Re: [lkp-robot] [x86/entry/64/compat] 8bb2610bc4: kernel_selftests.x86.test_syscall_vdso_32.fail

[Linus Torvalds]

On Wed, Jul 4, 2018 at 11:30 AM Andy Lutomirski <luto@xxxxxxxxxxxxxx> wrote:
>
> Nah, itâs a legit error.
>
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/arch/x86/entry?id=22cd978e598618e82c3c3348d2069184f6884182

Ahh. I looked at the patch and it looked "obviously correct", but only
tested my current tree (which had the above fix), not that actual
tree.

> I blame git apply for being a bit overzealous.

Heh. git apply is actually very careful and refuses to apply with
fuzz, but yes, it will apply with a big offset if the original place
has changed and it can find another place that is an exact match (why?
Think "somebody moved code around").

But yeah, apparently that _right_ point had changed, and there was an
exact match at the wrong spot. Duplicate code is bad, hmm.

And I was then confused because I was looking at an *old* dmesg
report, and saw the old kernel version from there, rather than the
right one, so I incorrectly blamed "you're running the wrong kernel".

              Linus