ci: Avoid duplicate builds due to GitHub size limits

[cgturner1]

Updates workflow to stop running --all builds and instead specify platforms not already built in earlier steps.

Signed-off-by: Christopher Turner christopher.g.turner@intel.com

Fixes: 10430

[cgturner1]

Previously we were running one step with build option --all as a final catch all. But with the github runner size limitations we can't afford to build all the targets in one step.

For now I propose running any platform that would've run in --all step that is not already previously built (using the mnfst option). Today it is wcl and imx95. In the future if we are able to increase our runner size or figure out a way to further slim down the docker+zephyr SDK (which has already been done once) we can revert to the --all level of coverage if not then we will have to be mindful if we add another build option to --all in the build script:

sof/scripts/xtensa-build-zephyr.py

Line 157 in 23c4020

platform_configs_all = {

then we will also need to update this workflow with the new platform to increase the test coverage

Successful build of the new de-duplicated step here: https://github.com/thesofproject/sof/actions/runs/20138093625/job/57797124606?pr=10434

[Copilot]

Pull request overview

This PR addresses GitHub storage constraints by modifying the CI workflow to avoid redundant platform builds. Instead of using the --all flag which builds every platform (including those already built in previous steps), the workflow now explicitly specifies only the platforms that haven't been built yet.

Key changes:

• Replaces --all platform flag with explicit platform list (wcl imx95)
• Updates comments to reflect the new approach and reasoning

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[kv2019i]

Thanks @cgturner1 , that seems to do the trick!

[softwarecki]

It would also be nice to add a comment in xtensa-build-zephyr.py about the need to separate enabling CI for a new platform in .github/workflows/zephyr.yml.

[kv2019i]

Only known fails in CI results. The comment to build script can be added in a separate PR. Proceeding with merge as this allows us to clean the CI results now and ensure nothing slips into main.