workflows: Zephyr workflow uses new 0.16.1 Zephyr SDK

[aborisovich]

Updated Zephyr workflow - installed new Zephyr SDK v0.16.1 that is required by Zephyr main branch.
New Zephyr SDK is NOT backward compatible to older Zephyr revisions so the upgrade in west.yml is done to the commit: zephyrproject-rtos/zephyr@9fc99928caf0 that requires this toolchain.
Build-windows job changes:

• new Zephyr SDK had changed extension and now is compressed using 7z format.
• from build-windows workflow removed unzip package that is no longer required by the setup.cmd Zephyr SDK installation script

Build-linux job changes:

• updated script zephyr\docker-run.sh that clones Zephyr Docker container to clone latest tag versioned v0.26.4.
• new container contains Zephyr SDK v0.16.1
• upgraded runner OS to Ubuntu 22.04 to match OS from Docker container

Sparse-Zephyr workflow:

• upgraded runner OS to Ubuntu 22.04 to match OS from the Zephyr workflow.

Signed-off-by: Andrey Borisovich andrey.borisovich@intel.com

[aborisovich]

@marc-hb we got all green on the Zephyr workflows.
Internal Intel CI had not finished yet but I think no point waiting for those.
This was with west.yml set to Zephyr main branch (what checks build-windows with zmain revision).
Dropping DNM and zmain revision from west.yml, let's see whether current SOF manifest rev works ok with new SDK.

[marc-hb]

Looks great, thank you! Very glad it seems to have gone smoothly.

Just to be on the safe side please add this:

--- a/.github/workflows/sparse-zephyr.yml
+++ b/.github/workflows/sparse-zephyr.yml
@@ -15,9 +15,9 @@ jobs:
   # sof/scripts/parse_sparse_output.sh
   warnings-subset:
 
-    # We're sharing binaries with the zephyr-build container so keep
+    # We're sharing the sparse binary with the zephyr-build container so keep
     # this in sync with it.
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-22.04
 
     strategy:
       fail-fast: false

[aborisovich]

Ok, upgraded OS for Sparse Zephyr workflow to Ubuntu 22,04.
Looks like we are good to go if all checks are green.

[aborisovich]

Wow - codestyle / checkpatch failed with the following message:

WARNING: A patch subject line should describe the change not the tool that found it
#4:
Subject: [PATCH] workflows: Sparse Zephyr upgraded runner OS to Ubuntu 22.04

I sometimes think this checkpatch is chatGTP powered or something...

[plbossart]

Wow - codestyle / checkpatch failed with the following message:

WARNING: A patch subject line should describe the change not the tool that found it
#4:
Subject: [PATCH] workflows: Sparse Zephyr upgraded runner OS to Ubuntu 22.04

I sometimes think this checkpatch is chatGTP powered or something...

it's been that way for a very long time...

[marc-hb]

checkpatch.pl is a Linux kernel tool. The Linux kernel tool has no CI scripts or configuration in the same git repo, nothing like the .github/ directory or similar. So these false positives are rare in the kernel.

While this warning is wrong in this case, it is unfortunately very common to see people submitting bad commit messages like "fix warning from tool X" without describing what the actual code change is. Then this warning is very useful.

[marc-hb]

Also: https://www.kernel.org/doc/html/latest/dev-tools/checkpatch.html

Checkpatch is not always right. Your judgement takes precedence over checkpatch messages. If your code looks better with the violations, then its probably best left alone.

False positives are OK because checkpatch only check PATCHES which means warnings do NOT stick after merge like with other linters.

[aborisovich]

Ok, upgraded Zephyr revision in west.yml. We should be good to go now, however we need to wait for Intel Internal CI results to make sure we have no regression introduced.

[marc-hb]

Thanks again @aborisovich , this solves a big problem!

Something's not right in
https://github.com/thesofproject/sof/actions/runs/5017216786/jobs/8995052856?pr=7645, sparse compilation still looks for the old toolchain. I will look at it after a few meetings.

CMake Error at /zep_workspace/zephyr/cmake/compiler/gcc/target.cmake:12 (message):
  C compiler
  /home/cwd_user/zephyr-sdk-0.16.1/xtensa-intel_s1000_zephyr-elf/bin/xtensa-intel_s1000_zephyr-elf-gcc
  not found - Please check your toolchain installation

EDIT: it's not sparse, it's everything MTL: https://github.com/thesofproject/sof/actions/runs/5017216783/jobs/8995053803?pr=7645

[marc-hb]

I will look at it after a few meetings.

OK this didn't take long: @aborisovich can you please bump sof/west.yml a bit further to zephyrproject-rtos/zephyr@9fc99928caf0 ? We need that one too for MTL

however we need to wait for Intel Internal CI results to make sure we have no regression introduced.

Note the Zephyr SDK is NOT used in sof-ci/jenkins, XCC is used instead. However the Zephyr uprev can of course make a difference! Speaking of which, TGLU_UP_HDA_IPC4ZPH looks pretty bad in https://sof-ci.01.org/sofpr/PR7645/build8107/devicetest/index.html :-( No idea why yet.

https://sof-ci.01.org/sofpr/PR7645/build8106/devicetest/index.html is mostly fine.

[kv2019i]

Thanks @aborisovich !

[kv2019i]

The manifest mtl fails to

https://github.com/thesofproject/sof/actions/runs/5017216783/jobs/8995053803?pr=7645
/home/cwd_user/zephyr-sdk-0.16.1/xtensa-intel_s1000_zephyr-elf/bin/xtensa-intel_s1000_zephyr-elf-gcc

While main mtl test passes and correct compiler is found:

https://github.com/thesofproject/sof/actions/runs/5017216783/jobs/8995053803?pr=7645
-- Found assembler: /home/cwd_user/zephyr-sdk-0.16.1/xtensa-intel_ace15_mtpm_zephyr-elf/bin/xtensa-intel_ace15_mtpm_zephyr-elf-gcc

Why is this happening, the PR is updating the manifest, right...?

[aborisovich]

All green now, thanks @marc-hb .
Let's wait for Intel Internal CI results and merge if positive.

[marc-hb]

Great job, thanks a lot @aborisovich ! Please merge this first thing on Monday, thanks!

https://sof-ci.01.org/sofpr/PR7645/build8158/devicetest/index.html?model=TGLU_RVP_NOCODEC_IPC4ZPH&testcase=multiple-pause-resume-5 has the known corrupted heap issue 7191

https://sof-ci.01.org/sofpr/PR7645/build8157/devicetest is all green.

Checkpatch is wrong.

Everything else is green!