Re: [OE-core] [PATCH 04/10] cmake: update 3.20.2 -> 3.20.3

["Richard Purdie" <richard.purdie@linuxfoundation.org>]

On Tue, 2021-06-08 at 20:15 +0200, Alexander Kanavin wrote:
> I can confirm that the previous kernel behaves correctly:
> utimensat_time64(AT_FDCWD, "../install/usr/local/lib/cmake/assimp-4.1/assimp-config.cmake",
> [{tv_sec=1623176017, tv_nsec=6290007298941124608}, {tv_sec=1622966579, tv_nsec=17838646317425885184}], 0) = -1
> ENOSYS (Function not implemented)
> utimensat(AT_FDCWD, "../install/usr/local/lib/cmake/assimp-4.1/assimp-config.cmake", [{tv_sec=1623176017,
> tv_nsec=0} /* 2021-06-08T18:13:37+0000 */, {tv_sec=1622966579, tv_nsec=0} /* 2021-06-06T08:02:59+0000 */], 0)
> = 0
> 
> Does look like a botched backport of time64 syscalls to me.
> 
> Alex
> 
> On Tue, 8 Jun 2021 at 17:40, Michael Halstead <mhalstead@linuxfoundation.org> wrote:
> > I've rebooted centos8-ty-1 using the previous 4.18.0-240.15.1.el8_3.x86_64 kernel and kept it in the pool.
> > I've paused centos8-ty-2 so it won't interfere with builds and left it at the current 4.18.0-
> > 305.3.1.el8.x86_64 kernel for testing. 

Thanks Michael and Alex. We can therefore pretty safely say that something was 
broken between the 240 and 305 of the centos8 kernel for the 32 bit syscall
utimensat syscalls.

I did poke around https://github.com/kernelim/linux/tree/centos8 which does
have a kernel diff but it is huge and you have to clone the repo to get it.

git show a3342908613ba72a84f652ca7a56c3e2113bda12 | grep sys_utimensat -C 40

shows they did add the syscalls in that kernel.

So this does look to be a RedHat issue. Not sure if we want to report it
to them? Can we run the autobuilders on the older kernel for now until
they hopefully fix it?

Cheers,

Richard