partitions mounted in the boot. If I have local DNS it is not up yet and the server names are not resolved at all (in a small installation they are in /etc/hosts but I have decided not to do that). If I do not have local bind, it correctly asks tha names from configured external nameservers. BTW: I have tried to use network masks in /etc/export. I have got them to work in *the last line* of the file. If I have three lines in the file, just the third export allows mounts according the mask. BTW2: Are you sure that nameservers should be separated with comma in /etc/resolv.conf like suggested in the boot set dialog. I have used just blank and it seems to work. Never tried comma. I would be happy with CD drivers in the kernel if there would be a way to disable autodetection of the devices that are not there. Is there a boot option to do that? Space is not such a big issue anymore. I have not found xload yet. fvwm95 is missing it. The old configuration file is missing xterm_color as well but it seems to be obsolete now(?) (I prefer this, the green cursor is awful.) Esa Turtiainen --------- Esa Turtiainen http://www.turtiainen.dna.fi [EMAIL PROTECTED] fax: +358-0-801 3008 -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED]
Re: Serial port IRQ strangeness
Hi, You are trusting that kerneld will load the module serial. However, this will happen just when kerneld is running, but it is just later in the boot sequence. Your system, just like mine, requires setserial for ttyS2 and ttyS3 to use the right interrupts. Compile a kernel with serial built in. Esa - Esa Turtiainen http://www.turtiainen.dna.fi [EMAIL PROTECTED] fax: +358-0-801 3008 On Wed, 20 Nov 1996, Steve Gaarder wrote: > I have a Debian 1.1 machine with 4 serial ports, each on its own IRQ. > I edited /etc/rc.boot/0setserial to configure the appropriate IRQs. > > Question #1: At boot, when /etc/rc.boot/0setserial runs, it complains > "no such device" for each of the devices. If I do the setserial as > /etc/rc2.d/S14setserial (after ), it works. Why? > > Question #2: I am running diald. Periodically, the following messages > appear in /var/log/messages: > > kernel: Serial driver version 4.13 with no serial options enabled > kernel: tty00 at 0x03f8 (irq = 4) is a 16550A > kernel: tty01 at 0x02f8 (irq = 3) is a 16550A > kernel: tty02 at 0x03e8 (irq = 4) is a 16550A > kernel: tty03 at 0x02e8 (irq = 3) is a 16550A > > and the IRQs for cua2 and cua3 are changed from 2 and 5 back to 4 and > 3! Any ideas? > > thanks, > > Steven Gaarder Network and Systems Administrator > [EMAIL PROTECTED] C-MOLD, Ithaca, N.Y., USA > > -- > TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to > [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED] > -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED]
dbackup
Hi, I just tested a package dbackup in experimental. It is exactly what I want: a list of files in my computer that are not a part of debian system. Is there another, more supported way to accomplish the same? Esa - Esa Turtiainen http://www.turtiainen.dna.fi [EMAIL PROTECTED] fax: +358-0-801 3008 -- This message was distributed manually by [EMAIL PROTECTED] after the list initially failed to distribute it.
kernel-source-2.0.23_0.deb corrupted
Hi, I just retrieved kernel-source-2.0.23_0.deb from a mirror in Sweden and from a mirror in Finland. Both were corrupted. I got it from subdirectory 'bo' and transfer went OK. The places are not mirrors of each other, the problem is likely in master. The error is: Unpacking kernel-source-2.0.23 (from kernel-source-2.0.23_0.deb) ... dpkg: error processing kernel-source-2.0.23_0.deb (--install): corrupted filesystem tarfile - corrupted package archive dpkg-deb: subprocess paste killed by signal (Broken pipe) Errors were encountered while processing: kernel-source-2.0.23_0.deb Esa ----- Esa Turtiainen http://www.turtiainen.dna.fi [EMAIL PROTECTED] fax: +358-0-801 3008 -- This message was distributed manually by [EMAIL PROTECTED] after the list initially failed to distribute it.
Emacs and XEmacs
Hi, I'd like to use both Emacs and XEmacs at the same time. Are there plans to allow this in the future. I have heard that people have patched this to work. Esa -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED]
Re: how to download in the background?
Hi, In libwww-perl there is a command GET that makes the trick. If you have a squid proxy, you can read it next time directly from cache. Otherwise, you can put it to a file. Esa On Mon, 11 Nov 1996, Lawrence Chim wrote: > Is it possible to download something, say http://xxx.xxx.xxx/abc.zip > in the background? For example, I can run ncftp, get a file, press > CTRL-Z, type bg, then I can logout and ncftp still downloading the > file for me. Is there an equivalent command for downloading from http? > > lawrence, > > -- > TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to > [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED] > -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED]
Re: Boot up problem...
