Your message dated Wed, 08 Sep 2010 03:58:12 +0000 with message-id <E1OtBnY-00059T-RN@ravel.debian.org> and subject line Closing old installation report #364110 has caused the Debian Bug report #364110, regarding installation-report to be marked as done. This means that you claim that the problem has been dealt with. If this is not the case it is now your responsibility to reopen the Bug report if necessary, and/or fix the problem forthwith. (NB: If you are a system administrator and have no idea what this message is talking about, this may indicate a serious mail system misconfiguration somewhere. Please contact owner@bugs.debian.org immediately.) -- 364110: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=364110 Debian Bug Tracking System Contact owner@bugs.debian.org with problems
--- Begin Message ---
- To: submit@bugs.debian.org
- Subject: installation-report
- From: jeremy@jsbygott.fsnet.co.uk
- Date: Fri, 21 Apr 2006 16:32:53 +0100 (BST)
- Message-id: <E1FWxcr-0000jT-Uf@debian>
Package: installation-reports INSTALL REPORT Debian-installer-version: Debian GNU/Linux 3.1 r1 _Sarge_ - Official i386 Binary-1 (20051220) uname -a: Linux wolfgang 2.6.8-2-386 #1 Tue Aug 16 12:46:35 UTC 2005 i686 GNU/Linux Date: 2006, April 14, circa 5pm-7pm (for the fifth attempt) Method: How did you install? What did you boot off? Installation media: 4 CD-ROMs as follows: 1: [Debian GNU/Linux 3.1 r1 _Sarge_ - Official i386 Binary-1 (20051220)] 2: [Debian GNU/Linux 3.1 r1 _Sarge_ - Official i386 Binary-2 (20051220)] 3: [Debian GNU/Linux 3.1 r1 _Sarge_ - Official i386 Binary-3 (20051220)] 4: a custom CD containing (all in its root directory): -- a few other packages, plus recent security updates for 1-3, copied from /var/cache/apt/archives on a working sarge system -- a file Packages.gz built with null overrides (clumsy but it works). I booted off the Binary-1 CD. Machine: rebuilt IBM PC compatible, motherboard Jetway V266DM (socket A) Processor: AMD Duron 700 Memory: 256KB Kingston Value Ram PC3200 Root Device: /dev/hda (ide) Root Size/partition table: Feel free to paste the full partition table, with notes on which partitions are mounted where. My problems with the partitioner are discussed in my main report. Output of lspci and lspci -n: wolfgang:~# lspci 0000:00:00.0 Host bridge: VIA Technologies, Inc. VT8366/A/7 [Apollo KT266/A/333] 0000:00:01.0 PCI bridge: VIA Technologies, Inc. VT8366/A/7 [Apollo KT266/A/333 AGP] 0000:00:11.0 ISA bridge: VIA Technologies, Inc. VT8233A ISA Bridge 0000:00:11.1 IDE interface: VIA Technologies, Inc. VT82C586A/B/VT82C686/A/B/VT823x/A/C PIPC Bus Master IDE (rev 06) 0000:00:11.2 USB Controller: VIA Technologies, Inc. VT82xxxxx UHCI USB 1.1 Controller (rev 23) 0000:00:11.3 USB Controller: VIA Technologies, Inc. VT82xxxxx UHCI USB 1.1 Controller (rev 23) 0000:00:11.5 Multimedia audio controller: VIA Technologies, Inc. VT8233/A/8235/8237 AC97 Audio Controller (rev 40) 0000:01:00.0 VGA compatible controller: nVidia Corporation NV5M64 [RIVA TNT2 Model 64/Model 64 Pro] (rev 15) wolfgang:~# lspci -n 0000:00:00.0 0600: 1106:3099 0000:00:01.0 0604: 1106:b099 0000:00:11.0 0601: 1106:3147 0000:00:11.1 0101: 1106:0571 (rev 06) 0000:00:11.2 0c03: 1106:3038 (rev 23) 0000:00:11.3 0c03: 1106:3038 (rev 23) 0000:00:11.5 0401: 1106:3059 (rev 40) 0000:01:00.0 0300: 10de:002d (rev 15) Base System Installation Checklist: [O] = OK, [E] = Error (please elaborate below), [ ] = didn't try it Initial boot worked: [O] Configure network HW: [ ] Config network: [ ] Detect CD: [O] Load installer modules: [ ] Detect hard drives: [O] Partition hard drives: [E] Create file systems: [O] Mount partitions: [O] Install base system: [O] Install boot loader: [E] Reboot: [O] Comments/Problems: <Description of the install, in prose, and any thoughts, comments and ideas you had during the initial install.> My response to this section has outgrown this template. I attach a separate file: report.htm Install logs and other status info is available in /var/log/debian-installer/. Once you have filled out this report, mail it to submit@bugs.debian.org.Title: Sarge installation reportDebian Sarge 3.1r1 installation report
This report describes my attempts to install Debian Sarge 3.1r1 from CD-ROM on an i386 platform. It describes my problems with the partitioner, various bugs and QA issues both during installation and on the freshly installed system, and my thoughts on improving usability. Though sarge is stable, it has rough edges.
