Thus spake Jesper Louis Andersen ("JLA> ") sometime Today...

JLA> I have a 1.6ZH here which crashes in some VOP_STRATEGY code. This is
JLA> completely unusable unless I somehow can get the backtrace from DDB
JLA> to the net. I know people are not sitting there and writing the stuff
JLA> down on paper, so:
JLA>
JLA> What is the usual way of getting a backtrace into a file in the file-
JLA> system?

usually /dev/scribe onto /dev/notepad or a serial interface...

JLA> One method would be a serial interface. I do not have such one handy,
JLA> but I can probably build one quite easily. Another could be to play
JLA> with savecore I expect, but I have no knowledge of that (yet, hehe).

In my experience, savecore works, but the underlying gdb support i have
found to be completely lacking -- I have yet to have a kernel crash dump
on which gdb works properly.  I shouldn't need all the debug symbols
on a kernel to gdb it, should I?

[Either the core or the kernel gets corrupted on compression.]

JLA> Thus my question: How are people out there doing this stuff? I am
JLA> willing to write appropriate sections in the kernel FAQ.



				--*greywolf;
--
22 Ways to Get Yourself Killed While Watching 'The Lord Of The Rings':

#14: Talk loudly about how you heard that there is a single frame of a nude
    Elf hidden somewhere in the movie.