Re: slink: Which directory am I supposed to install from?
On Tue, 25 May 1999, Marc Haber wrote:
>On Mon, 24 May 1999 14:42:17 +0100 (BST), you wrote:
>>On Mon, 24 May 1999 Marc.Haberemail@example.com wrote:
>>OK, that makes things a little clearer. Anyway, checking on the CD,
>>/install/boot.bat and /dists.../install.bat do effectively the same job.
>Why are they named differently then? This is more than only mildly
To be honest, I didn't even know install.bat existed until now. boot.bat
is created by my CD creation scripts at runtime. I'll fix this
>>This confuses me a little - how are you getting a kernel without the SCSI
>>drivers from one but not the other?
>This is what I am asking :-)
>>Ah, hang on - you're not using the
>>laptop (tecra) kernel on one of these are you?
>How do I select this? No, I am not aware of having done different
>procedures with both installation tries. Does the script silently
>decide to install a "smaller" kernel?
>When starting /install/boot.bat, I need the CD (or the files from
>/dists/.../disks-i386/current/) to install kernel and device driver
>modules (it seems to need drv1440.bin which is not included in
>/install). The system installed by that procedure seems to work
>This installation has this kernel:
>/vmlinuz -> boot/vmlinuz-2.0.36
>715260 May 24 1999 /boot/vmlinuz-2.0.36
>Installing from /dists.../install.bat naturally doesn't need any
>additional files and yields the same installation this time.
OK, which is what I'd expect.
>I have to apologize. Obviously, I must have made some funky mistake
>last week. In fact, both /install and /dists..../disks-i386/current
>seem to yield working systems with NCR support.
That's better. What I think may have happened is you used (by mistake) a
laptop kernel image, which is by necessity made smaller to work on
laptops. To achieve this lots of drivers are removed, by the looks of it
including the NCR SCSI driver.
>>Explanation: the .../disks-i386/ tree is the place where the boot-floppies
>>and documentation are placed first. The CD build script tries to make
>>things easier for new users by adding copies of most of this stuff (docs
>>and boot kernels) in an easier-to-find location (/install). The two should
>>be functionally identical.
>But this script surely doesn't rename a file from install.bat to
>boot.bat, does it? There must be something different going on than
It's not copied, but created. See above.
>>Actually, thinking about things a bit more: why are you copying just the
>>base files to a local hard disk? Are you then using NFS or similar to
>>install the rest?
>Yes. On some systems, I only transfer the debs needed to get the box
>on to the network on the hard disk.
>>If so, just using the boot/root floppy should cope with
>These are two floppies. And I _hate_ floppies. Slow and unreliable. A
I can understand that alright...!
>>>I failed it since it needed files from the
>>>/dists/slink/main/disks-i386/2.1.8-1999-02-22 directory. I thought
>>>"cool, there is a base system too, so I probably don't need that
>>>/install stuff" and started over with the other base system.
>If this way of doing an install is a non-option, it should be clearly
In theory it should work. I'll get this tested a bit better in future.
>Oh btw, dbootstrap's error reporting needs to be improved. While doing
>the experiments, dbootstrap at one time gave me the error message
>"installing device driver floppy failed". This error message is next
>to useless. The install system does not have syslog so there is no
>chance to learn what went wrong. Repeating the mkfs on that partition
>solved the issue, but I still don't know what exactly went wrong.
OK... Probably best to file a bug against boot-floppies.
>Also, the html documents in /install have links to a document called
>"index.html". Such a file does not exist, I suppose that install.html
>is meant instead. Which package do I file that bug against?
That'll be boot-floppies again.
>Also, I feel that there should be an option to skip dselect
>A third thing: dbootstrap denies installing lilo into a logical drive
>and insists on installing it into the extended partition instead. The
>reason is that the standard MBR bootstrap code can't load lilo from a
>logical drive. Some boot managers (most notably my favorite, xfdisk)
>can indeed boot from a logical drive. So I have to manually edit
>/target/etc/lilo.conf and re-invoke lilo to correct dbootstrap. I
>strongly feel that dbootstrap has a point in reminding me that I won't
>be able to boot my system using standard mechanisms, but shouldn't
>deny me my wish. There should be an "I mean it, really" option with
And again. :-)
It might sound silly to open several different bugs against boot-floppies
for these problems, but it's normally the easiest way for maintainers to
Steve McIntyre, CURS CCE, Cambridge, UK. firstname.lastname@example.org
<a href=http://www.chiark.greenend.org.uk/~stevem/>My home page</a>
"Can't keep my eyes from the circling sky, +------------------
"Tongue-tied & twisted, Just an earth-bound misfit, I..." |Finger for PGP key