This document is not a tutorial. Rather, it is aimed at the installer team and other debian developers who are in a position to address the issues raised. I lack the time to track each issue in unstable and file several dozen bug reports, but I do hope many of the points will be forwarded to the relevant DDs.
The remaining sections are as follows:
- System details (though much of this report is probably platform-independent)
- Why I ended up needing five installs
- Partitioning problems
- Other problems (during installation or on the freshly installed system)
- Walk-through of the fifth ("production") installation, further bugs, usability comments etc
Overall, I find the installer very impressive: a big step forward for debian. Accepting obvious or default options generally gives good results. Non-obvious choices can have odd effects. Many of my suggestions are aimed at making the experience smoother for people who don't install often. Indeed, it may be their first "serious" experience of debian. Every little improvement can win us new friends! (I am not a DD, but I have been running debian at home since slink. My background is in mathematics.)
System
IBM PC compatible, AMD Duron 700 (Socket A), Jetway V266DM motherboard (based on VIA Apollo KT266A chipset) with onboard sound, MS-8808 Vanta/TNT 2M64 AGP 4x graphics card, 52x CD-ROM, primary IDE 40GB, secondary IDE 3GB.
My aim was to install a triple-boot system: win98 on /dev/hda1, debian on the rest of /dev/hda, and legacy win95 on /dev/hdb. It is for a family home/office desktop that is migrating to debian.
To keep win98 happy I first booted the win98 installer and partitioned the brand-new 40GB disk 25% FAT32(LBA), 75% other, then installed win98. Debian was going to have to work around this.
I used four CD-ROMs:
1: [Debian GNU/Linux 3.1 r1 _Sarge_ - Official i386 Binary-1 (20051220)] 2: [Debian GNU/Linux 3.1 r1 _Sarge_ - Official i386 Binary-2 (20051220)] 3: [Debian GNU/Linux 3.1 r1 _Sarge_ - Official i386 Binary-3 (20051220)] 4: a custom CDThe custom CD contained (all in its root directory):
- a few other packages, plus recent security updates for disks 1-3, copied from /var/cache/apt/archives on a working sarge system;
- a file Packages.gz built by "dpkg-scanpackages" with null overrides (a bit clumsy but it works).
Why five installs?
The default install (option "linux") was a breeze. It produced a good, working system. However, in trying to improve on my initial choices, I ran into problems with the partitioner. In sorting them out, I had to install more than once, and to make this less boring I tried each of the options "linux", "expert", "linux26", "expert26". Having learnt various things - and feeling that there were enough issues to talk about - I did a final "production" install which this report is mostly about.
