Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
FS#10562 - CONFIG_PREEMPT option changed? Problems with BOINCs FPU-stack
Attached to Project:
Arch Linux
Opened by Holger Jorra (Trac3R) - Monday, 02 June 2008, 12:48 GMT
Last edited by Aaron Griffin (phrakture) - Friday, 18 July 2008, 16:18 GMT
Opened by Holger Jorra (Trac3R) - Monday, 02 June 2008, 12:48 GMT
Last edited by Aaron Griffin (phrakture) - Friday, 18 July 2008, 16:18 GMT
|
DetailsDescription:
Using the package community/boinc I have problems to crunch numbers for the Einstein@home project for some time now, while other ones work. The reply to my request there may point on a kernel bug explained at LKML: http://www.nabble.com/CONFIG_PREEMPT-causes-corruption-of-application's-FPU-stack-td17293854.html So I wonder if the CONFIG_PREEMPT option has been changed into the kernel package recently (last 3-5 months). Additional info: * boinc 5.10.45-2 * kernel26 2.6.25.4-1 and predecessors Steps to reproduce: * run boinc as deamon * join project einstein@home |
This task depends upon
Closed by Aaron Griffin (phrakture)
Friday, 18 July 2008, 16:18 GMT
Reason for closing: Upstream
Additional comments about closing: Fixed in 2.6.26
Friday, 18 July 2008, 16:18 GMT
Reason for closing: Upstream
Additional comments about closing: Fixed in 2.6.26
I post the link to the Einstein forums and the "fix" at LKML here.
http://einstein.phys.uwm.edu//forum_thread.php?id=6700
http://lkml.org/lkml/2008/6/1/34
The error occured definately because of a kernel bug, I tested this. Because of the fact that you prefer vanilla kernels I suggest to use the
CONFIG_PREEMPT_VOLUNTARY option instead of CONFIG_PREEMPT for kernel builds as long as the fix isn't official.