For got to cc this to the bug...-( On 21 July 2014 09:29, Ben Hutchings <b...@decadent.org.uk> wrote: ... > > Assuming that this regression only happened once, we can be fairly > confident that it happened before > 4c823cc3d568277aa6340d8df6981e34f4c4dee5, but as that is around v3.2-rc1 > it doesn't reduce the range very much. :-( >
Thank for looking at this issue. I don't get what is going on here either but it seemed to reliably cause the problem BUT when I reverted this change in the v3.2 kernel it still had the fault! (Broke my heart!). Then I built the version 8a9c59 (just prior to the bad version) from scratch and this time it had the fault where as before worked! So I think I have to build completely cleanly each time... So I am doing a git bisect again using a cleaner way of rebuilding the kernels. I did compile one on either side of it and it seemed solid. i.e. rebooting the same kernel reliably results in a failure. I didn't reboot every kernel but certainly the squeeze one is always good and the wheezy is almost always bad (sometimes you can partially write a BluRay before it completely bombs but it always gives these errors on kernel booting). One idea was that it might be a kernel configuration dependant because I had to do make oldconfig and as I switched back and forth between revisions it might have different configurations. But I just diff'ed the config file from the last 4 builds and there was *NO* difference so that should cover good and bad builds. As I didn't rebuild the last set of kernels from scratch I suspect this is the cause of the inconstancies. I am now trying the bisection again but now doing this each time. 1. Rebuilding from scratch (make-kpkg clean) 2. Copying the same config file and doing make oldconfig and always accepting the default. So far I have: git bisect start # bad: [8a9c594422ecad912d6470888acdee9a1236ad68] drivers/block/loop.c: emit uevent on auto release git bisect bad 8a9c594422ecad912d6470888acdee9a1236ad68 # good: [02f8c6aee8df3cdc935e9bdd4f2d020306035dbe] Linux 3.0 git bisect good 02f8c6aee8df3cdc935e9bdd4f2d020306035dbe # good: [0003230e8200699860f0b10af524dc47bf8aecad] Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6 git bisect good 0003230e8200699860f0b10af524dc47bf8aecad # good: [206d440f64030b6425841bf7cb38e26a5ea0c382] xfs: Fix build breakage in xfs_iops.c when CONFIG_FS_POSIX_ACL is not set git bisect good 206d440f64030b6425841bf7cb38e26a5ea0c382 # skip: [c11abbbaa3252875c5740a6880b9a1a6f1e2a870] Merge branch 'slub/lockless' of git://git.kernel.org/pub/scm/linux/kernel/git/penberg/slab-2.6 git bisect skip c11abbbaa3252875c5740a6880b9a1a6f1e2a870 The last skip because I got compile errors. ... arch/x86/built-in.o: In function `atomic_read': /movies3/work/linux/arch/x86/include/asm/atomic.h:25: undefined reference to `__tracepoint_xen_cpu_write_gdt_entry' ... Actually I couldn't build the squeeze kernel under wheezy (lots of problems) so I was building that under the squeeze in a virtual machine. This is a real annoying bug because I have to build so many kernels and hit compile problems but I will crush it. Suggestions on improving this debug plan would be much welcomed. Andrew -- To UNSUBSCRIBE, email to debian-kernel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/CA+Y=x3nb3qzsjykqnai0jwyo4f+__z1jab5e8ut+cxwnh0g...@mail.gmail.com