During "run 1" (the default "linux" install) I set up the following partitions. [Note that this table was produced under a 2.4 kernel. It turns out that 2.6 reports a different, better geometry. The C/H/S figures are: BIOS 19679/16/255, sfdisk (2.4 kernel) 4998/255/63, sfdisk (2.6 kernel) 79656/16/63. Multiplying up, you find that only linux2.6 wants to use all 41,110,142,976 bytes on the disk. But it is forced to co-operate with what win98 left behind. Meanwhile the BIOS can't multiply and claims the size is 41111 MB. Weird stuff. Moral: linux is a better bios than the bios.]
wolfgang:~# sfdisk -l /dev/hda Disk /dev/hda: 4998 cylinders, 255 heads, 63 sectors/track Units = cylinders of 8225280 bytes, blocks of 1024 bytes, counting from 0 Intended Linux Device Boot Start End #cyls #blocks Id System mount point /dev/hda1 * 0+ 1249 1250- 10040593+ c W95 FAT32 (LBA) end: (c,h,s) expected (1023,254,63) found (225,254,63) /dev/hda2 1250 1272 23 184747+ 83 Linux / /dev/hda3 1273 4997 3725 29921062+ 5 Extended /dev/hda4 0 - 0 0 0 Empty /dev/hda5 1273+ 1368 96- 771088+ 82 Linux swap / Solaris /dev/hda6 1369+ 1880 512- 4112608+ 83 Linux /usr /dev/hda7 1881+ 2136 256- 2056288+ 83 Linux /var /dev/hda8 2137+ 3460 1324- 10634998+ 83 Linux /home /dev/hda9 3461+ 3716 256- 2056288+ 83 Linux /tmp /dev/hda10 3717+ 4997 1281- 10289601 83 Linux /reservedI was not satisfied for three reasons: I wanted to repartition (to make my swap space a primary not a logical partition), I wanted to make (hd0,1) active and boot from there (not from the mbr, as I don't trust windows not to wreck it), and I wanted to install with a 2.6 kernel (simpler than upgrading).
Moreover, I wanted to preserve the /reserved partition across this process. [This partition was big enough for three full backups of the 3GB /dev/hdb drive, and enabled me to troubleshoot some horrible win95 problems risk-free, a separate project.]
The next three install runs made poor partition tables (whilst flawlessly shrinking win98 to 1000 cylinders). "Run 5" was the "production" install described in detail later. Apart from the geometry (not debian's fault), this gave the partition table I wanted:
wolfgang:~# sfdisk -l /dev/hda | head -7 Disk /dev/hda: 79656 cylinders, 16 heads, 63 sectors/track Warning: extended partition does not start at a cylinder boundary. DOS and Linux will interpret the contents differently. Warning: The partition table looks like it was made for C/H/S=*/255/63 (instead of 79656/16/63). For this listing I'll assume that geometry. wolfgang:~# sfdisk -l -uC -C4998 -H255 -S63 /dev/hda Warning: HDIO_GETGEO says that there are 16 heads Disk /dev/hda: 4998 cylinders, 255 heads, 63 sectors/track Units = cylinders of 8225280 bytes, blocks of 1024 bytes, counting from 0 Device Boot Start End #cyls #blocks Id System /dev/hda1 * 0+ 999 1000- 8032468+ c W95 FAT32 (LBA) /dev/hda2 1000 1032 33 265072+ 83 Linux /dev/hda3 1033 1128 96 771120 82 Linux swap / Solaris /dev/hda4 1129 4997 3869 31077742+ 5 Extended /dev/hda5 1129+ 1768 640- 5140768+ 83 Linux /dev/hda6 1769+ 1864 96- 771088+ 83 Linux /dev/hda7 1865+ 3214 1350- 10843843+ 83 Linux /dev/hda8 3215+ 3716 502- 4032283+ 83 Linux /dev/hda9 3717+ 4997 1281- 10289601 83 LinuxProblems with the partitioner
The front end is pleasant to use. The smileys and the shaggy dog symbol are intelligently chosen. But results were mixed. The partitioner did a good job on "run 1" (when keeping only hda1) but did rather poorly when asked to keep a logical partition.
