Re: Install PPC 8.2 on beige G3

[Christian Walther]

I've tried simply pressing and holding "C" after turning on the switch.
I've tried Command-C, Alt-C, Command-Alt-C, Command-Ctrl-Alt-C : all to
no avail, the hard disk just kicks in and boots OS9.2.
I tried Command-O-F and that gives me an error message on my monitor
screen saying "SIGNAL OUT OF BOUNDS".
I even tried installing the "Mandrake Linux Install.sit" file, from CD,
or copying to disk and then starting. In both cases, I get the error
message "The application Mandrake Linux Install.sit can not be opened
because an error of type -39 has occurred".
The beige G3 is an "Old World" machine, which basically means it has a 
big ROM chip containing half of Mac OS, and some older version of the 
boot code that doesn't support booting non-Mac-OS directly. You have to 
use BootX to start Linux out of Mac OS. BootX is contained in the 
"Mandrake Linux Install.sit" archive. Due to a misconfiguration of the 
Linux program that created the CD, this file is marked as being an 
application, although it isn't - that's why you get the error message 
when you doubleclick it. Just drop it on StuffIt Expander to expand it.

The AppleScript applet that should automate the BootX setup has some 
flaws too. Try it, but don't be surprised if it doesn't work - I 
recommend you configuring BootX by hand. I don't know the necessary 
options offhand, but if you have problems just ask and I'll look it up.

 -Christian

Re: Booting OS X off of 1394 drive with OF

[Christian Walther]

i copied all my OS X stuff over to the 1394 drive
How did you do that? Are you aware of the fact that just copying all 
the files using the finder does not yield a bootable Mac OS X 
installation (because the finder messes up file ownership and 
permissions)? See http://www.bombich.com/software/ccc.html for an 
utility that can make bootable HD copies.

 -Christian

zcip problem

[Christian Walther]

I'm having a problem with zcip on a Powerbook G4 Ti 867MHz with 9.1 
(final). It might already have been there in RC1, but I didn't 
encounter it then. When I start up without any network connection (i.e. 
no ethernet cable connected and no airport network in range), zcip 
assigns a 169.254.x.x address first to eth0 (ethernet), then to eth1 
(airport), and then nothing happens. The last line printed is 
"zcip[865]: Stored address 169.254.206.74 for eth1:9". The system isn't 
completely frozen, I can still type and change terminals, but it 
doesn't seem to do anything (at least for about 10 minutes), and I have 
to force restart. Everything works correctly if an ethernet connection 
is there (eth0 is configured by dhcp, and eth1 by zcip). I didn't try 
what happens if I have airport but no ethernet. I currently just work 
around the problem by not starting the network interfaces at startup.

Has anyone seen this problem too? Or not? Is there a solution?

 -Christian

Re: Radeon 9000 Mobility 1280x854

[Christian Walther]

Is the ModeLine for PB G4 (Radeon 9000) still missing
from XF86Config-4?
I believe it's there.

[EMAIL PROTECTED] perl-install]$ grep 854 Xconfig/*
Xconfig/xfreeX.pm:ModeLine "1280x854"   79.816 1280 1296 1408 1536
854  855  858  866 -hsync -vsync
It's there, but XFdrake doesn't offer that resolution in its list, 
neither in the installer nor in Mandrake Control Center. You have to 
manually add "1280x854" to the screen section in XF96Config-4.

 -Christian

Re: MOL 0.9.68

[Christian Walther]

http://people.mandrakesoft.com/~sbenedict/cooker-ppc/mol-0.9.68- 
5mdk.ppc.rpm
Works. I just installed 10.2 to a disk image and everything seems to  
work.

 -Christian

Re: ibook & XFree86

[Christian Walther]

Almost half the time, it boots then a blank screen when X is suppose 
to start.
I can't go to the console, I have to hard reboot.
I think your problem might be the same that I had: It was CUPS crashing 
in the background while X is already starting. You can check by booting 
to runlevel 3 (add "3" at the yaboot prompt), where X isn't started 
automatically to hide the console messages.
I finally worked around it by just deinstalling CUPS because I don't 
need to print in Linux anyway.

 -Christian

Re: MOL 0.9.68

[Christian Walther]

Did you only update mol, not mol-kmods?
Both, the source tarball contains both. And to be sure, I urpmed 
everything mol-related (i.e. mol and mol-kmods-benh) before installing 
the newly built mol. So I don't know if it would have worked with your 
kmods (and I'm not in Linux right now to test it).

 -Christian

Re: MOL 0.9.68

[Christian Walther]

It seems that Samuel was right about the bug being fixed in the release 
0.9.68. I extracted the kernel headers for benh-9mdk with rpm -bp and 
compiled the source tarball from 
http://www.maconlinux.org/download.html , and this version doesn't 
exhibit the problem (the easiest method to test it is booting from the 
Max OS X install CD, opening the "Reset Password" utility from the 
Installer menu, and using the text field for resetting the root 
password of the CD itself).

 -Christian

Re: MOL 0.9.68

[Christian Walther]

If you build kernel-benh from the source rpm, and then leave the tree 
in
your rpm/BUILD dir...
How do I do that? rpm --rebuild and rpm --recompile delete the files in 
BUILD after rebuilding the packages, and in the rpm man page I haven't 
found any option to disable this. I'm sure there's an easier way than 
waiting for the right moment and copying the files out just before they 
get deleted.

 -Christian

MOL 0.9.68

[Christian Walther]

Stew, what sources are you using to build mol-0.9.68-3mdk.ppc.rpm (and  
-2mdk and kmods for benh-7mdk and benh-9mdk)? I'm having a problem with  
it that its author Samuel Rydh says has been fixed in the release  
0.9.68 version (keyboard behaving as if keys were still down after  
releasing them in Mac OS X, see  
http://lists.maconlinux.org/pipermail/mol-general/2003-March/ 
001368.html).

My original idea was trying to compile the release 0.9.68 myself, but I  
can't do that because I don't have the kernel headers for benh-9mdk in  
/usr/src/linux/include. I've looked at kernel-benh-2.4.20-9mdk.src.rpm,  
but it doesn't look to me as if I could easily get these headers out of  
it (Do all these patches have to be applied? The README says something  
about not placing stuff in /usr/src/linux... is that applicable? I  
don't have anything in /usr/src/linux.).

 -Christian

Re: Networking Difficulty in RC1 (fix and another problem)

[Christian Walther]

If you're using dhcpcd (and that's pretty standard) the nameserver
IP-adresses will be placed in the /etc/resolv.conf (together with 
search
domain). Only if you specify the option -R this file will not be 
changed
by dhcpcd. Look for dhcpcd in the ps aux listing to see which options
are used.
Apparently I'm not using dhcpcd but dhclient. Some debugging shows that 
/sbin/dhclient-script, which (among other things) should rewrite 
/etc/resolv.conf, is the culprit. I can fix the problem by changing 
lines 23-31 of /sbin/dhclient-script from

  function make_resolv_conf() {
if [ -n "$new_domain_name" -a -n "$new_domain_name_servers" ]; then
  echo search $new_domain_name >/etc/resolv.conf
  for nameserver in $new_domain_name_servers; do
echo nameserver $nameserver >>/etc/resolv.conf
  done
  [[ -x /sbin/update-resolvrdv ]] && /sbin/update-resolvrdv
fi
  }
to

  function make_resolv_conf() {
if [ -n "$new_domain_name" -o -n "$new_domain_name_servers" ]; then
  echo >/etc/resolv.conf
  if [ -n "$new_domain_name" ]; then
echo search $new_domain_name >/etc/resolv.conf
  fi
  if [ -n "$new_domain_name_servers" ]; then
for nameserver in $new_domain_name_servers; do
  echo nameserver $nameserver >>/etc/resolv.conf
done
  fi
  [[ -x /sbin/update-resolvrdv ]] && /sbin/update-resolvrdv
fi
  }
because for me, $new_domain_name is empty while 
$new_domain_name_servers contains the DNS addresses. Linux gurus out 
there, are these changes reasonable? Do they break anything?

Once I put a name server address into
/etc/resolv.conf, everything works, but according to man resolv.conf,
this isn't the recommended way.
Where did you find that? I read in the man-pages that /etc/resolv.conf
is not necessary if you're running a nameserver on the local machine,
which to my opinion is not reasonable for a desktop machine.
man resolv.conf says "On a normally configured system, this file should 
not be necessary. The only name server to be queried will be on the 
local machine, the domain name is determined from the host name, and 
the domain search path is constructed from the domain name." Apparently 
its author doesn't consider machines not running a nameserver "normally 
configured". This lead me to believe that running a local nameserver 
was the normal way of dealing with DNS in linux.