On Sat, 9 Nov 1996, Tim Sailer wrote: > In your email to me, Adam Shand, you wrote: > > > > Configuring serial portsdone. > > /dev/cua0: No such device > > /dev/cua1: No such device > > /dev/cua2: No such device > > /dev/cua3: No such device > > INIT: Entering runlevel: 2 > > > > I traced the problem down to /etc/rc.boot/0setserial, I read the script > > (removed all the stuff I don't need) and manually defined the com ports. > > The error persisted. > > You didn't install the serial.o module when you were installing > the system. Either add 'serial' to /etc/modules, or uncomment that > > #auto > > line there to enable the kerneld autoloading of modules. > In fact, in Debian it seems that /etc/rc.boot/0setserial is executed always before /etc/init.d/kerneld. This means that you must never make serial as a module. In my configuration I must use setserial for the right interrupt addresses. So, setserial should be run after kerneld is running to get this to work. With kernel boot diskettes I can not boot my configuration where I have to use setserial first and build ppp connection after that, unless 0setserial is between /etc/rc2.d/S12kerneld and /etc/rc2.d/S13ppp (I have not tried actually, I found it easier to build a new kernel with serial built in.) Esa -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED]
Hint: slow Netscape
(I just subscribed the newsgroup, this may be an old one...) I recently wondered why my Netscape is so slow to start. I found that some Debian packages adds entries to /etc/mailcap many times and Netscape is *very slow* in processing them. I do not know yet the fundamental problem but I guess that I am not the only one. If your Netscape is even slower than you could expect, check the file. There should be just one copy of each lines. And Netscape is still faster if you remove lines you do not think to use. Esa -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED]
My experiences of the installation
I send here some notes of by beta test installation. Everything
works, I have just couple packages left that I don't really know
(dselect don't either) if they are installed or not.
Esa Turtiainen
[EMAIL PROTECTED]
Some notes of my Debian installation
Esa Turtiainen
16.5.1996
Because my old a.out Linux is totally outdated, I decided
to make a new installation from scracth. I selected Debian
because some years ago I worked with Sys V packages. I noticed
that they are a nice concept but to be usable the whole Unix
should be rewritten. Now it seems that someone is trying to do
that.
I did this to my home computer. I started at saturday (May, 11)
and on tuesday everything mostly worked including
- ppp connection
- X
- keyboard (always a problem...)
- sound
- CD-ROM
- mirroring
- httpd (and old WN pages and scripts converted to Apache)
- Harvest cache (not a part of debian, shame)
- SSH (the same)
- Netscape
- ... and all the nice utilities I have used to in Linux
I document here some of the problems I found. Mostly they were
not serious and likely not bugs as well. A common problem is
that I should re-install to find out afterwards to find out what
was the truth.
So, this deals with Debian 1.1 beta, snapshot taken about May, 11th.
Keyboard
After booting the new system the keyboard was not right. I had
selected the Finnish keyboard but the layout was still American.
Luckily I know it quite well, just close the eyes.
I tried many things. Likely the problem disappeared after I
re-installed the kbd package and rebooted.
Could it be that the /etc/rc.boot/0console was not executable in the
system from floppies? That is the place where it is set. (I have to
admit that I did not found it before the problem was solved.)
A lesser problem is that the map file for Finnish keyboar is not
complete. It is not very nice for coding because ', `, ~ and ^ are
dead characters for composite characters. There is a fix for this
in HOWTO/Finnish and it is just cosmetical for coders. Worse problem
for me is that Emacs control characters are not there. I had to add
the following addition to /etc/rc.boot/0console:
loadkeys << '@loadkeys'
alt shift keycode 6 = Meta_percent
shift control keycode 53 = Control_underscore
alt keycode 86 = Meta_less
alt shift keycode 86 = Meta_greater
alt altgr keycode 86 = Meta_bar
@loadkeys
to get my favorite Emacs keys to work. Isn't there any easier method
to get all the meta characters to work? ^] should be still added to
get easily the Telnet quit character. But really, they should all be
there.
HOWTO/Finnish suggest
LC_CTYPE=finnish.iso88591
All the perl programs give a three-line warning after that.
8-bit modes
---
There are good hints in HOWTO/Finnish how to make less, emacs, etc.
to work with 8-bit characters. Too elaborous.
Xterm seems to filter all the alt-commands to some 8-bit characters.
No Emacs-editing is possible. Rxvt seems to work, though.
Sound
-
It was a big job to get the sound working. This was due to new
kernel modules that I did not quite understood.
Most available configuration files speak of the dynamic module
loading with kerneld. However, it seems that the default kernel
don't support that. Therefore, the modules should be added to
/etc/modules. However, I did not do that like it. I compiled a new
kernel that used loadable modules.
Main reason to compile a new kernel was to get sound configuration
right. I suspect that the default configuration used default
SoundBlaster configuration that is for some reason wrong to my
SoundBlaster 16V.
Low DMA 1, High DMA 5, IRQ 5
It seems that you have to understand modules quite well to do an
advanced installation (like any using sound). There could be a little
more help for this. It took many days to believe that
HOWTO/mini/Modules is a good file to read.
I accidentally loaded NAS and it made something that did not
work at all.
Kernel sources
--
One file got installed to a wrong place in 'make
modules'. misc/sound.o got installed to /usr/src/linux/modules/misc,
not to /usr/src/linux-1.3.64 that is a totally different place. Easy
fix is to create directory modules to /usr/src/linux and copy the
result to the right place.
One important thing is to remember
depmod -a
after installation (make modules_install). Debian does not do this
in every boot like the system I used to have.
CD-ROM
--
For my Old Panasonic CD-ROM it is better to add the following option
to file /etc/conf.modules:
alias block-major-25 sbpcd
options sbpcd sbpcd=0x300,0
It is still very slow to load this driver but essentially faster than
without.
I think that the system should find sbpcd using driver name like
eth0 but it can not (BUG?).
It is a good idea to add the followi