(wishlist) It should say that GB/MB is the disk-vendor's 10-power GB/MB, not the RAM-vendor's 2-power GB/MB. People who don't partition disks every day might not know this. I had to verify it with a pocket calculator: by creating a partition of specific size in MB, and comparing this with free space in C/H/S before and after. Tedious.
(wishlist) It would be nice if we could specify partition sizes in (whole numbers of) cylinders; if this option exists, I couldn't find it. To achieve that result, I had to use the pocket calculator again.
(bug) A strange error: 4129.090560 MB was rejected as "invalid disk size" for partition hd8, but 4.129090560 GB was accepted.
(bug?) It did a poor job of adding a primary partition whilst keeping the extended partition. Starting with the partition table reproduced above, I should be able to delete hda5 and use that free space to make another primary partition at hda3. The d-i partitioner seemed to let me do this (it looks especially natural on screen, because the d-i partitioner doesn't actually display the extended partition, only the logicals) but the primary came out as hda4. The installation produced a working system, with this hda4 mounted as swap. However:
wolfgang:~# cfdisk /dev/hda FATAL ERROR: Bad logical partition 6: enlarged logical partitions overlap Press any key to exit cfdiskNow, it shouldn't be hard to do this correctly. If I understand the mbr format, adding a third primary should only mean moving the 16 byte block starting at offset 0x1de to 0x1ee to make room for it.
(bug?) It did a poor job of adding logical partitions whilst keeping a later logical partition. The renumbering of logical partitions is irritating - isn't it possible to add into the middle of a linked list and renumber the partitions in disk order? But a worse problem was that I was able to create one logical partition inside another, so the data started at offset 2x63 sectors, not the normal 63. I don't recall the exact sequence which achieved this unhappy feat.
In the end, to get the partition table I wanted, I had to delete all the logicals, including (scary!) the one I wanted to keep. Then I could build them up in disk order, using my pocket calculator again and checking the C/H/S data carefully, to recreate exactly the partition that I had wanted to keep and reuse its data. As far as I can tell, this has worked without data loss. But I did find it scary.
I think improvements in these areas will help to win over people who don't partition disks every day, have a working winXX they value, and find this whole thing scary.
Other problems
This section describes non-partitioner issues arising from the install process, and some QA issues and bugs on my freshly installed system.
(Expert mode: add module parameters?) If I answer "no" each time, I get asked this question three times. That seems correct: I might have parameters for only one stage. But (if my memory is correct) when I answered "yes" the first time, I didn't get asked yes/no later, but got asked direct for the parameters. I'm not sure this behaviour is natural. (But if experts find it convenient, fine.)
(auto dial-in) On one of my trial runs (let's say "run 2"), I let the installer set up PPP for me. I got a bit nervous when it dialled in. I'm used to having a good look around /etc before connecting a new box to the internet. I'm sure the d-i authors were careful, and I'm sure they're more security-aware than I am. To start with, of course, there's almost no software installed. Nothing to worry about? And as we keep going, and install more and more stuff whilst still online?
Suggestion: it would be nice if the installer put up a note explaining exactly what its exposure to the internet will be. Then I can choose more intelligently whether to use this facility.
(CD changing) On those installation runs when I said "no" to PPP, aptitude installed a vast bunch of packages with minimal CD changing. On "run 2" though, I allowed it to get a packages file from security.debian.org before turning the modem off. Even though I had (almost?) all of those packages on my custom CD, aptitude gave connect errors and only installed handfuls of packages at a time. I accept this is my fault. Maybe it's not worth investigating. Was the problem that security.debian.org was higher up /etc/apt/sources.list than the custom CD?
(possible filename issue) Or was it that the filenames on the CD were suboptimal? I copied them from /var/cache/apt/archives on another machine, and lots of filenames contain "%3a" where I would expect to see ":". Updating using those packages gives dpkg output like this:
Preparing to replace kview 4:3.3.2-2sarge3 (using .../kview_4%3a3.3.2-2sarge4_i386.deb) ...I wonder whether jigdo will recognise those "%3a" files when I make ISO for the next revision of stable, or whether I will need to manually rename them.
