Bug#278746: prerc2-net partman? detect problems & 2.6 red screen
Package: installation-reports Debian-installer-version: http://cdimage.debian.org/pub/cdimage-testing/sid_d-i/i386/pre-rc2/sarge-i386-netinst.iso 20041024 uname -a: Linux Markov 2.4.18-bf2.4 #1 Mon Apr 12 11:37:50 UTC 2004 i686 unknown Date: 20041027 Method: booted off CD-RW in my Pioneer A03? DVD+-RW drive, chose expert, then expert26 Machine: OEM Processor: Pentium 4, 1.5GHz Memory: 128MiB RAM Root Device: fdisk -l /dev/hdc Disk /dev/hdc: 255 heads, 63 sectors, 1232 cylinders Units = cylinders of 16065 * 512 bytes Device BootStart EndBlocks Id System /dev/hdc1 1937469916 FAT16 /dev/hdc2 130 1232 8859847+ f Win95 Ext'd (LBA) /dev/hdc5 131 207618502+ 6 FAT16 /dev/hdc6 261 3718915766 FAT16 /dev/hdc7 391 482738958+ 6 FAT16 /dev/hdc8 521 966 3582463+ b Win95 FAT32 /dev/hdc9 967 1232 2136613+ 83 Linux This is after I tried to resize hdc1-7 under expert. Problem details to follow. Output of lspci and lspci -n: 00:00.0 Host bridge: Intel Corp. 82845 845 (Brookdale) Chipset Host Bridge (rev 03) 00:01.0 PCI bridge: Intel Corp. 82845 845 (Brookdale) Chipset AGP Bridge (rev 03) 00:1e.0 PCI bridge: Intel Corp. 82820 820 (Camino 2) Chipset PCI (rev 12) 00:1f.0 ISA bridge: Intel Corp. 82820 820 (Camino 2) Chipset ISA Bridge (ICH2) (rev 12) 00:1f.1 IDE interface: Intel Corp. 82820 820 (Camino 2) Chipset IDE U100 (rev 12) 00:1f.2 USB Controller: Intel Corp. 82820 820 (Camino 2) Chipset USB (Hub A) (rev 12) 00:1f.3 SMBus: Intel Corp. 82820 820 (Camino 2) Chipset SMBus (rev 12) 00:1f.4 USB Controller: Intel Corp. 82820 820 (Camino 2) Chipset USB (Hub B) (rev 12) 00:1f.5 Multimedia audio controller: Intel Corp. 82820 820 (Camino 2) Chipset AC'97 Audio Controller (rev 12) 02:0a.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL-8029(AS) 02:0b.0 VGA compatible controller: ATI Technologies Inc 3D Rage I/II 215GT [Mach64 GT] (rev 41) 00:00.0 Class 0600: 8086:1a30 (rev 03) 00:01.0 Class 0604: 8086:1a31 (rev 03) 00:1e.0 Class 0604: 8086:244e (rev 12) 00:1f.0 Class 0601: 8086:2440 (rev 12) 00:1f.1 Class 0101: 8086:244b (rev 12) 00:1f.2 Class 0c03: 8086:2442 (rev 12) 00:1f.3 Class 0c05: 8086:2443 (rev 12) 00:1f.4 Class 0c03: 8086:2444 (rev 12) 00:1f.5 Class 0401: 8086:2445 (rev 12) 02:0a.0 Class 0200: 10ec:8029 02:0b.0 Class 0300: 1002:4754 (rev 41) Base System Installation Checklist: Initial boot worked:[O] Configure network HW: [O] Config network: [ ] Detect CD: [O] Load installer modules: [O] Detect hard drives: [O] Partition hard drives: [E] Comments/Problems: The errata said to report when a red screen is seen for 2.6 boots. I saw one. If I can help with something there, let me know. My main reason for making this report is that the partitions didn't show up properly when I went to resize and move around the partitions. 2.4 showed all but the ext2 partition as FAT16 (which I thought was plausible at first). I went through and figured out how to resize all the partitions except for hdc8 (marked fat32 above). Trying to resize hdc8 told me that I'd found an impossible configuration or something like that. I think that it was smart enough to check the partition size and knew that FAT16 couldn't support that size? So I decided to try expert26. 2.6 showed only two partitions, and I think sfdisk said they were EZBIOS? I can't remember their type and I don't think partman and sfdisk agreed, but then I can't remember for sure. If it'll help I can check. I was also surprised that I couldn't find a move partition option. While I managed to free up disk space after several partitions, I can't merge the free space to make one larger partition. I'd really like to get a 4.7GB partition free... ;-) Yes, that's right, I'm trying to use the installer simply to repartition an extra drive that's full... all so I can back things up again. :-( Streaming backup to DVD sounds great, but then wham, there comes that 2GiB limit. Oh well, I'm on a tangent. Drew Daniels -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#192305: "q to end, b for begin" ambiguous
