Closed marc-hb closed 3 years ago
Unlike the original commit, this revert has been tested in TEST PR https://github.com/thesofproject/sof/pull/4783 and everything is passing.
merging to keep the daily tests (starts in 15mn) and week-end stress tests operational.
merging to keep the daily tests (starts in 15mn) and week-end stress tests operational.
Thanks for merging. There was no rush for validation because the rimage version in SOF is lagging a few commits behind and right now it has neither the regression (otherwise it would have been tested and caught) nor this revert.
It's still great that developers can use the latest rimage branch again.
Reminder: the README.md now has instructions on how to run SOF tests on rimage changes before rimage merge. See examples above.
There was no rush for validation because the rimage version in SOF is lagging a few commits behind and right now it has neither the regression (otherwise it would have been tested and caught) nor this revert.
I don't get why CI does not test the latest rimage, or why with a delay? This is nuts.
I don't get why CI does not test the latest rimage, or why with a delay? This is nuts.
This is how git submodules work. They're designed for relatively static libraries that get upgraded rarely and manually.
From https://docs.zephyrproject.org/latest/guides/west/why.html#rationale-for-a-custom-tool
git submodules do not support continuous tracking of the latest HEAD in external repositories (R4)
See also
https://github.com/thesofproject/sof/issues/3517 [FEATURE] drop git submodules and switch to Zephyr's "west" multi-repo tool
The counterpart of validating a moving dependency is you're never quite sure what you're testing because it can be changing at any moment; even in the middle of a MIDDLE of test run! This is not theoretical, it happened to me once in the middle of a Zephyr PR: two shards tested slightly different versions. One failed the other not. Fortunately I had added some version logging in the build logs before it happened, otherwise I would have lost much more hair.
I don't get why CI does not test the latest rimage, or why with a delay? This is nuts.
This is how git submodules work. They're designed for relatively static libraries that get upgraded rarely and manually.
Maybe, but if it's know to be isolated from the automatic tests, then the author should know better and trigger a manual test.
Maybe, but if it's know to be isolated from the automatic tests, then the author should know better and trigger a manual test.
Yes. Except the submodule indirection requires some effort and submodule proficiency.
I just added a guide in the README that should help a bit.
@plbossart think we are good now, we have a process for testing rimage updates now. @mrajwa fyi - can you re-submit with the cmd line / Kconfig option. Thanks.
@lgirdwood yep, I have it on my TODO list will upstream soon.
will this command line option be required only for the new stuff? it'd be a problem if we have to update the build and CI scripts for existing platforms.
will this command line option be required only for the new stuff? it'd be a problem if we have to update the build and CI scripts for existing platforms.
I know only two places invoking rimage right now: https://github.com/thesofproject/sof/blob/main/src/arch/xtensa/CMakeLists.txt https://github.com/thesofproject/sof/blob/main/scripts/xtensa-build-zephyr.sh
Great opportunity to review and cleanup the first one @mrajwa . Ideally, please submit the cleanup first and the new feature later; think Continuous Integration.
Actually: we already have per-platform config files with sizes, offsets and what not: https://github.com/thesofproject/rimage/tree/main/config
Does this really require a new command line option?
This reverts commit 241af313dab49db06d6e6f0af3330f1a916de664 so the DSP can boot again.
On my APL UP2 this fixes:
Test PR https://github.com/thesofproject/sof/pull/4782 show this commit broke the DSP boot on ALL platforms https://sof-ci.01.org/sofpr/PR4782/build10387/devicetest/
This also broke QEMU boot on a number of platforms: BDW, BYT and CHT, see
error: invalid firmware signature
inhttps://sof-ci.01.org/sofpr/PR4782/build10387/boottest/ and https://github.com/thesofproject/sof/pull/4782/checks?check_run_id=3634915322
How was it tested?
Signed-off-by: Marc Herbert marc.herbert@intel.com