While playing around with the DHCP stuff, I came across another 
weirdness: Whenever my machine is being configured by DHCP, its 
hostname is "x1-6-00-03-93-ce-13-e6". This contains the hardware 
address of the ethernet interface, but where does it come from? The DNS 
hostname for the IP address provided by DHCP is something like 
"dclient80-218-137-66.hispeed.ch". When logging into Gnome, this causes 
an error dialog about not being able to look up an IP address for 
"x1-6-00-03-93-ce-13-e6". When I'm not connected to the network, the 
hostname is "localhost".

 -Christian

Re: Networking Difficulty in RC1

[Christian Walther]

I can ping local and remote IP addresses from the terminal window, 
both as root and as a regular user, but Mozilla and the other network 
apps do  not work either in KDE or Gnome. IFCONFIG reports up and 
running built-in (eth0) or wireless (eth1) with appropriate IP's 
provided by DHCP in my router.
Maybe your problem is the same as mine (which I just forgot to report 
until now):

I get an IP address from my cable provider's DHCP server, and I can 
connect to everything using IP addresses, but domain name lookups don't 
work. It seems that the name server address provided by the DHCP server 
doesn't get to the proper place. Once I put a name server address into 
/etc/resolv.conf, everything works, but according to man resolv.conf, 
this isn't the recommended way. I have no idea however what's the 
recommended way domain name lookups should work on linux. Can anyone 
clear this up for me?

 -Christian

Re: RC1 ISOs

[Christian Walther]

How did you boot the install?  The "Read-only file system" is the issue
above - no modules are loading.

Did you specify kernel-options "root=/dev/ram3" and 
"ramdisk_size=36000"
(after selecting the right kernel & initrd)?
Thanks for the hint... adding "root=/dev/ram3" did it. (RTFM?) But I 
wonder, why did it work with the old kernel without that? What does 
BootX use as root by default? Ramdisk size was set to 4 using 
BootX' ramdisk size field. Graphical installer works well now. I can't 
do an install at the moment though because I don't have any HD space 
left on the Wallstreet. Maybe I'll dig the second SCSI HD out of the 
7300 to test it.

By the way, what I miss from the new installer is being able to click 
"Exit Install" in the steps list to reboot. There's a quit button in 
the license agreement dialog, but none in the later dialogs, so I 
haven't found any other way than force rebooting ("shutdown", "halt", 
"reboot" from console 2 don't work either).

 -Christian

Re: RC1 ISOs

[Christian Walther]

I haven't tried a full install on my PowerBook G4 Ti 867MHz yet because 
I'm still downloading CD 2 and 3, but the graphical installer comes up 
with both the benh and mdk kernels and seems to work well (and look 
great - much better that the old one!).

I can however already report that it doesn't work with my oldworld 
machines, a PowerBook G3 Wallstreet and a Power Mac 7300. The following 
is from the Wallstreet, but it's similar on the 7300.

With both the mdk and benh kernels from the boot folder on the CD, I 
get an error dialog "I can't access a Mandrake Linux Installation disc 
in you CDROM drive (MATSHITA CR-174)."

Console 2 at this moment reads:

* loading modules dependencies
* PCMCIA: probing PCI bus..
*   TI 1131 found, 2 sockets.
* have to insmod pcmcia_core
* needs pcmcia_core
* /tmp/pcmcia_core.o: Read-only file system
* warning, insmod failed (pcmcia_core (null)) (1)
* have to insmod yenta_socket
* needs pcmcia_core
* /tmp/pcmcia_core.o: Read-only file system
* needs yenta_socket
* /tmp/yenta_socket.o: Read-only file system
* warning, insmod failed (yenta_socket (null)) (1)
* have to insmod ds
* needs pcmcia_core
* /tmp/pcmcia_core.o: Read-only file system
* needs ds
* /tmp/ds.o: Read-only file system
* warning, insmod failed (ds (null)) (1)
* CM: cardmgr/hacked starting, version is 3.1.23
* CM: no pcmcia driver in /proc/devices
* CM: cardmgr: failed
* cardmgr rc: -1
* AUTOMATIC: parameter cdrom for method means returning CDROM drive
* have to insmod ide-cd
* needs cdrom
* /tmp/cdrom.o: Read-only file system
* needs ide-cd
* /tmp/ide-cd.o: Read-only file system
* warning, insmod failed (ide-cd (null)) (1)
* have to insmod sr_mod
* needs cdrom
* /tmp/cdrom.o: Read-only file system
* needs sr_mod
* /tmp/sr_mod.o: Read-only file system
* warning, insmod failed (sr_mod (null)) (1)
* looking for ide media
* IDE/1: hda is a IBM-DKLA-24320
* IDE/0: hdc is a MATSHITA CR-174
* looking for scsi media
* looking for Compaq Smart Array media
* looking for DAC960
* /dev/hdc: Read-only file system
* could not create required device file
Using the 2.4.18-4mdkBOOT kernel from 8.2 that's still in the "Mandrake 
Linux Install.sit" on the CD, loading stage 2 works and the graphical 
installer starts without problems, but eventually fails because it 
doesn't find /lib/modules.cz-2.4.18-4mdkBOOT (which of course isn't 
there because the stage 2 ramdisk wasn't made for that old kernel).

 -Christian

Re: Processor Speed and /proc/cpufreq

[Christian Walther]

On my PowerBook G4 Ti 867MHz with 2.4.20-benh-5mdk it's there and 
works. It allows setting the frequency between 667 and 867 MHz, and in 
a quick test with the Mesa demos (without hardware accelerations) it 
shows.

 -Christian

Re: sound on ibook 2002 w/ 9.1 beta 2

[Christian Walther]

Oops, that should have gone to the list, no to you personally... Is  
there a reason why you are using a reply-to header?

modprobe  dmasound_pmac
/lib/modules/2.4.20-2mdk/kernel/drivers/sound/dmasound/ 
dmasound_pmac.o.gz: init_module: No such device
Hint: insmod errors can be caused by incorrect module parameters,  
including invalid IO or IRQ parameters.
  You may find more information in syslog or the output from dmesg
modprobe: insmod  
/lib/modules/2.4.20-2mdk/kernel/drivers/sound/dmasound/ 
dmasound_pmac.o.gz failed
modprobe: insmod dmasound_pmac failed

I'm using the benh kernel because thats the only one that would aloow  
the installer to work.  Also, there are no entries in /dev/sound.   
It's an empty directory.

"/lib/modules/2.4.20-2mdk/..." doesn't look like you're really using  
the benh kernel. Even if you use benh for the installer, the installed  
system by default uses the 2.4.20-2mdk kernel. You have to install it  
manually from the CD (or even better, use the newer version from  
).

I had the same problem with 2.4.20-2mdk on my 867MHz TiBook, but with  
benh-2.4.20-5mdk, it works fine.

 -Christian

Re: New mdk kernel, kernel-benh

[Christian Walther]

Another thing: Recently, I'm frequently having freezes at startup,
sometimes at the black screen before X comes up, sometimes at the blue
X startup screen before any windows appear. Mouse is frozen, no
reaction to any key presses (including screen brightness, which works
from very early on with the benh kernels). This is independent on 
which
kernel I use. I'm having the impression that it starts when switching
from one kernel to another, then comes again for the next few 
restarts,
and once a startup finished successfully, it doesn't return until I
switch kernels again - but I'm not completely sure about this. I
suppose it's one of the things that continue getting started in the
background while X comes up. Any ideas on what it could be, or how I
could narrow it down?

A depmod is run with a new kernel, not sure what else might be going 
on in
the background.  You could try to boot into runlevel 3 and see if you 
can
see more of what's going on. (append "3" to your yaboot choice, ie:
2320-benh 3)

New results:
o It doesn't have anything to do with switching kernels, I have now 
seen it occur also when rebooting with the same kernel.
o Booting to runlevel 3 shows that it occurs at "Starting CUPS printing 
system:". As a workaround, I could deinstall CUPS, since I don't need 
printing support at the moment.

 -Christian

Re: New mdk kernel, kernel-benh

[Christian Walther]

My experiences with the two new kernels on an 867MHz TiBook (Radeon 
Mobility 9000):

kernel-2.4.21.0.pre4.5mdk-1-1mdk.ppc.rpm


Doesn't work very well. Sound seems to work (no error message at 
startup), but X doesn't: all I get is a flickering brown mess instead 
of the login screen on blue background. The mouse cursor is right in 
colors and size, but its left half is on the right side of its right 
half. The system seems to be alive and responsive (the mouse cursor 
moves and changes its shape according to what's supposed to be below 
it), but except for the cursor nothing is recognizable. Trying to 
switch to a text console gives the moving northern lights I've already 
seen with the last mdk kernel.

kernel-benh-2.4.20-5mdk.ppc.rpm


Works fine. To be exact, exactly as fine as benh-2.4.20-3mdk. X display 
after waking up from sleep is still wrong, but can be cured by 
switching to a text console and back. CPU_FREQ works too. What I'm 
missing is kernel headers to build new mol-kmods against - or can I get 
these out of the srpm somehow?