(console keymap) This problem has not arisen since my final install - I can't reproduce it - but I did notice it on one of the earlier systems. Error messages appeared during bootup, approximately as follows:
unknown keysym "kappa" /etc/console/boottime.kmap.gz: 661: syntax error syntax error in map file key bindings not changed Problem loading keymap Use install-keymapand sure enough, the console wasn't set to "uk" keyboard after all. Running "dpkg-reconfigure console-data" seemed to fix it. Anyhow, my production install hasn't suffered this bug. Perhaps it was due to booting randomly between 2.4.27-2-k7 and 2.6.8-2-386 kernels? (But I had to, to learn the disk geometry!)
(logfile messages) A few lines from "cat -n /var/log/debian-installer/messages" don't look great. They may be normal, of course. I don't know.
696 umount: /target/dev/pts: Invalid argument 697 umount: /target/dev/shm: Invalid argument 698 umount: /target/proc/bus/usb: Invalid argumentIs that right?
879 Reading Package Lists... 880 Building Dependency Tree... 881 console-tools is already the newest version. 882 0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded. 883 perl: warning: Setting locale failed. 884 perl: warning: Please check that your locale settings: 885 LANGUAGE = (unset), 886 LC_ALL = (unset), 887 LANG = "C.UTF-8" 888 are supported and installed on your system. 889 perl: warning: Falling back to the standard locale ("C"). 890 locale 'en_GB' not available 891 Generating locales... 892 en_GB.ISO-8859-1... done 893 Generation complete. 894 Reading Package Lists... 895 Building Dependency Tree... 896 console-tools is already the newest version. 897 console-data is already the newest version. 898 console-common is already the newest version. 899 0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded. (END)Is that normal, or is something being done in the wrong order?
(grub wishlist) When grub detects win9x on /dev/hdb and sets up a boot stanza in /boot/grub/menu.lst, would it be possible to include the following lines before chainloading:- ?
map (hd0) (hd1) map (hd1) (hd0)This is needed for booting win95 off /dev/hdb without the irritating "missing system disk - press any key" message. I imagine this could be a big improvement for some of our end users. Not everyone is up to hunting around grub-doc and doing this for themselves! Early impressions of debian - particularly among possible converts - really count. They won't be impressed if their local guru has to fiddle with things like this after installation.
(ssh) The default Desktop system package selection contained ssh. When "configuring ssh" I answered "no" (don't run sshd) to the debconf question, but "ps -A" later revealed sshd to be running. Is this a bug?
(failure of "dpkg-reconfigure -a") Not strictly an installer bug, but a problem on a freshly installed system.
When I ran that command it got through configuring packages a-g, then:
ERROR: Could not find slib/require.scm in ("/usr/share/guile/site" "/usr/share/guile/1.6" "/usr/share/guile" ".") wolfgang:~#I traced this to package guile-1.6-slib:
wolfgang:~# dpkg-reconfigure guile-1.6-slib ERROR: Could not find slib/require.scm in ("/usr/share/guile/site" "/usr/share/guile/1.6" "/usr/share/guile" ".")So (bug) guile installs brokenly by default. And there's a side effect.
It is rather vexing to crash out of "dpkg-reconfigure -a" because one package won't configure. To proceed with this test, I had to remove guile-1.6-slib (and so gnucash and so gnome, and so I had to fiddle with dependencies to as not to lose a lot more which aptitude now considered un-wanted). Then I had to repeat "dpkg-reconfigure -a" from the beginning.
Wishlist: could dpkg-reconfigure trap errors like that and carry on?
(gdm wishlist) When "dpkg-reconfigure -a" (running on tty2) got to "configuring gdm", I kept gdm as default display manager. No change, but I was abruptly switched to the gdm login screen on tty7 and had to switch back manually. A bit irritating.
