Re: [GIT PULL] Networking for 5.17

From: Nathan Chancellor
Date: Tue Jan 11 2022 - 11:01:43 EST


On Mon, Jan 10, 2022 at 07:31:30PM -0800, Linus Torvalds wrote:
> I really wish we had more automation doing clang builds. Yes, some
> parts of the kernel are still broken with clang, but a lot isn't, and
> this isn't the first time my clang build setup has found issues.

As far as I know, we have four major groups doing regular build testing
with clang:

* Intel's kernel test robot
* KernelCI
* RedHat's Continuous Kernel Integration (CKI)
* Linaro's Linux Kernel Functional Testing (LKFT)

I regularly check the daily -next report that we get from KernelCI to
see what breakage there is and triage it as needed. The rest email us as
things break. The Intel folks are the only ones building from the
mailing list as far as I can tell, everyone else mainly targets your
tree and/or -next.

I don't think this particular issue was an automation fail, more of a
timing one, as the warning was reported by the kernel test robot:

https://lore.kernel.org/r/202201101850.vQyjtIwg-lkp@xxxxxxxxx/

However, it was reported a little under a day after the patch hit the
mailing list according to the lore timestamps at the bottom, after it
had already been merged into net-next (it looks like they were applied
to the netfilter tree and merged into net-next within an hour or so).

Pablo did sent a follow up fix rather quickly, which I noticed because
my own local builds were broken.

https://lore.kernel.org/r/20220110221419.60994-1-pablo@xxxxxxxxxxxxx/

Normally, I try to review patches like this so that the maintainers are
aware that the warning will break a build with CONFIG_WERROR. In this
case, I assumed that the netdev build tests would catch it and it would
be applied before the pull request was sent, as they have started
testing with clang and catching these warnings before accepting patches
but as Jakub said, that did not happen.

I'll try to keep an eye out for this stuff in the future, so that it is
dealt with by the time you get it, especially now that passing -Werror
is expected. Most standard arm64 and x86_64 configs should be completely
warning free with clang now, arm and some of the more exotic
architectures are still a WIP.

Cheers,
Nathan