On Thu, 13 Nov 2003, Martin [ISO-8859-1] Sjögren wrote: > ons 2003-11-12 klockan 00.33 skrev Denis Barbier: > > * Prompt is: > > q to quit select, b for back, n for next page > > How about "b to back up"? It would be either that or "b for previous > page" IMHO. > > Or to be consistent also "n to go to the next page" or alike... fwiw, I think the important change was "q to quit select". Throughly something like: Press q to quit the selection menu Press b to go back one page in the current selection menu Press n to go to the next page in the current selection menu But all this can be left for an extended help option or a manual... unless of course it's clear that enough people are still confused. Drew Daniels -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#192305: "q to end, b for begin" ambiguous
> q to quit select, b for back, n for next page Sounds great to me. Thanks! Drew Daniels -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: d-i minimum system requirements
My paraphrased minimum requirements from various sources and some guesses is at: http://wiki.debian.net/index.cgi?DebianInstallerMinReqs - In response to ??? - Where Do you live ??? In Africa ??? Last Place in Congo ??? (Nice holliday here) Or do you install via Cell-Phone and 9600 BpS ??? ;-) I live in Winnipeg, Manitoba, Canada but sometimes I spend time at Lake of the Woods where I don't have a good land line. I'm also looking to get together a computer for a local church and the best I've found in my budget range is a 16MB 486, single speed CD-ROM drive. There's a very aggressive "Computers for Schools" program here which seems to grab all the cheap computers at the Pentium level. Also try finding cheap large RAM SIMMS for a 486. Ebay would be good if it weren't for shipping costs... I've also managed to dig up two other 486's for which I'm personally interested in maintaining... I believe in reuse before refuse. I've seen others comment on debian-devel that they still use 486's for routers and alike. There were at least some grumblings when 386 support was dropped from iirc the C++ library, and I know there were grumblings about the speed of the Kernel with 486 emulation on 386's, not to mention the few K of "bloat". - >From all of this I've basically learned that the best way to find out is to give it a try. I'm still curious though as to how hard it would be to create an installer that could have a ramdisk with just enough to create a hard drive partition and then have the installer use that? This would probably impose the burden of restarting the computer, managing another ramdisk image with it's related software (d-i installer?), and adding hard drive support into the installer... likely at least a bit of a mess. Drew Daniels -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
d-i minimum system requirements
Arg, resending... Drew Daniels -- Forwarded message -- Date: Wed, 24 Sep 2003 11:06:20 -0500 (CDT) From: Drew Scott Daniels <[EMAIL PROTECTED]> To: [EMAIL PROTECTED] Subject: d-i minimum system requirements Hi, I was wondering if there was any speculation on what the minimum system requirements for d-i are. Is 4MB, 8MB, 12MB enough ram? How close are things to any real RAM limit? The current kernel used should make d-i usable on a 386 and newer for ia32 arches due to math emulation being turned on? Does the installer require a CD-ROM drive or a network connection? I.e. does the cd install work, and can one install via floppy disks? If it is possible to install via floppy disks, what is the minimum requirements for the disks? Single sided, single density? Can anna recover from download failures? High latency, corrupted packets, low bandwidth are still issues that users deal with these days... although TCP usually does a good job of recovering from corrupted packets. If it does have the ability to recover, how nice is it? The current discover was having problems with finding ISA devices unlike an older version... Is this fixed? Is discover 2.0 out yet? Is discover still used on the installer? A good test for this might be using boches' ethernet card emulation. What other minimum requirements are there? I'll look at doing some testing on Boches again once I have time. Drew Daniels -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
