Arjan van de Ven from Intel Open source centre has posted the news that http://kerneloops.org has recorded its 100,000 oops. An oops in the Linux kernel is a deviation from correct behavior of the Linux kernel which produces a certain error log. kerneloops is a client side software that helps record oops more automatically on the website with the same name and is available as part of many distribution repositories and even included by default in Fedora. This is part of the QA efforts in the Linux kernel and when posting the news, Arjan has noted that Linux kernel developers have been fixes most of the top oopses quickly
This is actually very nice. Too bad that fatal oopses can’t be reported the same way 😉
Interesting is that most of the oopses seem to come from binary-only parts (external blob drivers)
Is that unexpected when you have binary code targetting a constantly changing interface?
The problem is not the changing interface, but the lack of peer review. That’s why Free Software drivers don’t oops the kernel so much and they also have to deal with changing interfaces.
“The problem is not the changing interface, but the lack of peer review. That’s why Free Software drivers don’t oops the kernel so much and they also have to deal with changing interfaces.”
The interface does change. The peer review helps, however the people writing the majority of the drivers for Linux are kernel developers, so would be aware of the changes. That is actually why free software drivers do not oops the kernel as much as binary. Please correct me if I am wrong, as I very well could be not being involved in kernel development at all.
Yes, the lack of peer review would definitely explain why nVidia has delivered the best OpenGL driver for Linux.
Most fully featured perhaps, but it’s hardly the most stable driver, which is after all what’s being discussed here.
Nvidia driver has bugs for years, seen what happened to kde 4 when tried to deliver a better accelerated desktop. You may say that was kde bug but there are proof on the contrary. If it was a open driver it would be fixed at the month
yeah , if they kept the interface stable, at least they could paint the screen Blue and print some hex numbers !!!
[edit:sorry wrong thread]
Edited 2008-11-20 10:11 UTC
have been fixes most of the top oopses quickly
Should read “have been fixing” or just “fix” (“fixes” if you wish to imitate the British english.)
Not fixES as it’s plural (the Intel developerS)
just ‘fixing’. ‘fix’ and ‘fixes’ would both be wrong in that context.