more arches for clang

[nickdesaulniers]

We should be able to build with clang:

• malta_kvm_guest_defconfig for mips and mipsel
• ppc44x_defconfig for powerpc (32b)
• pseries_defconfig and powernv_defconfig for ppc64 (BE and LE respectively)
• build-configs.yaml: add riscv to the clang configs #534
• defconfig for s390
• build-configs.yaml: enable i386 for clang-10 and clang-11 of next #577

[gctucker]

Is this all to be built with clang 11 on linux-next?

(see my other question on #518 about mainline and the extent of the build coverage on each branch)

[broonie]

These all seem like reasonable targets to me, especially if we don't do any allmodconfigs which would consume lots of build resource.

Might also be worth throwing allnoconfigs in there for the ones that it works on, they're tiny and do throw up issues from time to time.

[gctucker]

Might also be worth throwing allnoconfigs in there for the ones that it works on, they're tiny and do throw up issues from time to time.

Yep it's there for riscv:
https://github.com/kernelci/kernelci-core/pull/534/files#diff-ec5c0ba0497e23f334c288a5ca3580e72c83e4bd1eac5a6627c33c22475cc2d2R300

[gctucker]

@nickdesaulniers I see you're now using tuxbuild with your own CI system, so are you still looking at kernelci.org results?

[nickdesaulniers]

Absolutely; I have to file a bunch of bugs in our issue tracker from the latest email report. It's on my TODO list.

My goal with the various kernel CI systems is redundancy. Each service has its tradeoffs, and https://blog.travis-ci.com/oss-announcement blew up our initial CI so it's not given that any one service sticks around forever.

[gctucker]

OK great, thanks for confirming. Please continue to provide feedback so we can keep improving things. The KernelCI build tool is being redesigned now to become more flexible, there may be ways for Clang builds and your workflow to benefit from that.

[padovan]

legacy material. Today we build everything with both compilers