d-i minimum system requirements
Hi, I was wondering if there was any speculation on what the minimum system requirements for d-i are. Is 4MB, 8MB, 12MB enough ram? How close are things to any real RAM limit? The current kernel used should make d-i usable on a 386 and newer for ia32 arches due to math emulation being turned on? Does the installer require a CD-ROM drive or a network connection? I.e. does the cd install work, and can one install via floppy disks? If it is possible to install via floppy disks, what is the minimum requirements for the disks? Single sided, single density? Can anna recover from download failures? High latency, corrupted packets, low bandwidth are still issues that users deal with these days... although TCP usually does a good job of recovering from corrupted packets. If it does have the ability to recover, how nice is it? The current discover was having problems with finding ISA devices unlike an older version... Is this fixed? Is discover 2.0 out yet? Is discover still used on the installer? A good test for this might be using boches' ethernet card emulation. What other minimum requirements are there? I'll look at doing some testing on Boches again once I have time. Drew Daniels -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
squashfs
On Wed, 5 Mar 2003, Herbert Xu wrote: > On Tue, Mar 04, 2003 at 12:24:49PM -0600, Drew Scott Daniels wrote: ... > > Where would I file a wishlist bug to get squashfs included in > > kernel-images? It's value is discussed in > > http://lists.debian.org/debian-boot/2003/debian-boot-200302/msg00412.html > > There are three or four separate threads about squashfs in the archives > > of debian-boot for Feb 2003. > > I have followed those threads and I don't see any reason why we must have > it. However, any Debian developper can package it as a kernel-patch > package. I don't see a reason why it *must* be included, but in http://lists.debian.org/debian-boot/2003/debian-boot-200302/msg00449.html Glenn McGrath <[EMAIL PROTECTED]> says: "The comparisons to cramfs does favour squashfs for lowmem installs." I can see situations where it must be used. I may be facing such a situation myself with a lowmem install on an old system of mine, but I may be able to pull things off with 4MB and no system reserved blocks. I would like to have an easy way of creating a debian install that uses squashfs. Whether this means having an official kernel-patch package and distributing an unofficial udeb or any other method. To be honest, I'd just like someone else to do it, and I'm unsure as to how much value it has. Squashfs' usefulness over cramfs seems to be in situations where cramfs images are too large for storage on media. Since the advantage is only 13% in the given example, there would not be a huge number of cases, but it is significant. 13% could be very useful to the installer team, and I would like the option somehow made available. Perhaps Koopnix may see some advantage to using squashfs as the CD filesystem. If someone demonstrates that squashfs is useful enough to create an official debian kernel-image udeb/deb then they would likely be the first maintainer and it would be made official. Useful enough should be qualified. I don't know what kind of limitations on filesystem size or memory requirements have currently been reached. Can anyone point to things that are wanted in, but not in Debian's images? Could squashfs reduce the number of disks used to install Debian? I doubt that squashfs would be any more useful than gzip for the base images. Drew Daniels -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Re: squashfs compressed file system update?