(gdm --> configure login manager --> standard greeter)
Bug! Attempting to drag the scrollbar with the swirl logo pops up a window with a red-and-white no-entry road sign and:
The application "gdmsetup" has quit unexpectedly. You can inform the developers of what happened to help them fix it. Or you can restart the application right now.(hotplug) "configuring hotplug" produced a dialog box too wide for the screen:
Before running hotplug support, you can preload some USB modules. Which USB modules do you want to be preloaded? [ ] aiptex ati_remote audio auerswald belkin_sa catc cdc-acm cyberjack cytherm dI assume each of those was supposed to have its own checkbox, and that it's probably just a small error in a hotplug template. I had to hit TAB to select "Ok" before proceeding.
(wishlist) Fix this bug; include the fix in revisions of stable?
(xscreensaver) The default Desktop system package selection installed the gnome desktop as default with xscreensaver (version 4.21-3) in mode "Random Screen Saver". Many of those screensavers just give yellow text saying things like:
xscreensaver: 11:50:54: 0: child pid 3952 (xanalogtv) terminated with SIGABRT.or/bin/sh: line 1: exec: gltext: not found xscreensaver: 11:50:54: 0: child pid 3719 (gltext) exited abnormally (code 127).or evenxscreensaver: 11:50:54: could not execute "moebius": No such file or directory Current directory is: / PATH is: /usr/lib/xscreensaver /usr/local/bin /usr/bin /bin /usr/bin/X11 /usr/games xscreensaver: 11:50:54: 0: child pid 3719 (moebius) exited abnormally (code 1).It's a pity - and makes a poor impression - that the default debian desktop has so many error-message screensavers! (A partial list: antspotlight, atlantis, blocktube, bouncingcow, bubble3d, circuit, cubenetic, cubestorm, endgame, engine, extrusion, flipflop, flipscreen3d, fliptext, flurry, flyingtoasters, gflux, glmatrix, gltext, hypertorus, jigglypuff, klein, lament, morph3d, noof, pinion, polytopes, providence, pulsar, sierpinski3d, spheremonics, superquadrics, stairs, starwars, stonerview, ...)
(xscreensaver: unsuitable material) After my first install, I wanted to verify that my graphics card was working under debian (it was). So I tried a few random gl screensavers. One of the first I tried was glsnake, and as luck would have it, one of the first shapes was unsuitable.
I consider this a serious bug, a time-bomb waiting to go off. I strongly agree with bugreport #313492, and was dismayed to read the maintainer's "wontfix" response.
It is *not* acceptable to install risqué material by default. It's no good saying "you can disable it" - the damage has been done. A professional desktop OS does not include this kind of thing: such things must be limited to opt-in end-users.
This tarnishes debian as a whole. How many other packages have stuff lurking? Fruity mahjongg tiles? Solitaire with special playing cards just a random number away? Lost trust is hard to regain. I've been happily handing out sarge CDs to people I don't know very well; young families. I am now more reluctant to do so.
This is bad for advocacy. In certain countries it could be used to block adoption of debian altogether, or to hassle debian-using folks with trumped-up charges. It could also hamper efforts to get debian into schools. (Daughter comes home from computer club. "Mummy, Daddy, what's a ***** ? It came up on screen and teacher didn't want me to tell anyone.") What teacher wants to explain that?
Detailed installation walk-through
Having learnt various things through trial and error, I'm ready for my fifth and final installation run! Let's see what happens ...Dr J S Bygott, Oxford Last modified: Fri Apr 21 16:15:49 BST 2006
- I boot off CD (official i386 Binary-1). Attractive debian swirl splash screen with simple message: Press F1 for help, or ENTER to boot. The help screens are well structured and well written. There does not seem to be a way to return to the splash screen. Whilst this would be of cosmetic use only, a user would expect to be able to hit, say, ESC to go back.
- Having read the various screens already, I go straight to F3 to confirm the syntax, then type "expert26" and hit return. A 2.6 kernel quickly boots.
- "[?] Debian installer main menu". Clean, simple layout. I'm not sure about the punctuation in the title. Is it deliberate, or an undisplayable character? Okay, deliberate, the next screen has "[!!]" in that position.
- Choose language: good to do this as early as possible. I choose English (also on every trial run), so can't comment further on i18n.