Another thing: Recently, I'm frequently having freezes at startup, 
sometimes at the black screen before X comes up, sometimes at the blue 
X startup screen before any windows appear. Mouse is frozen, no 
reaction to any key presses (including screen brightness, which works 
from very early on with the benh kernels). This is independent on which 
kernel I use. I'm having the impression that it starts when switching 
from one kernel to another, then comes again for the next few restarts, 
and once a startup finished successfully, it doesn't return until I 
switch kernels again - but I'm not completely sure about this. I 
suppose it's one of the things that continue getting started in the 
background while X comes up. Any ideas on what it could be, or how I 
could narrow it down?

 -Christian

Re: 9.1b2 on Nov 02 Powerbook

[Christian Walther]

o install-gui-benh works, except for the colors, which are still 
wrong.

I'm trying to remember.  Did we try different bit depths on you machine
for install to see if the color issue was corrected?  If we can narrow
down the machine ID, I could force an appropriate bit depth if that is 
the
needed fix.

No, changing the bit depth didn't correct the colors. Here's what I 
wrote about it on Jan 11:

I've tried with 16, 15 and 8, and the colors are different every time, 
but still wrong. 15 gives the most readable display. With 8, the 
colors are dithered, but so to speak before the color distortion is 
applied, so the result is multicolored noise that makes the text 
almost unreadable.
I can't really explain this to myself, since I have examined the 
XF86Config file generated by the installer and have found no 
fundamental difference to the one that works fine on my installed 
system. The XFree86 log doesn't show anything insightful (at least to 
me) either.

I didn't test 24 or 32 at that time because it wasn't supported by XF86.


o I chose the radeon X driver and 1280x854 in 24bit, testing the
configuration works, and after the testing, the colors in the 
installer
are right!

Interesting.  I wonder what triggered that?  There is a stage2 
installer
image on my web space, that has a newer X built into it, if you are 
setup
to be able to test an ftp install (You don't need to do a full install,
just see if the GUI comes up correctly).

That works! The colors are right.


...


Hmm - I patched in the gatos drm into -4mdk, hoping it would help your 
DRI
issue.  Sounds like it broke other things.  Maybe I should roll it 
back.

DRI is not really important for me - it's rather a toy factor. And as 
it doesn't work with 4mdk either, I'd rather have sleeping and console 
switching back. And XFree86 has made great progress with respect to 
radeon compatibility in the last weeks, so I expect to have it working 
eventually anyway.

Sound works with the benh kernel?


Yes.

Thanks

 -Christian

9.1b2 on Nov 02 Powerbook

[Christian Walther]

My experiences with 9.1 beta 2 on a Novermber 2002 Ti PowerBook G4:  
Most things work out of the box, but not everything. Here is a random  
list of problems:

Installer:

o install-gui-benh works, except for the colors, which are still wrong.
o The "Choose a monitor" step doesn't list a TiBook 1280x854, so I  
chose a TiBook 1152x768.
o I chose the radeon X driver and 1280x854 in 24bit, testing the  
configuration works, and after the testing, the colors in the installer  
are right!


Installed system with Mandrake kernel:

o X works, using the radeon driver and 24bit color, even 2D accelerated.
o When switching from the X console to a text console, I don't get a  
console display but some strange moving multicolored northern lights.
o The powerbook doesn't sleep when the lid is closed, even after  
manually installing pmud which wasn't installed by default.
o Adding 'Load "dri"' to XF86Config-4 produces the following error  
message in the XF86 log:
	(EE) RADEON(0): [dri] RADEONDRIScreenInit failed because of a version  
mismatch.
	[dri] radeon.o kernel module version is 1.2.0 but version 1.5.0 or  
newer is needed.
	[dri] Disabling DRI.
o Sound doesn't work: At startup, there are some error messages that I  
can't read because they go away so fast, when I manually "modprobe  
dmasound_pmac" I get the following error message:
	/lib/modules/2.4.20-2mdk/kernel/drivers/sound/dmasound/ 
dmasound_pmac.o.gz: init_module: No such device
	Hint: insmod errors can be caused by incorrect module parameters,  
including invalid IO or IRQ parameters.
	  You may find more information in syslog or the output from dmesg
	modprobe: insmod  
/lib/modules/2.4.20-2mdk/kernel/drivers/sound/dmasound/ 
dmasound_pmac.o.gz failed
	modprobe: insmod dmasound_pmac failed
o Adjusting the display brightness (with the keys whose secondary  
functions are F1 and F2) doesn't work (this works with the benh kernel).


With 2.4.20-benh-4mdk kernel:

o When switching from the X console to a text console, I get a row of 4  
(unreadably) small copies of the console display at the top of the  
screen, the rest of the screen doesn't change.
o When waking up from sleep, I just get a black screen and no sign of  
any reaction to key presses, so I have to reboot.
(these two things worked with the 2.4.20-benh-3mdk kernel on my cooker  
system)
o Adding 'Load "dri"' to XF86Config-4 produces the following error  
message in the XF86 log:
	(EE) RADEON(0): [dri] DRIScreenInit failed.  Disabling DRI.


Independent of kernel:

o gedit doesn't work: When I start it from the command line, it says:
	gedit: error while loading shared libraries: libgnomeprintui-2-2.so.0:  
cannot open shared object file: No such file or directory
o No XFree86-75dpi-fonts are installed, which causes GTK applications  
that don't use antialiased fonts, such as GIMP, to use an ugly  
monospaced font everywhere.
o When installing additional packages with rpmdrake, I get an error  
message 'Medium "Cooker_CD_1 (cdrom1)" is not selected' when that CD is  
not inserted, although it is selected in the media selection dialog.  
When the CD is inserted, everything works.
o On the subject of missing packages, one thing I missed is Evolution.


 -Christian

Re: New benh-mdk kernel build

[Christian Walther]

> kernel-benh-2.4.20-3mdk.ppc.rpm

Works well on my Nov 02 PowerBook G4. In fact, I don't see any difference
compared to benh-2.4.20-2mdk. I still have the error messages at startup about
"IP Virtual Server" not being available as a module or built into the kernel.

Moreover (I don't know if that's due to the new kernel or the new XFree86),
I do now get 24-bit video with the radeon driver (Radeon Mobility 9000 here).
According to the XF86 log, it's even 2D-accelerated. If I try 32 bit, I just
get a black screen that doesn't change even if I switch to other consoles or
sleep and wake up again. The system remains functional however, and I can
restart it by typing blindly.

After waking up from sleep, the X display is weird: somehow scaled up, but
as individual pixels in wrong colors on a black background, except for the
mouse cursor, which appears normal (hard to describe, but I guess it doesn't
matter anyway). Switching to a text console and back returns it to normal
operation.

When I add 'Load "dri"' to the module section of XF86Config, I get the
following lines in the XFree86 log:

(EE) RADEON(0): [dri] RADEONDRIScreenInit failed because of a version
mismatch.
[dri] radeon.o kernel module version is 1.1.1 but version 1.5.0 or newer is
needed.
[dri] Disabling DRI.

 -Christian

-- 
+++ GMX - Mail, Messaging & more  http://www.gmx.net +++
NEU: Mit GMX ins Internet. Rund um die Uhr für 1 ct/ Min. surfen!

Mac side of hybrid ISOs

[Christian Walther]

Stew,

as you're making ISOs again, this could be the time for some contribution to
enhance the user experience on the Mac side of the hybrid CD, namely the
type and creator properties that tell the Mac OS Finder what application to
launch when one double-clicks a file.

>From examining the latest ISO, it seems that you are making these ISOs using
the (CD)/misc/mkcd_ppc.pl script, is that right?

My modifications are:

o I have modified the files (CD)/misc/tools/ppc/mapping and
(CD)/misc/tools/ppc/magic in order to produce more appropriate
types/creators for the files on the CD. I don't attach these files here
because there's no point in the whole list getting them, but I can send them
to you privately, if you like.

o (CD)/misc/mkcd_ppc.pl, line 213, reads:

system("hattrib -t APPL -c STi0 'Mandrake Linux Install.sit'");

This is just plain wrong, and causes the "error -39" problem described on
http://www.mandrakelinux.com/en/82errata-ppc.php3 . For this file to open
with StuffIt Expander when double-clicked, this line should be

system("hattrib -t SITD -c SITx 'Mandrake Linux Install.sit'");

but I doubt that it's necessary at all, since my updated mapping file should
take care of this.

I haven't tried out all that yet, because I suspect that making my own ISO
would require some work getting my installer tree into a suitable form
first, but I see no reason why it shouldn't work.

