west.yml: upgrade Zephyr revision - update usage of IPC4 ipc_platform_send_msg function signature

[aborisovich]

During last changes in Zephyr that implemented power transition for the IPC Device, signature of the intel_adsp_ipc_send_message had changed and now returns negative int error codes (previously bool on success). Updated single function reference in SOF IPC4 ipc_platform_send_msg().
Implementation had not changed as the function may return only -EBUSY error code until the Zephyr Device Power Management option is disabled.

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

[marc-hb]

Thanks, for submitting this separately from #7325, this will provide finer-grained daily test results. Continuous Integration and git bisect for the win!

[aborisovich]

Corrected commit SHA (zephyrproject-rtos/zephyr@23b3cae)
Previous was incorrect (for some reason SHA on PR is different than SHA after rebase merge).

[tmleman]

We should fix this before merging:

In file included from <command-line>:
/zep_workspace/zephyr/boards/xtensa/nxp_adsp_imx8m/nxp_adsp_imx8m.dts:10:10: fatal error: nxp/nxp_imx/mimx8ml8dvnlz-pinctrl.dtsi: No such file or directory
   10 | #include <nxp/nxp_imx/mimx8ml8dvnlz-pinctrl.dtsi>
      |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.

[kv2019i]

@tmleman wrote:

We should fix this before merging:

Although isn't this error in the mainline build? In the build with Zephyr from this PR (commit 23b3cae1b ), the error is not seen. Or am I missing something?

[aborisovich]

Although isn't this error in the mainline build? In the build with Zephyr from this PR (commit 23b3cae1b ), the error is not seen. Or am I missing something?

Yeah, it's actually zmain revision failing not the manifest. I think we can merge, nxp will resolve their refactors in other revision upgrade.

[tmleman]

Although isn't this error in the mainline build? In the build with Zephyr from this PR (commit 23b3cae1b ), the error is not seen. Or am I missing something?

@aborisovich @kv2019i I haven't seen this error in any other PR so my assumption was it was caused by the zephyr update. If that's not the case we can proceed with merge.

[tmleman]

Off-topic: I was convinced that those are build created using Zephyr SDK (with revisions from manifest). Maybe they should be labeled better?

[kv2019i]

One unrelated PM fail in https://sof-ci.01.org/sofpr/PR7857/build10051/devicetest/index.html . One known pause-resume fail in https://sof-ci.01.org/sofpr/PR7857/build10052/devicetest/index.html . Proceeding with merge.

[marc-hb]

@tmleman naming is hard :-) https://www.martinfowler.com/bliki/TwoHardThings.html

It's especially hard here because of the very limited space on the fixed-size, left column.

Unlike many other CIs, there is absolutely nothing hidden/secret in Github Actions, it's ALL in the repo itself. So by all means please feel free to propose and even experiment with alternatives, see commit 4311da1 for a starting point.

If you do experiment, for pure .github/ changes please add a [SKIP SOF-TEST] prefix in the PR title to avoid wasting hardware cycles. Commit titles don't matter for this, only the PR title does.

There is no "self-hosted" runner used in Github Actions right now, it's all running on the default Ubuntu images provided by Github. There are 1 or 2 docker images used on top but they're all open-source too, no proprietary toolchain has ever been used in Github Actions.

Cadence toolchains are used outside Github Actions, see #7841 for some recent discussion on that topic.

[marc-hb]

Thanks, for submitting this separately from #7325, this will provide finer-grained daily test results. Continuous Integration and git bisect for the win!

In the "I hate to be right" category: this Zephyr PR is a prime suspect for a regression, see discussion in #7869 and others.

Except #7325 has already been merged now :-(

[marc-hb]

New bug filed by @fredoh9 (thx!) for this regression:

• [BUG][MTL] MOD_SET_DX causes DSP exception (very recent regression) #7875