From: ftp://sources.redhat.com/pub/bzip2/docs/manual_2.html#SEC7 Compress Decompress Decompress Corpus Flag usage usage -s usage Size -1 1200k 500k 350k 914704 -2 2000k 900k 600k 877703 -3 2800k 1300k 850k 860338 -4 3600k 1700k1100k 846899 -5 4400k 2100k1350k 845160 -6 5200k 2500k1600k 838626 -7 6100k 2900k1850k 834096 -8 6800k 3300k2100k 828642 -9 7600k 3700k2350k 828642 It's interesting to note that in this case -8 and -9 yield the same results, but -9 requires more memory at the time of decompression and compression. I suspect the memory usage statistics will remain constant for any file (inherent in block sorting), not just the corpus. The compression results will of course vary depending on the input file. If memory usage is a concern and small files like initrd are being compressed, the choice between gzip and bzip2 leans towards gzip. bzip2 only really compresses better than gzip with large blocks (thus high memory requirements) and on small files like initrd it doesn't make much of a difference in nominal (non percent) terms. Prediction by Partial Match (PPM) algorithms are usually very good at compression, take lots of memory and are very slow. Thus PPM algorithms would be good to compress an entire initrd file to make the media space requirements small, but require more memory and CPU time than other compressors. The order of the files in the initrd file can make a difference if the run length or context of the compressor can span multiple files (large block sizes). I don't know how one could change the order of files in an initrd file, but I suspect it's like tar, just pass them in a different order. The order of files would make no difference if files are compressed independently like squashfs does. Phillip Lougher: I'm also not subscribed to debian-boot thus I can't get the threading right, but you did. I read the mailing list at http://lists.debian.org/debian-boot and several others on lists.debian.org. Would you be willing to try to get squashfs to be made part of the standard Debian Linux kernel? I'm not sure how, or what the reaction would be. I've submitted a request for packaging which you can see at http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=179672 It's likely that the Debian Linux kernel maintainers would not include squashfs, but would suggest that it be made a patch package. If the boot system team finds it useful enough then maybe that would be enough to have squashfs include in the Debian Linux kernel. Drew Daniels -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
PPMd (PPMII) memory, file order in archives, special tar/debs/udebs
Sorry about the cross post, but the topic here are relevent to the bug, and discussed in both lists. On Thu, 20 Feb 2003, Magnus Ekdahl wrote: > > There is a good paper on the PPMII algorithm (based on PPM) which ppmd > > uses at: http://DataCompression.info/Miscellaneous/PPMII_DCC02.pdf > > Basically, as I understand it, the more memory that you allow a PPM > > algorithm to use, the more of the file it can check for similar contexts > > or come up with better prediction values by checking more of and more in > > the file. > > True, the PPMII can use a lot of memory and if that memory ain't available > the compression model have to be crushed somehow. With decreased > compression ratio as a result. But if PPMII only needs 10Mb for its model, > another 100Mb won't make the compression any better. (As far as I understand). > It's possible. It depends the maximum context size. If you have a very large chunk in the same context (and it's allowed), then the statistics table generator should imho be checking all predictors and predicted blocks. I didn't check to see if it actually allows large contexts and actually builds tables using all the predictors and predicted blocks in larger contexts. I'm also not sure how the different models would affect maximum context sizes, but I wouldn't be surprised if it was expected that binaries have small contexts. Another possible reason for allowing a PPM algorithm to use more memory is to allow it to be sloppy and quick by using full ints or equivalent in its in memory table generation. There are memory usage optimizations that can be put in place as it is known ahead of time the maximum number of possible occurrences of a block. If you take the amount of space left in the file and divide it by the block size you get the maximum possible number of occurrences (a rough description). Having the maximum number of occurrences be small enough, there is extra space in the int or whatever kind of counter that could be shared with another table entry (counter/statistic). I doubt ppmd would be using such an optimization though and I'm not sure if it's counters are slightly lossy (ie, floats) although lossy counters