 -Christian

Re: x wont boot

[Christian Walther]

> G4Ti.

If it's a november 2002 machine (867MHz or 1GHz, 1280x854 screen, Radeon
Mobility 9000), you need a BenH kernel for X to work
(http://people.mandrakesoft.com/~sbenedict/cooker-ppc/kernel-benh-2.4.20-2md
k.ppc.rpm).

See my posts to the list (and Stew's replies) starting on 2002/11/28.

 -Christian

Re: BenH kernel based install

[Christian Walther]

>>> I'm using 15bpp by default, which seems OK on a number of machines.  You
>>> could try editing line 362 of install_gtk.pm
>>> (mdkinst/usr/bin/perl-install) and change it to see if 16bpp is better for
>>> you.  I could then force this the same way I do the resolution, up at line
>>> 278.
>> 
>> I've tried with 16, 15 and 8, and the colors are different every time, but
>> still wrong. 15 gives the most readable display. With 8, the colors are
>> dithered, but so to speak before the color distortion is applied, so the
>> result is multicolored noise that makes the text almost unreadable.
>> I can't really explain this to myself, since I have examined the XF86Config
>> file generated by the installer and have found no fundamental difference to
>> the one that works fine on my installed system. The XFree86 log doesn't show
>> anything insightful (at least to me) either.
>> 
> 
> At this point the installer and the installed system are now using
> different versions of XFree. I'm a little hesitant to rebuild the
> installer on my cooker system. (see Phil's broken XFree report). YOur
> installed system may be running at 24bpp or 32.

No, unfortunately 24 or 32 is not supported by the fbdev driver. At install
time, I chose 15 and later changed it to 16 (and I was astonished about the
huge difference that extra green bit makes...)

>> o The installer didn't autodetect my airport card, and in the driver list
>> that appears when you answer "Yes" to the "Do you have another ethernet
>> interface" question, I didn't find anything that looked like an airport
>> driver.
> 
> If you care to do another install, this could be tested, or perhaps on
> your live system:
> 
> list_modules.pm, line 17:
> 
> if_(arch() =~ /ppc/, qw(mace bmac gmac)),
> 
> becomes 
> 
> if_(arch() =~ /ppc/, qw(mace bmac gmac airport)),
> 
> On the live system it's in /usr/lib/libDrakX/list_modules.pm.
> drakconnect should go through similar motions as the installer routine.
> If you could test this I can readily make that change.  I don't have
> airport.

With this change made, I can now choose airport from the list (with
drakconnect on the live system). Additionally, if I change line 89 of
/usr/lib/libDrakX/modules.pm from

if_($category =~ /net/, 'bmac', 'gmac', 'mace'),

to

if_($category =~ /net/, 'bmac', 'gmac', 'mace', 'airport'),

the airport card is even autodetected.

 -Christian

Re: BenH kernel based install

[Christian Walther]

>>> If you have the latest yaboot (yaboot-1.3.8-2mdk.ppc.rpm), I patched the
>>> issue I mentioned before.

Where do I get this? The mirrors I use, and they are up to date otherwise,
only habe 1.3.8-1mdk.

>> Thanks, I'll report how it goes.

HD install works now. See below for the rest...

> I'm using 15bpp by default, which seems OK on a number of machines.  You
> could try editing line 362 of install_gtk.pm
> (mdkinst/usr/bin/perl-install) and change it to see if 16bpp is better for
> you.  I could then force this the same way I do the resolution, up at line
> 278.

I've tried with 16, 15 and 8, and the colors are different every time, but
still wrong. 15 gives the most readable display. With 8, the colors are
dithered, but so to speak before the color distortion is applied, so the
result is multicolored noise that makes the text almost unreadable.
I can't really explain this to myself, since I have examined the XF86Config
file generated by the installer and have found no fundamental difference to
the one that works fine on my installed system. The XFree86 log doesn't show
anything insightful (at least to me) either.

Otherwise, the install worked mostly flawlessly. After manually installing
the 2.4.20-benh-2mdk kernel, X works as well, without having to mess with
the XF86Config.

Some issues:

o The installer didn't autodetect my airport card, and in the driver list
that appears when you answer "Yes" to the "Do you have another ethernet
interface" question, I didn't find anything that looked like an airport
driver.

o When booting with the benh kernel, I get error messages about an "IPVS IP
virtual server" (or something similar) not being available as a kernel
module or built into the kernel.

o When fsck runs at startup and asks for input, the keyboard layout is
completely messed up. I believe this is connected to the fact that at this
point the only message shown about keymaps yet is "loading default keymap:",
whereas I recall this being "loading default keymap: us" in previous
installs (where the keyboard was right at that stage).

o Gnome-panel still segfaults when I log out of Gnome.

 -Christian

Re: BenH kernel based install

[Christian Walther]

> My bad on this one. I had devfs enabled.  New kernel is at:
> 
> http://people.mandrakesoft.com/~sbenedict/cooker-ppc/install-benh/vmlinux-benh
> 
> You need new all-benh.gz, and  modules.cz-2.4.20-benh-2mdkBOOT since
> module symbols changed too.
> 
> I did both an nfs and HD install with it.  If you have the latest yaboot
> (yaboot-1.3.8-2mdk.ppc.rpm), I patched the issue I mentioned before.

Thanks, I'll report how it goes.

In the meantime, I have tried an FTP install with the last benh installer. I
get a graphical installer, but the colors are wrong. They don't change when
I switch to a text console and back either. It looks like it uses a wrong
color palette (does it use 8 bit color?).

Besides, in the instructions you posted to the list, you forgot to mention
that we should run make_mdkinst_stage2 after copying the modules file into
mdkinst/lib (at least that was what fixed the "can't load modules" error for
me).

Otherwise, everything seemed to work well up to the hard disk partitioning,
where I stopped because I didn't want to copy all the packages over to the
machine I used as FTP server.

 -Christian

Re: BenH kernel based install

[Christian Walther]

I wonder - am I the only one to have problems with the new BenH kernel based
installer (see my post of Dec 27)? Or am I just the only one who has tried
it?

 -Christian

Re: BenH kernel based install

[Christian Walther]

That was fast, thanks! I hope you didn't sacrifice your whole enjoying the
holidays for it?

I have not come very far with it however. I have installed the kernel and
stuff on my HD, set up yaboot, and the installer starts fine. When I choose
to install from HD, it says "no partitions found". VT3 lists the partitions
in lines such as:
* looking at partition table: 3 14 4718592
ide/host0/bus0/target0/lun0/part14 0 0 0 0 0 0 0 0 0 0 0

I believe this is the same bug as you described, but I can't follow your
instructions for a workaround because I don't have a shell on VT2 (VT3 says
"* cannot open shell - /tmp/sh doesn't exist" after "* spawning a shell").

I tried mounting the ramdisk image and digging around in it, but didn't gain
much insight from that (in particular, i found neither a shell nor ofpath).

 -Christian

Tidbits from Nov 02 PBG4

[Christian Walther]

Some tidbits from my efforts to get Linux onto my new PowerBook G4 867MHz:

o Kernel 2.4.20-benh-2mdk works fine. Not more or less sleep or X issues
than before: colors on the X console are wrong after waking up, right again
after switching to a text console and back. X only works in 16bit frambuffer
mode, but that's probably XF86's fault.

o Sound works after I "modprobe dmasound_pmac". Why is this not done on
startup?

o Sometimes (but as far as I remember not always), gnome-panel segfaults
when logging out of gnome.

o I've found this page from someone who uses yellowdog on such a powerbook:
  http://www.staikos.on.ca/~staikos/tibook/

o Recently, something went wrong with my linux install (I did so many
experiments that I've no idea what it was): Files have vanished. perl is
gone, xinetd is gone, bzip2 is gone, wget is gone, which is gone, and
probably some more that I didn't discover yet. Trying to urpmi them again
just says "everything already installed". So my system is now in a pretty
unusable state (which is no catastrophe because it's a purely experimental
system anyway), and I'm in for a reinstall. However, I'd like to wait until
an installer with a benh kernel is out, so that I can try out the graphical
install, is that going to happen anytime soon? Or can I easily build such an
installer myself?

 -Christian

Re: 'gendistrib' idea

[Christian Walther]

> Another thought: maybe run it from rescue, if I could get the proper bits
> on the rescue image (if they're not already there)

I've run into this problem too, and also had the idea of trying it from
rescue, either with the 8.2 CD or with the first cooker install ISO (don't
remember), but it didn't work (missing libraries or wrong versions or
whatever, I don't remember exactly). I had to install 8.2 to run gendistrib
over my cooker packages.

