west.yml: update zephyr to 740d7f735e2 #9291
Conversation
|
@lyakh I don't think the power_down assembly changes are working out -- lot of fails on ACE20 |
Total of 369 commits. Changes include: 46b356a6b0e intel-adsp: fix ACE power-off assembly fb53d2ef8dd ace: power: replace pseudo-assembly movi Signed-off-by: Tomasz Leman <tomasz.m.leman@intel.com>
4f3da74 to
6667f81
Compare
|
Good to go, waiting for second +1 review so we can merge. |
|
This Zephyr upgrade is likely part of the reasons that caused MTL to stop working entirely: #9308 |
|
I don't fully get how this Zephyr upgrade was successful CI-wise, but somehow post-merge we are in the weeds. Does this point to a testing problem or an obsolete configuration used in tests @kv2019i ? |
|
@plbossart I think this is a feature of the current system. We can have cases where two PRs pass the tests independetly, but when they are both in the tree, the test fails. I think this happened here. To catch these, we'd have to trigger CI again for all other PRs, when we merge a PR and this would be very costly. @marc-hb can comment much better on the options we have. |
This is a well-known CI issue and Gitlab and Github have a solution for it: https://about.gitlab.com/blog/2020/01/30/all-aboard-merge-trains/ The Chrome project also had some project-specific "merge queue" concept for a very long time. I don't know whether Jenkins has something like this. Even if it has, this would be a major change not worth the effort for something that happens very rarely in SOF. We have much more frequent problems to solve first (for instance: merging known broken PRs). The solution above are designed for repos with a much higher churn, like Zephyr for instance. But even Zephyr does not use this; they just revert very quickly when this happens. |
Total of 369 commits.
Changes include:
46b356a6b0e intel-adsp: fix ACE power-off assembly
fb53d2ef8dd ace: power: replace pseudo-assembly movi