can be optimized too. > > Another thing to note when creating compressed *archives* is that the > > order of files may be important. It has been documented (I don't remember > > where) that grouping files of similar types together can yield better > > compression. It might yield optimum compression to go through all the > > different ways of ordering files. I suspect this is how the 7-zip ( > > http://www.7-zip.org ) algorithm outperforms other zip algorithms and > > still maintains compatibility with all other zip archivers. > > > > A program or script to put files into different orders for archiving (in a > > list or in a list passed to something like tar) is something that I have > > been meaning to do. Going through all possible combinations would be nice, > > but it may be faster to use something like file to group like files > > together. File has quite a few bugs against it though ( > > http://bugs.debian.org/file ) and it may be far from an optimal choice. > > Certainly using file would be a good place for me to start. It'd be nice if > > someone wrote such a script before me though. ;-) > > Hmm... this might be more of a tar issue than a PPMd issue. Tar has > additional problems too, since it adds extra things in the archive. > Christian Fasshauer <[EMAIL PROTECTED]> has done some work with tars > problems, and his solution was supposed to decrease the size of the debian > packages. Perhaps he has more suggestions of what can be done. > I like the extra information that tar includes, more so in source files, and especially in "original" source files. When looking through source code, seeing the date that a file was change/modified/created and it's attributes can tell allot about the history of the file. For binary packages, the value of the information is debatable. I'm on the fence as far as whether removing things like the number of user/group entries available and other things suggested in http://lists.debian.org/debian-dpkg/2003/debian-dpkg-200301/msg00049.html but I'm leaning towards their inclusion in standard debs, and their removal in udebs. Glenn, did you ever get around to making a busybox dar applet? If so what were your results. Does it implement the suggestions of Christian Fasshauer? If so maybe Russell Coker could get his comparisons done. http://bugs.debian.org/cgi-bin/bugreport.cgi?archive=no&bug=174308 is also an interesting discussion about gtar, star... I really don't know enough about any tar to have an opinion as to what kind of tar is best. To save even more space, a special compressor with a dictionary of Debian blocks and/or statistics could be used. Thus if few users/groups etc were used then such a dictionary would pick up on this and be used. As far as compression goes, if there's a common dictionary to all the files, then the
PPMd (PPMII), bzip2 -9 misconception, order of files in archives
On 19 Feb 2003 17:16:48 +1100, Glenn McGrath <[EMAIL PROTECTED]> wrote: >I have looked at PPMd, ive read thats good but it took me a while to >workout how to use it to get maximum ocmpression, here is a comparison. > >8921088 Packages >2375680 Packages.gz >1814528 Packages.bz2 >1335296 Packages.pmd > >That was done with "PPMd e -o16 -m256 -r2 Packages", it surprised me >that the -m option (amount of memory to use) effects the compression >ratio. > >Thats a pretty improvement over bz2. I would guess that memory usage improving compression is a concept that is widely unknown so I filed a bug against ppmd: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=181665 I'm ccing the bug as in this message I talk about the -m option in this next paragraph. There is a good paper on the PPMII algorithm (based on PPM) which ppmd uses at: http://DataCompression.info/Miscellaneous/PPMII_DCC02.pdf Basically, as I understand it, the more memory that you allow a PPM algorithm to use, the more of the file it can check for similar contexts or come up with better prediction values by checking more of and more in the file. Some interesting stuff about bzip2 form: ftp://sources.redhat.com/pub/bzip2/docs/manual_1.html#SEC1 - bzip2 compresses files using the Burrows-Wheeler block-sorting text compression algorithm, and Huffman coding. Compression is generally considerably better than that achieved by more conventional LZ77/LZ78-based compressors, and approaches the performance of the PPM family of statistical compressors. - Also, when using bzip2 note that -9 does not indicate to use the best compression (best as in size), but instead specifies the largest blocksize which *usually* gives the smallest compressed file. Sometimes different block sizes can actually be better than -9. I'd have to test it to be sure, but the case I was testing was compressing several megabytes of IRC logs which are ASCII text with many repeating words and a defined structure. The reference fwiw: ftp://sources.redhat.com/pub/bzip2/docs/manual_2.html#SEC6 - -1 (or --fast) to -9 (or --best) Set the block size to 100 k, 200 k .. 