Camera testcases are skipping on rb3gen2-core-kit, passing on iq-9075-evk

[qcom-anilyada]

[Libcamera_cam] - No sensors reported by 'cam -l' - marking SKIP
[Camera_RDI_FrameCapture] SKIP – Media node not found

Latest Job: https://lava.infra.foundries.io/scheduler/job/376472

while on wrynose branch it is passing

When tracing back, the issue was observed around the below PRs:

#2712

• https://github.com/qualcomm-linux/meta-qcom/actions/runs/29155835384 [SKIPPED]
• https://github.com/qualcomm-linux/meta-qcom/actions/runs/29056606175 [PASSED]

#2684

• https://github.com/qualcomm-linux/meta-qcom/actions/runs/28843072885 [PASSED]
• https://github.com/qualcomm-linux/meta-qcom/actions/runs/28853513248 [PASSED]
• https://github.com/qualcomm-linux/meta-qcom/actions/runs/28867146019 [PASSED]
• https://github.com/qualcomm-linux/meta-qcom/actions/runs/29246527224 [SKIPPED]

#2711

• https://github.com/qualcomm-linux/meta-qcom/actions/runs/29074060595 [PASSED]
• https://github.com/qualcomm-linux/meta-qcom/actions/runs/29223322233 [SKIPPED]
• https://github.com/qualcomm-linux/meta-qcom/actions/runs/29247504880 [SKIPPED]

[mwasilew]

Naresh Kamboju (@nareshkamboju) could you take a look at devices in Hyd?

[nareshkamboju]

Milosz Wasilewski (@mwasilew) I am working with Anil (@qcom-anilyada) to find out the real reason for these skips.

[qcom-anilyada]

the root cause of the skips in rb3gen2-core-kit specifically is the cdt bins we were flashing as part of post processing step which seems to be affected by the new qdl flashing method

This part of the code:
https://github.com/qualcomm-linux/lava-test-plans/blob/d880bca463efc30f226e1fce1e797c8d7d93f7b6/lava_test_plans/devices/rb3gen2-core-kit#L7-L14

is there a way to add post processing steps in the qdl method of flashing as well?

[lumag]

Could you provide an explanation, why/how the CDT affects the skips and what would you like to change?

[mwasilew]

As I remember the default CDT is for "core-kit". In the old jobs we updated it to use the "vision-kit" CDT. Both are shipped with the build. TBH I don't know what is the proper solution here. Maybe multimedia image should enforce vision-kit CDT as default?

[lumag]

I'd rather not infer machine type from the image type.
Anyway, I see that the testing script already checks the tests and switches to the vision kit. Is there anything to be done other than that?

[mwasilew]

Dmitry Baryshkov (@lumag) the problem is that we're currently only using "core-kit" CDT. The old approach with "flasher" script was able to change the CDT before flashing. It is possible to repack the CDT binary in the tarball before flashing, but someone has to implement this in the test plan.

[lumag]

Dmitry Baryshkov (@lumag) the problem is that we're currently only using "core-kit" CDT. The old approach with "flasher" script was able to change the CDT before flashing. It is possible to repack the CDT binary in the tarball before flashing, but someone has to implement this in the test plan.

Ok, thanks for the explanation.