I apologize if this is inappropriate, but I am sure someone will delete it if it is.
I could not get an address for Chris Wright and I wanted to reply to a comment that was reported as his made at Linux Foundation Collaboration Summit in Austin, Texas concerning bug reporting and the kernel.
I have used 2.6.8 through 2.6.14 and probably beyond as I ran the 6.1 release of Kubuntu (I think that is the correct one). I have used mostly Debian distributions and I had problems that to me are kernel related but I do not have the equipment (and probably not the time) to figure it out. Linux always ran good but when I started using the BOINC applications to do mostly protein research (Rosetta project), terrible things happened. I have enough experience testing and developing custom OS code (CPM and several custom Motorola processor based - it dates me doesn't it!) to feel that it is really a kernel problem. I have run NT since version 4, and have had many things happen, but I have never had drives get thrashed as I did in Linux. I posted some comments on forums but the usual response is it is a hardware problem (my hardware works with Windows NT/2000 but not Linux - actually 2 different hardware platforms). I have got more than a little grief on some sites concerning my comments about Linux being flaky. I support Linux and have gotten my wife to move from Windows 2000 to Ubuntu. After having several occasions where after a power failure one or more of my drives were trashed, I moved to Solaris. I am not sure it is any better but I have not found a BOINC version that runs on it and it has been stable. With BOINC running on Windows 2000 I had power failures and a power supply die, but it always came back OK.
Thanks,
Chris Wright can be found in
Chris Wright can be found in LKML as well as many other developers, that can handle difficult problems. He also manager of "stable" branch:
http://lwn.net/Articles/278804/
When posting to LKML be ready to be ignored (i.e. "we handle only vanilla kernel.org stuff") or be asked for more details, hard work, including (in)famous bisecting: (until kernel_is_OK; do patching && compiling kernel && run kernel; done).
But some applications (better Open Source) may reveal really subtle kernel bugs, like it was with apt-get/rtorrent [ext3 corruption] -> [corrupted memory handling bug] in 2.6.19/20 cycle (looong thread, easy to find).
Good luck.
_____
Reporting bugs and bisection
Andrew Morton (Linus' right hand):
[...]
We dont' do that as much nowadays - there's a tendency to
a) throw the problem back at the reporter, often asking them to bisect.
If the reporter is running a distro kernel (eg: Fedora) then that's
quite hard, and often isn't a think they have knowledge to do. So
they'll just disappear. Or
b) just ignore the report altogether.
http://article.gmane.org/gmane.linux.network/90836