900 k when compressing. - Perhaps I should make a wishlist bug against bzip2 about this option's documentation. --best could also be said to be short for best guess at blocksize which uses the 900 k block size. Also people using bzip2 might need to be explicitly informed. Perhaps all bzip2'd files in Debian should be checked to see what blocksize is best and file bugs against the few that don't have optimal blocksizes chosen. I might warn of a mass bug filing in debian-devel if I decide I want to test all relevant bz2 files. It should also be noted that bzip2 may not be an optimal implementation of the Burrows-Wheeler block-sorting algorithm. In ftp://sources.redhat.com/pub/bzip2/docs/manual_4.html#SEC44 Julian Seward says: "It would be fair to say that the bzip2 format was frozen before I properly and fully understood the performance consequences of doing so." The Archive Comparison Test (ACT) (usually available at http://compression.ca ) has a comparison of archivers. You can find the algorithm that each archiver uses in the Index of archivers. I found the best compressors for size use Prediction by Partial Match (PPM which is based on Markov Chains) and Context Tree Weighting (CTW) algorithms. Another thing to note when creating compressed *archives* is that the order of files may be important. It has been documented (I don't remember where) that grouping files of similar types together can yield better compression. It might yield optimum compression to go through all the different ways of ordering files. I suspect this is how the 7-zip ( http://www.7-zip.org ) algorithm outperforms other zip algorithms and still maintains compatibility with all other zip archivers. A program or script to put files into different orders for archiving (in a list or in a list passed to something like tar) is something that I have been meaning to do. Going through all possible combinations would be nice, but it may be faster to use something like file to group like files together. File has quite a few bugs against it though ( http://bugs.debian.org/file ) and it may be far from an optimal choice. Certainly using file would be a good place for me to start. It'd be nice if someone wrote such a script before me though. ;-) Drew Daniels -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
squashfs compressed file system update?
On Wed, 5 Feb 2003 14:08:14 +1100 Glenn McGrath wrote: > On Tue, 4 Feb 2003 10:27:59 -0600 (CST) Drew Scott Daniels > <[EMAIL PROTECTED]> wrote: > >> It looks stable enough now, but I wonder why it hasn't been included >> in the Linux kernel (even 2.5?). I also have to question the value of >> zlib compression vs other types of compression, but then I suppose >> that zlib usually gives better compression than none. If I ever get to >> it I'll try writing a zlib replacement that uses a PPM variant, but it >> seems unlikely that I will. > >The reason gzip compression is used on intrd's is becasue the kernel and >some bootloaders (grub at least) support it. > Would squashfs work? I doubt this is an optimal compression, but it might be better than alternatives. Currently the best free solution might be based on PPMd which has been packaged into Debian. >It would be great is we could use better compression on the initrd. > What support would be needed for this? >Ill have a look at squash fs for comparison, but if it isnt in the main >kernel its hard to depend on it for the installer, it could be used in >custom builds though i guess. Have you done a comparison? Would it be possible to include the option to use squashfs if it's deemed better than at least some of the other options? Thanks Drew Daniels -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
squashfs compressed file system
I've been meaning to file an RFP on squashfs for a while, and I finally got around to it ( http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=179672 ). While I was at it, I figured I might ask if this filesystem has been looked at for use in any debian media. I know there has recently been some comparisons of cramfs, ext2, romfs... It looks stable enough now, but I wonder why it hasn't been included in the Linux kernel (even 2.5?). I also have to question the value of zlib compression vs other types of compression, but then I suppose that zlib usually gives better compression than none. If I ever get to it I'll try writing a zlib replacement that uses a PPM variant, but it seems unlikely that I will. Squashfs is hosted at http://squashfs.sourceforge.net Drew Daniels -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]