 -Christian

Re: New Installer on Nov 02 PBG4

[Christian Walther]

> kernel-benh-2.4.20-1mdk.ppc.rpm

Works mostly fine. Only troubles so far were:

o When installing it with urpmi, it complains about "unresolved symbols in
lib/modules/2.4.20-benh-1mdk/kernel/net/sunrpc/sunrpc.o"

o When booting it for the first time, it complains about not being able to
write to "/lib/modules/2.4.20benh-1mdk/modules.dep". That's because the
modules are in 2.4.20-benh-1mdk (with a dash). I worked around this by
making 2.4.20benh-1mdk a symlink to 2.4.20-benh-1mdk.

o When mounting filesystems at boot it "can't locate module supermount" and
thus "fs type supermount not supported by kernel".

Otherwise, everything seems to work alright. My Powerbook even sleeps with
this kernel, which it didn't do with the other ones (the colors on the X
console are wrong after waking up, but this can be cured by switching to a
text console and back).

And two unrelated notes:

o Everytime I log out from Gnome or KDE to the graphical login (kdm?), the
mouse pointer is frozen (but keyboard navigation works), and I have to
restart the X server to get it working again.

o Somehow my configuration of using F10 and F11 as middle and right mouse
buttons got lost. How can I get this back? I vaguely remember that there
were some lines about it in some configuration file (maybe
/etc/sysctl.conf), but what are they exactly?

 -Christian

Re: New Installer on Nov 02 PBG4

[Christian Walther]

> OK - so it sounds like we need the radeon driver from Ben's kernel, or an
> alternate kernel.  I've been putting together an alternative Benh-Mdk
> kernel, which I think I'll provide as an alternative install/system
> kernel.  Would you like to test it?

Sure. Just tell me where to get it.

 -Christian

Re: Newbie to Linux on PPC

[Christian Walther]

> The first stumbling block I have run into is it came with mac os x jaguar
> installed, with 9.2 in a sepirate partion. This powebook boots into os x,
> from there when I launch bootX it launches 9.2 to run bootX, I set it up
> (which is the best video driver for a powerbook wallstreet?) once configured
> I select linux and it shuts down the 9.2 but does not reboot the system. Only
> returns me to os x. I have thought about just formatting the drive with
> disktools however untill I am certain I can boot from cd and initiate the
> install process i am leary of ending up with no os and no way to acess
> somthing like bootX
> What is the procedure for directing a powerbook wallstreet to boot from cd?

The Wallstreet can not directly boot anything but MacOS from CD (because it
is a so called "Old World" machine), you have to use BootX. BootX doesn't
work in the Classic environment of Mac OS X (as you have seen), you need to
be booted directly into 9.2. Choose "System Preferences" from the apple
menu, then "Startup Disk", select the 9.2 partition and click "Restart".

As far as I remember (it's been some time since I last booted Linux on my
Wallstreet) the right video driver is atyfb. You might need some special
vmode argument, there is a note about it at the bottom of some documentation
file on the install CD.

You might like to read what I wrote to the list when I installed Mandrake on
my Wallstreet, starting at
http://www.mandrake.com/en/archives/cooker-ppc/2002-02/

 -Christian

New Installer on Nov 02 PBG4

[Christian Walther]

I have now tried out the new installer ISO with an up-to-date local cooker
mirror on my PowerBook G4 867MHz, and these are my experiences:

o I still don't get a graphical install (not surprisingly). When I choose
install-gui at the boot prompt, I get the following output after stage 2
loading:

> in second stage install
> VERSION file missing
> 
> XFree86 Version 4.2.1 / X Window System
> (protocol Version 11, revision 0, vendor release 6600)
> Release Date: 3 September 2002
> If the server is older than 6-12 months, or if your card is
> newer than the above date, look for a newer version before
> reporting problems.  (See http://www.XFree86.Org/)
> Build Operating System: Linux 2.4.19-19mdk ppc [ELF]
> Module Loader present
> Markers: (--) probed, (**) from config file, (==) default setting,
>  (++) from command line, (!!) notice, (II) informational,
>  (WW) warning, (EE) error, (NI) not implemented, (??) unknown.
> (==) Log file: "/var/log/XFree86.0.log", Time: Thu Dec 12 17:12:18 2002
> (++) Using config file: "/tmp/Xconf"
> Using vt 7
> Couldn't open RGB_DB '/usr/X11R6/lib/X11/rgb'
> FBIOPUT_VSCREENINFO: Invalid argument
> (EE) FBDEV(0): Mode init failed
> 
> Fatal server error:
> AddScreen/ScreenInit failed for driver 0
> 
> 
> When reporting a problem related to a server crash, please send
> the full server output, not just the last messages.
> This can be found in the log file "/var/log/XFree86.0.log".
> Please report problems to [EMAIL PROTECTED]
> 
> XIO:  fatal IO error 104 (Connection reset by peer) on X server ":0.0"
>   after 0 requests (0 known processed) with 0 events remaining.
> Thu Dec 12 17:12:19 2002 Gtk-Warning **: cannot open display: :0 at
> /usr/bin/perl-install/ugtk.pm line 146.
> install exited abnormally :-(
> sending termination signals...done
> sending kill signals...done
> unmounting filesystems...
> /proc
> /tmp/hdimage
> /tmp/stage2
> you may safely reboot your system

o Text install works however (no crash at bootloader config this time), I
did a recommended install, with two minor quirks: - The window that says
"Formatting hda11" (or something like this) doesn't get closed and stays
visible in the background during the whole install. - The "Package
Selection" dialog is empty except for the "OK" and "Cancel" buttons. When I
chose "OK", the install went on and installed a lot of packages (that
eventually led to a working system), but I have no idea on which criteria
these packages were chosen.

o At X setup, the monitor choice dialog doesn't list any 1280x854 LCD, so I
chose the TiBook 1152x768. For the X server, I chose fbdev.

o The newly installed system (with kernel 2.4.20-2mdk) boots fine and seems
to work alright, except for X: when I do "startx", I get the same error
messages as above. Dmesg still doesn't show any "radeonfb" lines.

o With a BenH kernel from ppckernel.org (2.4.20-rc4-ben0), the radeon gets
recognized, and X works well in 16bit color in fbdev mode. I have not yet
tried to get it to work in native radeon mode.

 -Christian

Re: Graphical Install on 1280x854 Powerbook?

[Christian Walther]

>>> Hmm, cat /proc/fb should also be informative.
>> 
>> Here:
>> 
>> 0 OFfb /pci@f000/ATY,XiaParent@10/ATY,
>> 
>> I can't make much sense of this, can you?
>> 
> 
> Looks like your booting with some video=ofonly arguments or something?
> 
>> o An excerpt from dmesg:
>> 
>> ...
>> Using unsupported 1280x854 ATY,Xia_A at bc008000, depth=8, pitch=1280
>> Console: switching to colour frame buffer device 160x53
>> fb0: Open Firmware frame buffer device on
>> /pci@f000/ATY,XiaParent@10/ATY,Xia_A
>> no framebuffer address found for /pci@f000/ATY,XiaParent@10/ATY,Xia_B
>> ...
> 
> I have a 1280x854 config that was sent to me, which has been added to the
> new installer. Have you tried with no "video=" arguments passed to the
> kernel?

You were right, there was indeed a video argument (leftover from my earlier
trials), namely "video=radeon:1280x854-8@60". I took that out (the "Kernel
command line" line in dmesg confirms that it was really taken out) , but
that didn't change anything. /proc/fb is still the same, as are the dmesg
lines above.

 -Christian

Re: Graphical Install on 1280x854 Powerbook?

[Christian Walther]

>> clock   : 667MHz
>> bogomips: 665.19

>> (That 667MHz puzzles me. This machine is supposed to run at 867MHz - is this
>> some power saving feature that only MacOS knows how to turn off?)
> 
> I wouldn't sweat the bogomips, it's not all that reliable anyway.

I wasn't referring to the bogomips (I don't even know what that is). I meant
the "clock : 667MHz" line.

>> But I'm far away from X: It seems that the kernel doesn't even recognize the
>> Radeon card, with both Mandrake kernels (2.4.18 from 8.2 and 2.4.19 from the
>> cooker mirrors) and a bk 2.4.20 kernel from ppckernel.org. According to
>> something I read on some yellowdog mailing list archive, dmesg should contain
>> some lines starting with "radeonfb", which doesn't happen for me.
> 
> Hmm, cat /proc/fb should also be informative.

Here:

0 OFfb /pci@f000/ATY,XiaParent@10/ATY,

I can't make much sense of this, can you?