- Choose country: the default of "US" is wrong for me, but reasonable on population grounds. "UK" is only one arrow keystroke away.
- Select a keyboard layout: the default "PC-style" is correct
- Keymap to use: the default "British English" is correct
- Detect and mount CD-ROM - works fine; I leave module selections unchanged; Prompt for module params? No
- Start PC card services? No (I have no PCMCIA interface)
- Unable to load some modules (not available yet): via-ircc, ide-mod, ide-probe-mode, ide-detect, ide-floppy. Oh well, I did ask for expert mode. I press ENTER to continue anyway.
- Scanning CD-ROM pool works nicely; "CD-ROM detected". It's a nice touch to confirm that I have Debian GNU/Linux 3.1 r1 "Sarge" - Official i386 Binary-1 (20051220)
- Load installer components from CD Nothing on the list seems relevant. I hit TAB and ENTER to continue with default components.
- Detect network hardware (no *again* to module params; no *again* to PCMCIA; list of unavailable modules now includes "ide-scsi" as well. Continue. Correct detection of "no ethernet card". Correct detection of "No network interfaces detected".
- Enter hostname: wolfgang
Detect hardware (no *again* to module params; no *again* to PCMCIA; same list of six unavailable modules; continue)
Partition disks Correct detection of:
IDE1 master (hda) - 41.1 GB Maxtor 6K040L0 IDE1 slave (hdb) - 3.2 GB QUANTUM FIREBALL_TM3I choose "manually edit partition table". I would rather specify sizes in (whole numbers of) cylinders, but I've learnt that I have to use GB/MB. No indication that these are power-of-ten sizes except by trial and error (looking at the CHS info of the remaining free space).
strange error: 4129.090560 MB is rejected as "invalid disk size", but 4.129090560 GB is accepted, for partition hd8. I repeat this carefully several times.
- Please wait... The installer quickly creates the specified file systems.
- Install the Debian base system This stage takes only about two minutes on my system. The progress indicator spends quite some time stuck on 61%.
- Choose kernel: I choose kernel-image-2.6.8-2-386 (I really want k7, but I don't expect this flavour on binary-1). After only 56 more seconds:
- Install the GRUB boot loader I choose mbr, "grub-install (hd0)". I'm resigned to using this option. [My second or third install attempt went like this: make grub floppy just in case, get win98 to restore its mbr, boot installer, partition with /dev/hda2 marked as active, install grub to partition (hd0,1), reboot: missing operating system. Ouch. Lucky I had that grub floppy. Did I make a mistake? Or was it a problem that /dev/hda1 was >1024 cylinders? (I thought *that* issue no longer applies.) Perhaps I did something wrong; I lacked the time to experiment. On "run 4", I did get the d-i partitioner to reduce hda1 from 1250 to 1000 cylinders in case I want to try again in the future; that step seems to have worked non-destructively, the win98 disk checker is happy.]
- Finish the installation The system ejects the CD; I like that, good psychology. I press enter; it reboots; I hit DEL to enter BIOS setup and take the CD out of the boot sequence.
- Display introductory message - another nice touch
- Time in GMT? Correct default of "no".
- Timezone? Correct default "Europe/London". Good, it's making use of my earlier choice of country.
- configuring passwd: enable shadow passwords? Yes
- no problems setting up root user/normal user (with passwords) and hostname
- use a PPP connection to install? No (see earlier section for my issues with this)
- configure apt First, a nice interface to apt-cdrom. I add binary-2 and binary-3 as usual. Then I add my custom CD. Another good touch: I get prompted to supply a name for it ("Debian Sarge - custom selection March 2006") which it uses later to ask for the CD.
(package selection) I now select a whole load of packages by hand, loosely based on what "Desktop System" would do. Slow! But that's my fault!
(wishlist) I would find it more convenient if I could choose "Desktop System" to set initial package selections without installing, then launch aptitude to refine the selections, and only then install. I don't recall being offered this option.
