FS#10209 - 'glibc 2.7-9' ruined whole system

Attached to Project: Arch Linux
Opened by sda (sda00) - Thursday, 17 April 2008, 16:16 GMT
Last edited by Greg (dolby) - Saturday, 24 May 2008, 08:25 GMT
Task Type Bug Report
Category Packages: Core
Status Closed
Assigned To Jan de Groot (JGC)
Andreas Radke (AndyRTR)
Architecture i686
Severity Critical
Priority Normal
Reported Version 2007.08-2
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description:

After upgrade to 'glibc 2.7-9' I lost ability to boot system natively (solved by booting from OpenSuSE grub), all new downloaded files had incorrect inodes reporting that their size is more than 20Gb each (real file size was from 800Mb to 1.6Gb, disk is a healthy reiserfs-3.6 with standard journal, and it's clear that all downloaded files were corrupted and unusable without manual adjustments of their 'tales'). System became extremely unstable. Tried to rebuild all system - hopeless.


Result: reinstalled system and downgraded glibc, klibc, binutils, coreutils, gcc, mkinitcpio.


http://bbs.archlinux.org/viewtopic.php?pid=356821#p356821

Additional info:
* package version(s)
* config and/or log files etc.


Steps to reproduce:

upgrade to glibc-2.7_9

System Info: AMD 3700+, 1.5Gb RAM, ext3 as / and reiserfs-3.6 as /home (all disks are healthy)
This task depends upon

Closed by  Greg (dolby)
Saturday, 24 May 2008, 08:25 GMT
Reason for closing:  Works for me
Comment by Dan McGee (toofishes) - Thursday, 17 April 2008, 19:05 GMT
uname -a output, pacman.log would be helpful if you could still get to it (if not, any other packages that may have been upgraded at the same time).

And since you have reinstalled your system and can't get to the above, do you think this bug is reproduceable? None of the devs have seen this or we obviously wouldn't have released it to core.
Comment by Thomas Bächler (brain0) - Thursday, 17 April 2008, 19:10 GMT
Your kernel is too old. Every kernel older than 2.6.24.4 will lead to breakage in binaries built with gcc 4.3. You need at least 2.6.24.4 or any older kernel that has been properly patches.
Comment by sda (sda00) - Thursday, 17 April 2008, 19:11 GMT
> uname -a
Linux black 2.6.24-ARCH #1 PREEMPT Wed Apr 2 12:59:54 NOVST 2008 i686 AMD Athlon(tm) 64 Processor 3700+ AuthenticAMD GNU/Linux


my kernel is/was

> pacman -Q | grep kernel
kernel-headers 2.6.24.3-1
kernel26 2.6.24.4-1
Comment by sda (sda00) - Thursday, 17 April 2008, 19:14 GMT
frankly speaking I'll upgrade once more, but if this remain unusable as it was - I just switch to OpenSuSE and won't even bother to inform you about it. I'm not going to to be a 'tester' right now. Forum post confirms that I'm not along with it.

My Best Regards.
Comment by Thomas Bächler (brain0) - Thursday, 17 April 2008, 19:36 GMT
Nobody here is interested in what you will switch to and how much of an ass you're gonna be about it. Stick to facts, please.
Comment by Jan de Groot (JGC) - Friday, 18 April 2008, 07:55 GMT
What do you mean with "can't boot natively"? Grub and glibc don't have anything to do with eachother, as grub isn't even linked dynamic to glibc. And what do you mean with "tales"?
Comment by sda (sda00) - Friday, 18 April 2008, 12:42 GMT
>> What do you mean with "can't boot natively"?

I was quite surprised to discover it also. Just finished full system rebuild (3-rd). Dunno why, but if not to reinstall grub to root disk bootsector - system refuses to start. Grub just stops and there's no even a prompt. For example:

Case 1
1) rebuild and reinstall all
2) reboot failed

Case 2
1) rebuild and reinstall all
2) boot from another 'grub'
3) reinstall grub to root disk bootsector
4) success

In MBR I have an OpenBSD loader which loads Arch grub (hd0,0) or OpenSuSE one (hd0,2). I get used to it and it works.

>> And what do you mean with "tales"?

As I said the first two attempts to upgrade to glibc 2.7-9 resulted in inconsistency between actual file size and inode markers. Command:

> tail -n 100 file_just_copied

returned 'nothing' (I admit that phrase "unusable without manual adjustments of their 'tails'" is more adequate, but sense remains the same).

I'm not used Arch after third rebuild much. Just reinstalled grub. Quick 'file copy' test passed ok, but I can't stand than now system is a stable one and that all my software runs as it should. Guess it's clear that after all I'll stick to Arch only when I have time for 'testing'.


Thanks.
Comment by sda (sda00) - Friday, 25 April 2008, 10:44 GMT
This is my last report here. Despite that gcc-4.3 is in terms of some polite people "more strict", Arch detected that my /home partition should be forced to check (more than 32 mount operations had passed). To keep consistency before my third upgrade I formatted my /home to ext3 (as my /). And I see that in front of my eyes it replaces good file inodes with something... Look:


> ls -lh

total 689M
-rw-r--r-- 1 sda users 20G 2008-04-22 23:44 Captain (1-7 serial).avi
-rw-r--r-- 1 sda users 20G 2008-04-22 23:45 Captain(8-13 serial).avi

> ls -lsh

total 689M
367M -rw-r--r-- 1 sda users 20G 2008-04-22 23:44 Captain (1-7 serial).avi
322M -rw-r--r-- 1 sda users 20G 2008-04-22 23:45 Captain(8-13 serial).avi


> ls -sh
total 689M
367M Captain (1-7 serial).avi
322M Captain(8-13 serial).avi


Size of all changed file inodes is 20990681088 bytes.

Guys, I'm Dumb Of Astinishment!

Loading...