And some other things that could help:

o An excerpt from dmesg:

...
Using unsupported 1280x854 ATY,Xia_A at bc008000, depth=8, pitch=1280
Console: switching to colour frame buffer device 160x53
fb0: Open Firmware frame buffer device on
/pci@f000/ATY,XiaParent@10/ATY,Xia_A
no framebuffer address found for /pci@f000/ATY,XiaParent@10/ATY,Xia_B
...

o What happens when I do XFree86 -configure:

XFree86 Version 4.2.1 / X Window System
(protocol Version 11, revision 0, vendor release 6600)
Release Date: 3 September 2002
If the server is older than 6-12 months, or if your card is
newer than the above date, look for a newer version before
reporting problems.  (See http://www.XFree86.Org/)
Build Operating System: Linux 2.4.19-16mdk ppc [ELF]
Module Loader present
Markers: (--) probed, (**) from config file, (==) default setting,
 (++) from command line, (!!) notice, (II) informational,
 (WW) warning, (EE) error, (NI) not implemented, (??) unknown.
(==) Log file: "/var/log/XFree86.0.log", Time: Wed Dec  4 23:26:09 2002
Using vt 7
List of video drivers:
atimisc
r128
radeon
mga
glint
s3virge
sis
savage
trident
chips
tdfx
fbdev
ati
vga
nv
v4l

Fatal server error:

XFree86 has found a valid card configuration.
Unfortunately the appropriate data has not been added to xf86PciInfo.h.
Please forward 'scanpci -v' output to XFree86 support team.

When reporting a problem related to a server crash, please send
the full server output, not just the last messages.
This can be found in the log file "/var/log/XFree86.0.log".
Please report problems to [EMAIL PROTECTED]


Tell me if I can do anything else to pinpoint the problems. But don't feel
pressed, Linux is not critical for me at the moment.

 -Christian

Re: Graphical Install on 1280x854 Powerbook?

[Christian Walther]

> If you happen to be up and running, could you post the output of:
> cat /proc/cpuinfo

I'm halfways up and running. Enough to have a look at cpuinfo. Here:

cpu : 7455, altivec supported
temperature : 53 C (uncalibrated)
clock   : 667MHz
revision: 3.2 (pvr 8001 0302)
bogomips: 665.19
machine : PowerBook3,5
motherboard : PowerBook3,5 MacRISC2 MacRISC Power Macintosh
detected as : 95 (Unknown Keylargo-based)
pmac flags  : 
L2 cache: 256K unified
memory  : 256MB
pmac-generation : NewWorld

(That 667MHz puzzles me. This machine is supposed to run at 867MHz - is this
some power saving feature that only MacOS knows how to turn off?)

As for the halfways: my install wasn't entirely successful. I did a text
install first with the 8.2 CDs, then with the 9.1 cooker install ISO
(installing packages from the 8.2 CDs), and both gave me a segmentation
fault when I dismissed the dialog that lets you customize the various boot
options for yaboot. I then booted from the rescue image and used mkofboot
and ybin to install and configure yaboot manually, and my system is now
bootable and seems functional. But I'm far away from X: It seems that the
kernel doesn't even recognize the Radeon card, with both Mandrake kernels
(2.4.18 from 8.2 and 2.4.19 from the cooker mirrors) and a bk 2.4.20 kernel
from ppckernel.org. According to something I read on some yellowdog mailing
list archive, dmesg should contain some lines starting with "radeonfb",
which doesn't happen for me.

 -Christian

Re: Graphical Install on 1280x854 Powerbook?

[Christian Walther]

> There is radeon support for the card/display in the first TiBooks that
> used radeon, that was the latest machine (& th eLCD iMac) at the time 8.2
> was created. I have an Xconfig for the 1280x854 if you want it, but I
> don't think it will help you in terms of the install.

Yes, there have also been reports by other people on the list who had got X
to work, so I'm confident that I'll get that far as well, sooner or later.

As I said, it was just a short question out of curiosity - I don't mind
doing a text install.

> The pending installer beta will be using XFree by default, rather than
> Xpmac, I'll add the 1280x854 config.  If you could test it, that would be
> wonderful.

Sure, just post a notice to the list when it's ready.

 -Christian

Graphical Install on 1280x854 Powerbook?

[Christian Walther]

Has by accident anyone managed to get a graphical install on a 1280x854
Powerbook (867MHz, in my case)? I didn't try very hard yet, and I'm not
afraid of doing a text install either, but I thought a quick question
couldn't hurt. My experiences up to now are that with install-novideo I get
a freeze at an X startup screen in false colors (what should be black
appears blue, white appears black), and with various ATI modes (is radeon
the right one for the Radeon Mobility 9000 in this machine?) I get a
responsive display, but which is horizontally stretched by a factor 4 and
overimposed on itself (and the colors don't seem to be right as well).

 -Christian

Re: Half-broken german keyboard - mine is full-broken

[Christian Walther]

> Another user provided what was supposed to be an improved X keymap for german
> keyboards, which I patched in:
> 
> * Thu Mar 28 2002 Stew Benedict <[EMAIL PROTECTED]> 4.2.0-10.2mdk
> 
> - updated de_CH keymap - Christian Walther

My keymap is for swiss german  (CH for Confoederatio Helvetica, ahem)
keyboards, which are not exactly the same as german ones. This is what we
call cultural diversity here in europe...

 -Christian

Re: 8.2 PPC Errata -- Request for comments

[Christian Walther]

> 5) The BootX installer package doesn't work with languages other than English.
> Error scenario: The BootX installer is an AppleScript that is used for
> automating the BootX settings; when it is run on a Macintosh that uses a
> language other than English, an error message appears: "Sorry, can't setup
> BootX". This is because the script looks for the Mac's "System folder" (in
> English) which isn't present on systems that use a different language.
> Solution: The BootX app can still be used, but you need to manually enter the
> BootX entries instead of using the Applescript wrapper. See these pages for
> more info:
> http://www.linux-mandrake.com/en/demos/PPC/Install/pages/install8B.php3

That's true, but unfortunately that's not the only problem with the
applescript. It doesn't work for me even though I'm on a US system (9.2.2).
That's because it doesn't find the settings file that it should copy into
the system folder, as apparently AppleScript doesn't know where to look if
you just tell it to do something with «file "x" of folder "y"». It seems to
expect an absolute path, which can be constructed e.g. by saying «file "x"
of folder "y" of insertion location». See the alternative script I've
written

 http://www.n.ethz.ch/student/walthec/mandrake-install.sit

and the thread about it

 Date: Wed, 01 May 2002 13:01:41 +0200
 Subject: Applescript


Another problem is that BootX's ramdisk file location setting is only
correctly setup if the user by accident has the file in the same place (same
absolute path) as Stew had it when he made the settings file. I don't know
any way of correctly automating this, but one can at least tell the user how
to manually set this up.

 -Christian

Re: Applescript

[Christian Walther]

> The ramdisk size is setup by the current Applescript

Size yes, but not location. The location of the ramdisk is stored as an
absolute path in the BootX Settings file, so this only works if the user by
accident has the ramdisk in the same place as you had it when you made the
settings file, otherwise BootX just unchecks the ramdisk checkbox.

Applescript

[Christian Walther]

Because of the various flaws of the OldWorld-Installer-Applescript I've
decided to write my own version:

 http://www.n.ethz.ch/student/walthec/mandrake-install.sit

Have a look at it. It should have no problems with international systems, it
should find the necessary files on its own (and asks for them if it
doesn't), and it tells the user how to specify the RAM disk in BootX
(because this can't be automated). It works well on my PowerBook G3
Wallstreet with MacOS 9.2.2 US, but I haven't tested it on any other system.

 -Christian

Re: error -39

[Christian Walther]

>> I have a PowerMac 440/160, when I lunch "Mandrake Linx Install.sit", I have
>> an error -39... what do I must do ??
>> 
> 
> I've seen this on the 7600 too.  Not sure what it means, but I was able to
> open the .sit by dropping it on Stuffit Expander, or opening Stuffit
> Expander and then selecting the file.

For some strange reason the file "Mandrake Linux Install.sit" has type
"APPL", so the Finder thinks it is an application and tries to execute it
when you doubleclick on it. The file should have type "SIT5" and creator
"SIT!". Do you know what I'm talking about, Stew? As far as I remember you
are making the HFS filesystem of the hybrid CD in Linux, so probably there
is something misconfigured there to give files ending in ".sit" type "APPL"
and creator "STi0".

 -Christian

Re: Maybe very stupid newbie question, but...

[Christian Walther]

> ... it always came out as a document, not CD image, and consequently wasn't
> recognized by DiskCopy.

DiskCopy can't mount ISO images (neither StuffIt/ShrinkWrap, by the way). At
least the OS 9 version, I haven't tried in OS X. Toast is the only program I
know that can.

> The problem seems to be that the .iso suffix is
> somehow known neither to my internet config nor to the ftp programmes. I
> could add the .iso suffix to my suffix mappings, but don't know what to
> put as "type of file" and "creator".