(During selection of packages by hand, I find it slightly fiddly to get dependencies sorted since some perl pkgs from the base install are already due for a security upgrade. My own fault for having a CD with extra security downloads. Won't do that again.)
Installation is fast, and each CD is requested only once! Good! (Unlike "run 2", as mentioned above.)
Based on the country settings that I gave the installer early on, the correct default "wordlist" is "wbritish" not "wamerican". Aptitude doesn't seem to know. I did choose most of the right dictionaries by hand, but I overlooked this one, and get:
Which wordlist should be the system's default? american (American English) manual symlinks setting(wishlist) I expect this is a bit tricky to implement, but "the best OS in the world" should try. When a package depends on "xxx | yyy" there are sometimes locale reasons for preferring one. Perhaps the installer could drop some country/locale info into /etc/apt/conf.d, and aptitude could use it to make intelligent choices? Packages might need co-operating control info.
cruft should use debconf (with priority low?) to display its "preliminary notes", rather than hold us up with "Please press enter to continue...". But I gather the message has gone completely from post-sarge versions (0.9.6-0.6 onwards).
Finish configuring the base system! Good news!
Start the graphical display manager? Yes.
Oops. Stupid me, I selected gdm but not x-window-system-core, so when I exit the installer it is unable to start gdm, leaves a message to that effect on tty7 and dumps me at a root prompt on tty1. I repeat: oops! Back into aptitude...
(wishlist) Does it ever make sense to install gdm but not x-window-system-core? (For remote displays or something?) If not, the "depends" control field of gdm should be improved.
(configuring xserver-xfree86) On my other runs, of course, this step was within the overall d-i process, not an afterthought. It works well. I note with approval that with the 2.6 kernel, the graphics card correctly defaults to "nv" (on the install runs with 2.4 kernel I had to pick "nv" from a long list, not quite obvious since my card is a clone which says MS-8808 Vanta/TNT 2M64). I choose the advanced install, and supply the frequencies from my monitor's manual.
Post-install steps that might have been avoidable:
- edit /boot/grub/menu.lst to add the "map" commands for /dev/hdb
- edit /etc/fstab - although the installer knew all about my hdb, it is not listed in my default /etc/fstab. Aha, I suppose that's my own fault, for not "using" those partitions in the partitioner? It's obvious with hindsight that it writes both the partition table and /etc/fstab (that's why it asked for mount points). But perhaps it could have said this explicitly?
- add file /etc/skel/.bash_logout (I think it came from a pre-sarge examples dir and has been removed; a pity, I like this default, and it's a lot easier for users to delete than to put there)
# ~/.bash_logout: executed by bash(1) when login shell exits. # when leaving the console clear the screen to increase privacy case "`tty`" in /dev/tty[0-9]) clear esac- submit installation report - well, you did ask for it! (If reports are not welcome, than asking for them is a serious bug in the installer.)
--- End Message ---
--- Begin Message ---
- To: 364110-done@bugs.debian.org
- Subject: Closing old installation report #364110
- From: Christian Perrier <bubulle@debian.org>
- Date: Wed, 08 Sep 2010 03:58:12 +0000
- Message-id: <E1OtBnY-00059T-RN@ravel.debian.org>
We are closing this installation report for one of the following reasons: - it was reported with a pre-lenny version of Debian Installer. - indications in the installation report give the feeling that the reported problem waslying in another software, unrelated to D-I, which we can't easily identify. - indications in the installation report suggest that it may have been fixed in a more recent version of a D-I component - it was successful and we forgot closing it..:-) - it has no information we consider useful The D-I team is currently in the process of cleaning out the old spool of installation reports that haven't bene processed yet. In case you think that the problem you reported has chances to be still present, please reiterate your installation test with a more recent image of D-I, if you're in position of doing this. You'll find daily builds at http://www.debian.org/devel/debian-installer. We recommend you choose the netboot image, in the "daily builds section", then choose to install "squeeze" when prompted. If some problems are found, please report them with a new bug sent against installation-reports. Many thanks for your understanding and your help improving Debian, past and present.
--- End Message ---