FS#10614 - 2008.04RC CD install : gcc-libs corrupted package

Attached to Project: Arch Linux
Opened by Damien Herraud (bapman) - Sunday, 08 June 2008, 23:04 GMT
Last edited by Simo Leone (neotuli) - Sunday, 22 June 2008, 00:04 GMT
Task Type Bug Report
Category Packages: Extra
Status Closed
Assigned To Simo Leone (neotuli)
Aaron Griffin (phrakture)
Architecture All
Severity High
Priority Normal
Reported Version 1.5.1
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 3
Private No

Details

Hi,

As reported on the forum, some people (and myself) had a problem with CD install from 2008.04RC images. We all tried to install it from usb stick. We all checked the md5sum, which was right.

Here is what we get at the "install packages" stage of installation :

Resolving dependencies....

Warning: Dependency cycle detected
Warning: bash will be installed before its glibc dependency
Warning: db will be installed before its coreutils dependency

Error:: failed to commit (corrupted package)

Error: md5:/var/cache/pacman/gcc-libs-4.3.0-1-i686
Error: md5:/var/cache/pacman/gcc-4.3.0-1-i686
Can't get md5 checksum for gcc-libs-4.3.0-1 ... can't get checksum for gcc-4.3.0-1


FTP install worked flawlessly though.

Here is the forum post : http://bbs.archlinux.org/viewtopic.php?pid=376897#p376897
This task depends upon

This task blocks these from closing
 FS#10702 - 2008 RC ISO related bugs 
Closed by  Simo Leone (neotuli)
Sunday, 22 June 2008, 00:04 GMT
Reason for closing:  Fixed
Additional comments about closing:  Partitioning a file is a bitch.

Thanks for the help guys!
Comment by Damien Herraud (bapman) - Sunday, 08 June 2008, 23:08 GMT
Ooooops I put it in the wrong project. Sorry about that !
Comment by don carpenter (pooldoo) - Friday, 13 June 2008, 01:53 GMT
I'm having exactly the same problem. I'll be happy to test when there's a fix.
Comment by Glenn Matthys (RedShift) - Wednesday, 18 June 2008, 08:16 GMT
I cannot confirm this, did you check if your ISO md5sum was correct after downloading? Can you try another CD-R, another brand of CD-R or a different CD-ROM/CD-RW drive?
Comment by don carpenter (pooldoo) - Wednesday, 18 June 2008, 12:49 GMT
Yes, I did check md5sum. The others did also, according to the posts. I believe everyone having the problem were installing from a usb flash drive using a .img image. I haven't tried a different usb stick - only have the one at the moment. I'm wondering if anyone has had a successful install from a .img on a usb stick?
Comment by Damien Herraud (bapman) - Wednesday, 18 June 2008, 13:53 GMT
I second that, it was on the usb key with .img image. I did check md5sum too.
Comment by Glenn Matthys (RedShift) - Thursday, 19 June 2008, 09:02 GMT
Confirmed with arch-core-install-2008.04-rc-i686.img:

[root@archlive pkg]# md5sum gcc-libs-4.3.0-1-i686.pkg.tar.gz
md5sum: gcc-libs-4.3.0-1-i686.pkg.tar.gz: Input/output error

The kernel has to say this about it:
attempt to access beyond end of device
sda1: rw=0, want=731462, limit=722862
attempt to access beyond end of device
sda1: rw=0, want=731470, limit=722862
attempt to access beyond end of device
sda1: rw=0, want=731478, limit=722862
attempt to access beyond end of device
sda1: rw=0, want=731480, limit=722862
attempt to access beyond end of device
sda1: rw=0, want=731480, limit=722862
attempt to access beyond end of device
sda1: rw=0, want=731480, limit=722862
attempt to access beyond end of device
sda1: rw=0, want=731480, limit=722862
attempt to access beyond end of device
sda1: rw=0, want=731480, limit=722862
attempt to access beyond end of device
sda1: rw=0, want=731462, limit=722862
Comment by Aaron Griffin (phrakture) - Thursday, 19 June 2008, 18:45 GMT
I will be building a new ISO soon. Thanks for reporting this
Comment by Glenn Matthys (RedShift) - Thursday, 19 June 2008, 22:26 GMT
Aaron: release your ISO whenever you're ready. I'll then go through the bugtracker again and request a close of the bugs. Is that ok for you?
Comment by Greg (dolby) - Thursday, 19 June 2008, 22:34 GMT
I have all ISO related bugs listed. If a new ISO comes out we can review them.
Comment by Aaron Griffin (phrakture) - Thursday, 19 June 2008, 23:11 GMT
FTR: We can create a "master bug" for the new ISO, and set all of these as dependencies for closing that one - it's real useful in flyspray. Do a search of closed bugs for "bugcatcher" 8)
Comment by Greg (dolby) - Thursday, 19 June 2008, 23:27 GMT
Done  FS#10702 
Comment by Simo Leone (neotuli) - Friday, 20 June 2008, 07:42 GMT
 FS#10225  is a dupe of this. I suspect this has to do with the already funky behavior with the USB filesystem. It needs to be made much larger than filesystem metadata and file data in order to even boot, and I have no idea why.
Comment by don carpenter (pooldoo) - Friday, 20 June 2008, 12:29 GMT
I noticed the comment in  FS#10225  by febeee "I reproduced the problem systematically on a Eee PC." This error occurred for me while installing on an Eee PC also, from a usb stick onto an external 4GB SanDisk SDHC card. (Not sure if it's even possible to do this.) Curious if others were using unconventional hardware for install.
Comment by Simo Leone (neotuli) - Friday, 20 June 2008, 12:38 GMT
I'm fairly convinced this has to do with something in the usb image generation, but I'm not sure what's causing it. If a release is made while we still don't know what the problem is, I'll make sure to
cat /src/core/pkg/* > /dev/null
to check for i/o errors as well as check all the md5sums.
Comment by Simo Leone (neotuli) - Saturday, 21 June 2008, 09:18 GMT
I simplified mkusbimg a bit, using some ideas from http://www.stickytapesolutions.com/imageForStick.html to avoid strange loopback games and things like that, but the random corruption remains an issue.
Comment by Glenn Matthys (RedShift) - Saturday, 21 June 2008, 09:51 GMT
Where can I find this mkusbimg?
Comment by Simo Leone (neotuli) - Saturday, 21 June 2008, 18:42 GMT
http://projects.archlinux.org/?p=archiso.git
Comment by Simo Leone (neotuli) - Saturday, 21 June 2008, 21:51 GMT
I think I got it. Turns out both all the *fdisk programs AND the kernel itself are absolutely horrible at calculating disk geometries for things that aren't really disks, like our usb image file. Consequently, we need to go against all these programs advice and forcefully write a correct partition table, even if sfdisk claims it goes past the end of the disk. So far this has been working, and it also solved the "won't boot if my ext2 partition is too small" problem. I'm gonna run it like 10 more times to make sure it isn't random luck, but I think I might have nailed it this time :D

Loading...