To have them opened by Toast, use Type "iImg" and Creator "CDr3".

But be warned when trying to mount the images: it has been said on this list
that they could be corrupted by mounting, making them un-bootable. I haven't
tried this, so I don't know. I've had no problems however with setting the
files "locked" in the Get Info dialog before mounting them.

Anyway, just check the md5 sums before burning the CDs. A utility for Mac
that can do this is Shorten, available from
http://www.hornig.net/shorten.html

 -Christian

Re: Crashes partout

[Christian Walther]

> In the case of Gnome, the 'bonobo-moniker' keeps crashing. Whatever that is.
> And of course, that brings down other things such as the background applet,
> etc... But at least Gnome doesn't really crash. Anyway, that's unacceptable.

I'm having this problem too. Everytime I start Gnome, I get a dialog saying
that "bonobo-moniker-archiver" crashed.

It might be that I am not using the most recent Gnome packages, so this
might be fixed already, but I don't have Linux running at the moment (and I
don't have time to start it) to check.

 -Christian

Re: Keyboard (Mac keycodes) issues

[Christian Walther]

> I must admit I'm not an expert in keycodes, but anyway, what I wrote
> works! ;-) 
> Examining your proposition, it seems you can either use macintosh and
> macintosh/xx or macintosh_old and xx, respectively for XkbModel and
> XkbLayout. But you can't use macintosh_old and macintosh/xx it seems, as
> you reported.

Well, my experience is a different one. I've tested all 4 possibilities:

o XkbModel macintosh, XkbLayout de_CH:
I get a screwed up keyboard (as expected, because Mac keycodes are being
generated and then interpreted as Linux keycodes), and the X server doesn't
show any errors (in its console output, you have to "startx" manually from
runlevel 3 to see this).

o XkbModel macintosh, XkbLayout macintosh/de_CH
o XkbModel macintosh_old, XkbLayout macintosh/de_CH
(same behavior for both):
I get a semi-functional swiss german keyboard: plain and shifted special
characters work, but not option-ed. The X server says:

The XKEYBOARD keymap compiler (xkbcomp) reports:
> Error:Can't find file "macintosh/macintosh/de_CH" for symbols include
>   Exiting
>   Abandoning symbols file "default"

I don't really understand this one. How does it know that I'm using a swiss
german keyboard if it can't find the file that should tell it so?

o XkbModel macintosh_old, XkbLayout de_CH:
Everything works, no error message from the X server.

Maybe we are talking about something different? I'm using Mac keycodes for
all examples; my XF86Config-4 keyboard section is as follows:

Section "InputDevice"

Identifier "Keyboard1"
Driver  "Keyboard"
Option "AutoRepeat"  "250 30"

Option "XkbRules" "xfree86"
Option "XkbModel" "macintosh_old"
Option "XkbLayout" "de_CH"

EndSection

and the relevant lines in xkb/rules/xfree86 are

! model =   keycodesgeometry
  macintosh =   xfree86 macintosh(macintosh)
  macintosh_old =   macintosh   macintosh(macintosh)

! model layout  =   symbols
  macintosh us  =   macintosh/us(extended)
  macintosh en_US   =   macintosh/us(extended)
  macintosh *   =   macintosh/us(extended)+macintosh/%l%(v)
  macintosh_old us  =   macintosh/us(extended)
  macintosh_old en_US   =   macintosh/us(extended)
  macintosh_old *   =   macintosh/us(extended)+macintosh/%l%(v)


> Setting the XkbModel to macintosh for linux keycodes doesn't give
> exactly the same result as pc105, so i prefer pc105.

Of course. macintosh gives you a more Mac-like keyboard, and pc105 gives you
a more PC-like.

Keyboard (Mac keycodes) issues

[Christian Walther]

I just finished configuring my keyboard to my full contentment, but it
involved a bit of hacking on which the experts out there might want to
comment. I'm quite a newbie concerning Linux, so I might have done some
things in a less-than-optimal way.

I want my keyboard (Swiss German layout, by the way) to behave as I'm used
to, i.e. as close to MacOS as possible. So I first switched to Mac keycodes,
following Sylvain Obegi's guide: dev.mac_hid.keyboard_sends_linux_keycodes=0
in /etc/sysctl.conf.

Then on to the console keymap: As there was no swiss german keymap in
/usr/lib/kbd/keymaps/mac/, I first used a german one (mac-de2-ext or
mac-de-latin1-nodeadkeys, can't remember), then one that I once had
downloaded somewhere and hand-tweaked that to suit my needs. Everything is
working fine now with that one, I can write special characters (among them
important ones like @, |, {}, #) like in MacOS using the Option key that I
mapped to AltGr, but there is one thing that I don't understand: At bootup
it displays

  Loading keymap: mac-de_CH-latin1 [  OK  ]
  Loading compose keys: compose.latin1.inc [  OK  ]
  The BackSpace key sends: ^?  [FAILED]

What does this "[FAILED]" mean? I always thought that thing about the
BackSpace key was an informational message only, how can this fail? This
happened with the the supplied mac keymaps too, but not when using Linux
keycodes.


Second, X keymap. Sylvain's guide says:

> in /etc/X11/XF86Config-4
> Option «XkbModel»
> pc105 (default) : Linux keycodes
> macintosh : mac keycodes
> 
> Option «XkbLayout»
> xx (default) : keymap for Linux keycodes
> macintosh/xx : keymap for mac keycodes (is it exists, see
> /etc/X11/xkb/symbols/macintosh/)

This is (in my opinion - but I don't understand that whole xkb system very
well) not entirely correct.

First, if you are using Linux keycodes, you can use "macintosh" as XkbModel
as well, because ...xkb/rules/xfree86 maps model "macintosh" to keycodes
"xfree86". (But this still doesn't give you a fully usable MacOS-like
keyboard, because the Mac keyboard has no AltGr key and thus you can't reach
the secondary characters. Also, the Command key doesn't seem to be mapped to
anything (but I couldn't verify this as I didn't have xev installed).)

Second, if you are using Mac keycodes, you should not use XkbModel
"macintosh", but "macintosh_old", because ...xkb/rules/xfree86 maps model
"macintosh" to keycodes "xfree86" (Linux keycodes) and "macintosh_old" to
keycodes "macintosh" (Mac keycodes).

Third, if you specify "macintosh/xx" as XkbLayout, the X server complains
about not finding "macintosh/macintosh/xx". You just have to use "xx"
because the "macintosh/" directoy part gets added automatically by
...xkb/rules/xfree86.

Except for those errors, Sylvain's guide was very helpful.

I made some changes to the supplied ..xkb/symbols/macintosh/de_CH as well to
make it as faithful as possible to the MacOS keyboard behavior. The most
important one is that the key that it calls  (the additional key on
european keyboards, Mac keycode 18, produces [section, degree] in swiss
german layout) is actually called  (in the keycodes file and in all
other european symbol files).

 -Christian

Re: some apps launch from shell, but not menu

[Christian Walther]

> some apps will launch from a shell or the "run command" dialog but not from
> the menu... particularly galeon and mozilla do this.

Same thing for me. I've just deleted "soundwrapper" from the menu command in
menu editor, and now it works. I've never tried playing sound in mozilla
though.

 -Christian

Re: PB t get CD boot from mac 9.1

[Christian Walther]

> But when i try to burn CD ROM with toast, it does work .
> 
> First i see that name is too long (BEWARE ANY NAME event in CD should be
> less than 32 caracters).
> two what option should i use ? (alias resolution ?) and repertorie name
> ? )
> 
> What is the name of CD ROM (? i get CD ROM ?)
> 
> 
> did you try to make cd under 9.1 ? How to do this correctly ?
> 
> Could you please tell my your option ? Why don't you explain on you web
> how to to with standard ?
> 
> i use a firewire burner. it could running under 9.1 our mac os X. i have
> toast titanium
> for 9.1.

Probably you tried to burn a MacOS CD containing the image file instead of
burning the image directly.

You have to choose "Disc Image" from the "Other" menu (rightmost of the blue
buttons) and then drag the image file into the window. It should then say
"649.5 MB Mac/ISO Hybrid CD-ROM".

 -Christian

Re: Hostname, MOL CD

[Christian Walther]

>> o Does Linux have any way of knowing the hostname that was set in MacOS X??
>> When starting Linux, my Powerbook suddenly had the hostname "CWalthers
>> PowerBook G3" that I had given it in MacOS X instead of the normal
>> "localhost", and I'm pretty sure that I had never mentioned "CWalthers
>> PowerBook G3" to Linux (because I suspected that Linux would not support
>> hostnames with spaces in them, while I expect MacOS (X) not to allow me to
>> do things that don't work).
> 
> The only way I know that would be picked up is if it's provided by a dhcpd
> server.

