steps in bisecting the kernel

[Doug S]

The git bisect start step gives the confused state, that I am pretty sure is a kernel.org kernel, which wouldn't have the ./debian/rules file and needs to be compiled via different methods (not working for me). However, if you just tell it "git bisect good", it will proceed to the next step, and the ./debian/rules file will be there. If that could be made to compile, then contining should be possible. Yes, it is unclear if that is the actual correct direction to proceed, but it wouldn't take long to try the end points (i.e. good, good, good ... from this point only building the last one and bad, bad, bad... from the same point, only building the last one)

Edit: As a sanity check I downloaded and compiled 3.2.0-32.51, the current 12.04 kernel. It compiled fine, although it took a few minutes longer than in the past.

[Lars Noodén]

... However, if you just tell it "git bisect good", it will proceed to the next step, and the ./debian/rules file will be there. If that could be made to compile...

That does indeed add the rules and other files, but the actual compilation has a metric buttload of y/n/m queries and then ultimately fails to compile and aborts with an error.

make[1]: *** [sub-make] Error 2
make[1]: Leaving directory `/home/foo/ubuntu-natty'
make: *** [/home/foo/ubuntu-natty/debian/stamps/stamp-build-generic] Error 2

[Lars Noodén]

Grrr.

/home/foo/ubuntu-natty/arch/x86/pci/xen.c:428:1: fatal error: closing dependency file arch/x86/pci/.xen.o.d: No space left on device
compilation terminated.

That's all the space I have on that machine. The only other partition is a smallish swap partition.

[Doug S]

That does indeed add the rules and other files, but the actual compilation has a metric buttload of y/n/m queries and then ultimately fails to compile and aborts with an error.

make[1]: *** [sub-make] Error 2
make[1]: Leaving directory `/home/foo/ubuntu-natty'
make: *** [/home/foo/ubuntu-natty/debian/stamps/stamp-build-generic] Error 2

I get the same as you. I tried both accepting defaults for the tons of configure questions and answering "no" to all of them. What I do not like with the compile method is that the error exit doesn't really tell us anything about why.