[Rule-list] Test of 0.6.3-small - the whole story

[Devon]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Thursday 21 February 2002 01:29 pm, Martin Stricker wrote:

> I first tested with mem=10M. It froze somehow while showing
> starting anaconda...
> Happily Red Hat provides four consoles ( switching as usual, [Alt][Fx]
> for x=1 to 4). The first is the install console, the second shows a
> minimal shell, the tird and fourth show different information on what
> the box currently does.
>
> Console 3 shows (in extracts):
> 7972 kB are available
> going to insmod (several kernel modules on several lines)
> looking for usb controllers
> no usb controller found
> (detecting several other hardware)
> looking for USB mouse...

All the above are normal output. I don't believe the installer froze. It 
just doesn't have enough memory to run. In some of my tests, this stage 
has taken quite a long while, and then continued.

> Here the install hangs. The really weird thing is I don't have any USB
> in my computer (heck, it's five years old!), and at the beginning the
> installer correctly detected this. Why does it look for an USB mouse
> now? This might be the point where the install really hangs. Is it
> possible to do an if/else and schecking for USB mice only if there *is*
> an USB controller? This might be worth a bugzilla entry, but first I
> need to check if it also happens with stock Red Hat Linux 7.2 CDs.
> Maybe this evening...

It happens with the stock installer as well. I'll have a look and see 
what can be done about it. I don't think the delay is the result of 
looking for the usb mouse. It is more likely that the search completed, 
and the installer is struggling to complete the _next_ stage.

> Here the inbstall hangs. It furiously spins the CD and the hard disk
> over and over again, always with exactly the same "melody". This kept
> driving me nuts so I turned the box off after half an hour.

I believe this indicates the installer is exceptionally low on memory, 
but still trying it's best to continue. Amazing that it doesn't just 
explode. :)

> Finally I tried mem=12M which worked quite nicely with RULE 0.5.3, but
> here it got trapped at the same place as below! Devon, did you change
> somethimg here? I'll have to check if 0.5.3 also looks for an USB
> mouse... It was the same noise than before, but not completely
> repeated, sometimes there was an additional noise. So I hoped and
> waited and tried to shield my ears against that horrifying noise... The
> console messages were exactly the same as with 10 MB. After about 10
> minutes it really went past  this stage and continued booting! console
> 4 gave no additional messages, but on console 3 appeared Detected 16 M
> of memorv
> Very weird! Then I went on with the install. Deleting the old partition
> was easy, but after I told "fdisk" (wich got me some GUI for the real
> fdisk I wanted) to create the new partition the install hung for
> another four minutes. Within this time console 3 got this message:
> Swap attempt of 96M to 192M

All normal (the delays) when you have stretched the memory limits of the 
installer. The message above is the amount of swap  that the 
autopartition utility would suggest, and is also normal.

> which didn't do anything visible to the size of my swap partition (128
> MB). After that (and the activation of 128 MB swap) all went nice.
> Again I installed only the base install, nothing else. This time, as
> promised by devon, mo unresolved dependencies popped up.

Yes, once we can get swap turned on, things go pretty well.

> Then it copied the installation image to hard disk (which wasn't there
> in 0.5.3 and is part of anaconda/Lepto's capability to install from
> everal CDs). Now I got this error message which wasn't there in 0.5.3:
> WARNING not all packages in hdlist had order tag
> Then the install finally begun.

I saw that warning as well. I believe it is because I included extra 
packages when I built the hdlists. I'll have to look at it more when I've 
had some sleep. :)

> The package install took 22 minutes, about the same time as with 0.5.3.
> The complete installation from turning on the computer until it
> rebooted took 41 minutes, definitely longer than 0.5.3.

Part of the time increase is due to using multiple disks, and the need to 
write the stage2 image to disk, I suspect.

> Booting into the system with mem=12M went fine. df reported:
> 1k Blocks   used       available   used
> 2329540   155825   2055380      8%
> The discrepancy between the total and the used plus available are the
> 5% security space to prevent users from filling the disk. Interesting
> side note: The installer told me it'd install 244 MB, but only 152 were
> used! Weird again...

The same thing happens on a stock install. The size of the install is an 
estimate, and I suspect is designed to leave space to write stage2 to 
disk. (which needs 47M of disk space.) Score one point there for doing 
this as a single cd containing a base install of stock RedHat RPMs.

> That's my story. Devon, Benoit, if you have any questions or any
> suggestions on how to perform the next tests don't hesitate to ask!
> Benoit, I'd like to help with setting up a test plan if you are
> interested.

Thanks for testing. 
We have multiple options at this point, which is good. The 3rd CD works, 
but obviously makes for a bigger download, for one.
The boot/update floppy is nice and small, but will somewhat limit us as 
to how much memory requirement we can shave off.

> I just grabbed the new floppy images. Devon, would you please be so
> kind to include the full RULE installer version number on any later
> file names so I always know which Version I'm testing? Oh, and which
> version did I download right now?

I recreated the files due to a screw up on my part. Since we have taken 
the installer in (yet) another new direction, I bumped up the version 
number to 0.7.0

> I'll ry a 11 MB or 10 MB tomorrow while I'm at work (which means my
> computer at home has about ten hours to get anaconda starting). Maybe
> it eventually does...

I don't think that this version will have any hope of installing at 10M, 
but by all means feel free to prove me wrong. Just look at the file sizes 
on the stock cds in the RedHat/base/ directory.

This is one place we do save on the 3rd CD option, as well as the single 
custom CD. We managed to shave 10M + off hdlist2, and 17M off the stage2 
image. It's a testament to the quality of the stock installer that it 
doesn't just explode when we force it to read in hdlist2 in 12M of 
memory. :)

Stock disks-
# ls -lh /mnt/cdrom/RedHat/base/
total 75M
- -r--r--r--    3 root     root          20k Sep  9 20:52 comps
- -rw-r--r--    2 root     root         1.6M Sep 24 16:22 hdlist
- -rw-r--r--    2 root     root          12M Sep 24 16:22 hdlist2
- -rw-r--r--    3 root     root         6.4M Sep  9 21:43 hdstg1.img
- -rw-r--r--    3 root     root         6.6M Sep  9 21:42 netstg1.img
- -rw-r--r--    3 root     root          47M Sep  9 21:45 stage2.img
- -r--r--r--    1 root     root          301 Sep 24 17:40 TRANS.TBL

Rule-6.2-
# ls -lh /mnt/cdrom1/RedHat/base/
total 46M
- -r--r--r--    1 root     root         3.0k Feb 18 23:34 comps
- -r--r--r--    1 root     root         310k Feb 18 22:19 hdlist
- -r--r--r--    1 root     root         1.8M Feb 18 22:19 hdlist2
- -r--r--r--    1 root     root         6.4M Feb 18 23:48 hdstg1.img
- -r--r--r--    1 root     root         6.6M Feb 18 23:48 netstg1.img
- -r--r--r--    1 root     root          30M Feb 18 23:48 stage2.img
- -r--r--r--    1 root     root          301 Feb 18 23:50 TRANS.TBL

- -D

- -- 

pgp key:  http://www.tuxfan.homeip.net:8080/pgpkey.txt

- --
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE8da04eMAUbzJhSVcRAtVxAJ9LVbdFTB/VCDt5q+CtdeiFNunR1QCfSHXa
UzhT+JJ38/1XLiSMzc7uU70=
=A3uN
-----END PGP SIGNATURE-----

_______________________________________________
Rule-list mailing list
Rule-list at mail.freesoftware.fsf.org
http://mail.freesoftware.fsf.org/mailman/listinfo/rule-list