Hmm - I use a DHCP server - but why on earth would the DHCP server of my
internet provider set the hostname to something I made up? In fact, my
hostname usually gets set to something like "dclient217-162-184-113" by the
DHCP server.

>> o I haven't been able to get Mac-On-Linux to recognize my CD drive. I added
>> "blkdev: /dev/hdc -cd -boot" to /etc/molrc, and all I get when starting MOL
>> is 
>> > (disk_open) Opening /dev/hdc1: No such device
>> ...
> 
> Never tried booting MOL from CD.  I have the same entry here, without
> -boot.

I've now got it to work by making /dev/cdrom a symlink to /dev/hdc and
specifying "blkdev: /dev/cdrom -cd -boot" in molrc. It seems that MOL
doesn't really look at /dev/hdc when you tell it to, but only the individual
partitions /dev/hdc1.../dev/hdc15.

Hostname, MOL CD

[Christian Walther]

Two issues that I recently encountered with 8.2b1 on my Wallstreet
Powerbook:

o Does Linux have any way of knowing the hostname that was set in MacOS X??
When starting Linux, my Powerbook suddenly had the hostname "CWalthers
PowerBook G3" that I had given it in MacOS X instead of the normal
"localhost", and I'm pretty sure that I had never mentioned "CWalthers
PowerBook G3" to Linux (because I suspected that Linux would not support
hostnames with spaces in them, while I expect MacOS (X) not to allow me to
do things that don't work).

This caused another problem: When I tried to start KDE, it failed
complaining about not being able to open some file ".DCOPserver..." (can't
remember the exact name, but it had the new hostname in it). This problem
vanished when I set the hostname back to "localhost".

o I haven't been able to get Mac-On-Linux to recognize my CD drive. I added
"blkdev: /dev/hdc -cd -boot" to /etc/molrc, and all I get when starting MOL
is 
> (disk_open) Opening /dev/hdc1: No such device
> (disk_open) Opening /dev/hdc2: No such device
> (disk_open) Opening /dev/hdc3: No such device
...and so on until...
> (disk_open) Opening /dev/hdc15: No such device

I tried switching off supermount, but that didn't change anything.

The CD drive is definitely /dev/hdc, as I could even mount the MacOS
installation CD from there to copy out the MacOS ROM file.

MOL works fine when booted from my MacOS 9.2.2 HD, apart from the fact that
it starts in 8-bit video and the color translation to my 32-bit X window is
somewhat broken - what should appear white appears black, what should appear
black appears black too, and the other colors are wrong as well. But that's
probably not a Mandrake problem.

 -Christian

Re: install problems

[Christian Walther]

> I am not able to boot the 8.2 beta cdrom.  I verified the checksums of the
> iso's prior to burning and they appeared to burn ok.  If I hold the "C" key,
> the mac os (8.1) on the hd still boots.  If I hold the "C" key and have the
> mac os cd in, the mac os cd boots.
> 
> HW Config:
> PowerMac 7600/132 with 32MB RAM, 1 SCSI HD, 1 SCSI CDROM.  I have tried to
> boot with SCSI HD at ID=0 and again at ID=1.  SCSI CDROM at ID=1 and later
> at ID=0.  No IDE drives, only SCSI.

Since nobody of the more experienced people seems to have heard your
question, I'll answer what I can:

The Powermac 7600 is Old-World, which basically means it has a big ROM chip
containing half of MacOS, and some older version of the boot code that
doesn't support booting non-MacOS directly.

You have to boot into Linux from MacOS using BootX, which you can find in
"Mandrake_Linux_8.2_Install.sit" in the "BootX" folder on the CD.

If you need additional instructions (on using BootX), just ask.

 -Christian

Re: linux ppc 8.0 .iso's..

[Christian Walther]

> I just downloaded what i thought was a disk image of a cd
> (MandrakePPC-8.0-inst.ppc.iso) but it comes up as a netscape file which
> is pure code.

That does not necessarily mean it's corrupted. Check the MD5 sums! A utility
for mac that can do this is Shorten, available from
http://www.hornig.net/shorten.html
If it is corrupted, I'd also recommend downloading it by FTP. Just make sure
it's downloaded as "binary", not as "text".

> stuffit expander won't mount the disk image, help?

Yes, DiskCopy neither, but Toast does.

Re: 8.2b1 on Wallstreet problems

[Christian Walther]

> If you were using SCSI drives, you would need the initrd.  You can get
> around it by specifying root=/dev/XXX in the kernel arguments.

Ah, OK. Thanks for the info.

> Gnome is known to be broken in the beta.  Apparently it was on x86 too,
> and I inherted some of their issues, plus some libraries I omitted on the
> CD's.

OK. I heard of that, but thought it was a different problem because of the
strange screen flashes, and because the same thing happens with Sawfish, but
let's see.

>> I log out of KDE and try to log into Gnome, but this bounces me back to the
>> login window after some screen flashes and colored vertical stripes. The
>> same thing when I try Sawfish.

>> I notice that I can't change consoles with cmd-ctrl-Fx

> No "Command" now with Linux keycodes.  It's just like a PC. Ctl-Alt-F2,
> etc.

Ah, right. I didn't think of that because while installing it worked. I'll
change back to Mac keycodes anyway because I'd like my keyboard to behave as
I'm used to...

 -Christian

8.2b1 on Wallstreet problems

[Christian Walther]

Hello!

I just tried installing Mandrake 8.2b1 on my PowerBook G3 Wallstreet
(OldWorld), and I've run into some problems that you might like to know
about.

First, the installer applescript doesn't work because the Finder doesn't
find the "Mandrake Linux Install" folder (which by the way is called
"Mandrake Linux 8.2 Install", but fixing that doesn't help). This can be
cured by replacing 'folder "Mandrake Linux Install"' by 'insertion location'
(2 times). This is not the most elegant solution, because "insertion
location" is simply the active window which may or may not be the desired
folder, but I don't know any way of getting the folder that contains the
script file at the moment.

I run the applescript, choose "Install/Rescue" - "Install" - "2.4" -
"Graphical" - "atyfb". When BootX starts, I notice that the "More kernel
arguments" line contains "ramdisk_size=32000", although you recommend 36000.
I change this and go on.

Linux starts booting, but stops with "Kernel panic: VFS: Unable to mount
root fs on 03.09", " <0>Rebooting in 180 seconds.."

I examine BootX's settings a little more and discover that it has no ramdisk
set. I add "all.gz" from the "Linux Kernels" folder in the "Options" dialog
and try again.

X starts and seems to work OK, apart from the familiar flickering that
surprisingly disappears after changing to one of the text consoles and back.

In the "Choose your language" dialog, I click "Advanced" because I want to
install additional files for swiss german. This makes the dialog fill the
whole screen height, and there is no scrollbar for the "other languages"
checkboxes, so that my desired swiss german is somewhere unreachable way off
the screen. The lowest I can see is "Euskara (Basque)", so I have to count
the offset from this to "German|Switzerland" in the list above and use the
arrow keys to blindly reach it.

I choose Expert Install, choose a "Swiss (German layout)" keyboard layout,
partition my hard drive and install the default set of packages. When it
comes to X Configuration, I choose XFree 4.2.0, confirm the display settings
iMac/PowerBook, 1024x768, 32bit, and choose to start X automatically.

Back in MacOS, I tell BootX to use kernel "vmlinuz-2.4.17-17mdk" that I
copied from the CD into "Linux Kernels". I'm not sure about what to do with
"initrd-2.4.17-17mdk.img", because if I select it as RAM disk in BootX, I
can't set the root device because that field is then used for ramdisk size.
So I leave it away, enter my root device and delete everything except
"video=atyfb:vmode:17" from "More kernel arguments".

Linux seems to boot OK, apart from complaining about some "Unresoved
symbols" in "Finding module dependencies" and at "Starting pcmcia":
"cardmgr[614]: config error, file './config.opts' line 8: no function
bindings".

I log in and choose to run Gnome. In the First Time Wizard I click "Cancel",
which starts KDE instead of Gnome. While starting, KDE displays an error
message "Error while initializing sound driver: device /dev/dsp can't be
opened (Device or resource busy)".

I log out of KDE and try to log into Gnome, but this bounces me back to the
login window after some screen flashes and colored vertical stripes. The
same thing when I try Sawfish.

I notice that I can't change consoles with cmd-ctrl-Fx. Repeating this with
a console window open in KDE shows that keys F1 to F4 produce numbers 0 to
3, F5 to F10 produce "~", F11 and F12 produce mouse clicks (as configured).

When booting the second time and logging into KDE (Gnome still doesn't
work), it displays that message about the sound driver again, but plays its
welcome sound nevertheless.


That's all for the moment - hope you can make any sense of it.

 -Christian