[marc-hb]

4 commits. Main one below. Also: fix toolchain coverage in CI

The -a option has been broken by the addition of platform(s) that don't
have a toolchain available in the Docker image. Every platform can be
built by someone but no one can built -a(ll) platforms right now.

Fix the -a option by changing its definition: make it build what
everyone can build using the public Docker image.

[dbaluta]

@TangleZ I was under the impression that GCC toolchain for imx8ulp is publicly available? Isn't that true anymore?

@marc-hb what do you mean by toolchain to be publicly available?

one can build imx8ulp toolchain from here: https://github.com/thesofproject/crosstool-ng

[lgirdwood]

Lets update our container to GCC11 https://github.com/thesofproject/crosstool-ng/releases/tag/gcc11.2 which has all targets.

[marc-hb]

@marc-hb what do you mean by toolchain to be publicly available?

I meant in the Docker image, see next sentence. Sorry for the confusion. Commit message fixed.

Now looking at https://github.com/thesofproject/sof/commits/main/scripts/docker_build/sof_builder I see that imx8ulp and run were actually added

However they are missing from this current image:

REPOSITORY              TAG       DIGEST                                                     IMAGE ID       CREATED         SIZE
thesofproject/sof       latest    sha256:5cffb0b235e0de1998b608d878f42c5e47d07150d84518d784  7941412b7be9   7 days ago      2.19GB

docker run thesofproject/sof@sha256:3bf321bd4fd7605599be9af6d8459e74a67453a156a72c8a4b76394aba4d5c1b from 4 months ago builds imx8ulp but not rn.

I think we just made a Docker mess, @greg-intel and/or @juimonen can you please help fix it?
Note this incomplete docker image would have been caught if imx8ulp had been part of .github! Just added to this PR - which now fails imx8ulp until the Docker image is fixed.

Lets update our container to GCC11 https://github.com/thesofproject/crosstool-ng/releases/tag/gcc11.2 which has all targets.

Sure but let's please first fix the simple imx8ulp drop/regression in the Docker image.

Note old [BUG] IMX8 QEMU boot failed after SOF docker image upgrade #4186

[lgirdwood]

I have GCC11 locally installed and test with build all -a so need to keep all targets.

[marc-hb]

I have GCC11 locally installed and test with build all -a so need to keep all targets.

Platforms will come and go as they have so the split is useful even if the two lists happen to be identical some of the time.

The split is also a great reminder that the work is not done after merely adding one platform to the script.

[marc-hb]

I think we just made a Docker mess, @greg-intel and/or @juimonen can you please help fix it?

Docker image has been updated to sha256:3d2f4f6c7eba57eb9e4e871365f801f3f0462e45b502665daf670890926c9e49 ded60930e71d (thx) and all platforms except mt8195 (which has never been in https://github.com/thesofproject/sof/commits/main/scripts/docker_build/sof_builder) are now building in Github Actions. So only mt8195 is being removed from -a. It can be re-added super quickly if/when needed.

All Github Actions are green; they build more configurations than ever and the cleanups make any future addition trivial. Please fast track this small PR to quickly fix these recent regressions and catch any future one.

APL_UP2_NOCODEC was unavailable and https://sof-ci.01.org/sofpr/PR5202/build11601/devicetest/?model=BYT_MB_NOCODEC&testcase=check-capture-3times is maybe related to #5086? Neither is related to this PR.

[greg-intel]

Not sure I understand the whitespace choices in pull-request.yml on lines 89 vs. 124, but that's irrelevant to the PR as a whole.

[marc-hb]

Not sure I understand the whitespace choices in pull-request.yml on lines 89 vs. 124

Each company on a different line.

Thanks for the review!

[lgirdwood]

I have GCC11 locally installed and test with build all -a so need to keep all targets.

Platforms will come and go as they have so the split is useful even if the two lists happen to be identical some of the time.

The split is also a great reminder that the work is not done after merely adding one platform to the script.

Can we update the container to GCC11. This will also allow us to enable static analysis as part of the CI build.

[marc-hb]

Can we update the container to GCC11. This will also allow us to enable static analysis as part of the CI build.

That will be great but it's at least one order of magnitude more work than this (which is already done and tested), assuming no issue. Upgrading a compiler always carries some level of risk.

Meanwhile -a is broken (which caused the earlier docker mess) and we have zero compile coverage for AMD and for the new imx8ulp.

Once we have a newer and better container with GCC11 and mt8195, adding mt8195 to -a is a one line change.

[cujomalainey]

I agree with @marc-hb revert first to fix, then update after

[lgirdwood]

Sorry all, I need mt1895 for Zephyr native API work to build on all platforms. We also need GCC11 static analysis as part of CI build. @marc-hb pls drop the last patch

[marc-hb]

@marc-hb pls drop the last patch

Done!

Sorry all, I need mt1895 for Zephyr native API work to build on all platforms.

This script does not build Zephyr...

[marc-hb]

In https://sof-ci.01.org/sof-pr-viewer/#/build/PR5202/build8372328 3 platforms fail with strange errors quoted below. Other platforms pass 100%. That CI does not use the script modified by this PR.

Looks like someone merged some untested (test?) code, @kkarask can you help?

03:41:30.764 => Environment initialization
03:41:30.811 ITERATION TestBasicPowerFlow.test === 1642732890.811658
03:41:30.811 => TestBasicPowerFlow.test
03:41:30.936   => Enter D0 - test setup
03:41:30.936     => Configure SOF
03:41:30.936       => Loading base firmware
03:41:31.280   => Enter D3 - test cleanup
03:41:31.483   => Exception in test.
03:41:31.514     Exception AttributeError occurred at 'C:\\work\\tools\\pyexecnetcache\\cavs_scripts-py\\utilities\\dma_trace_buffer.py':286
                 'NoneType' object has no attribute 'is_set'

03:40:43.283 => Configure SOF
03:40:43.283   => Environment initialization
03:40:43.298   => Loading base firmware
03:40:43.783 => Exception in test.
03:40:43.798   Exception FileNotFoundError occurred at 'C:\\work\\tools\\pyexecnetcache\\cavs_scripts-py\\utilities\\fw_loader_pci.py':550
               [Errno 2] No such file or directory: 'C:\\Testing\\Builds\\pull-5202-merge_414440\\CNL\\sof-cnl.ri'

EDIT: after git commit --amend -C HEAD + force push https://sof-ci.01.org/sof-pr-viewer/#/build/PR5202/build8373289 is now all green. Go figure.

[marc-hb]

https://sof-ci.01.org/sofpr/PR5202/build11721/devicetest/?model=CML_HEL_RT5682&testcase=check-suspend-resume-with-playback-5 looks like thesofproject/linux/issues/3370 and we also have the usual alsabat error on BYT. Everything else OK.

https://sof-ci.01.org/sofpr/PR5202/build11725/devicetest/ has some rtcwake issues + a stuck sof-logger = nothing related to this.

[lgirdwood]

@marc-hb thanks, btw we should update the container at earliest convenience. There is a new MTK platform and toolchain to add too.