Repository navigation
[BUG] IPC3/4 fuzzer build broken with Zephyr PR 71378 #9101
Description
Describe the bug
As discussed in zephyrproject-rtos/zephyr#71378 (review) , recent changes in Zephyr mainline break the SOF fuzzer builds and the github CI checks using these builds.
To Reproduce
https://github.com/thesofproject/sof/actions/runs/8972185189/job/24639608648?pr=9098
Reproduction Rate
100%
Expected behavior
Fuzzer build works.
Impact
Fuzzer not usable with SOF.
Environment
https://sof-ci.01.org/sofpr/PR9098/build4470/build/index.html
Parsing /home/runner/work/sof/sof/workspace/zephyr/Kconfig
/home/runner/work/sof/sof/workspace/build-fuzz/zephyr/misc/generated/extra_kconfig_options.conf:2: warning:
attempt to assign the value '100' to the undefined symbol ARCH_POSIX_FUZZ_TICKS
error: Aborting due to Kconfig warnings
Loaded configuration '/home/runner/work/sof/sof/workspace/zephyr/boards/native/native_posix/native_posix_defconfig'
Merged configuration '/home/runner/work/sof/sof/workspace/sof/app/prj.conf'
Merged configuration '/home/runner/work/sof/sof/workspace/sof/app/stub_build_all_ipc3.conf'
Merged configuration '/home/runner/work/sof/sof/workspace/build-fuzz/zephyr/misc/generated/extra_kconfig_options.conf'
CMake Error at /home/runner/work/sof/sof/workspace/zephyr/cmake/modules/kconfig.cmake:391 (message):
command failed with return code: 1
Call Stack (most recent call first):
/home/runner/work/sof/sof/workspace/zephyr/cmake/modules/zephyr_default.cmake:132 (include)
/home/runner/work/sof/sof/workspace/zephyr/share/zephyr-package/cmake/ZephyrConfig.cmake:[66](https://github.com/thesofproject/sof/actions/runs/8972185189/job/24639608648?pr=9098#step:7:67) (include)
/home/runner/work/sof/sof/workspace/zephyr/share/zephyr-package/cmake/ZephyrConfig.cmake:92 (include_boilerplate)
CMakeLists.txt:5 (find_package)
Activity
kv2019i commentedon May 7, 2024
Marking as P1 for now as we have the fuzzer tests in our github PR test flow.
andyross commentedon May 7, 2024
Hm. That PR's original version was breaking, but the refactored one was safe I thought. Looks like it's just a missing kconfig we can replace upstream, give me a moment to try it out.
andyross commentedon May 7, 2024
Confirmed that reverting the PR works to recover a fuzz build:
git revert --no-edit 2bff629d296f...82fa83048e66A few things got moved to the sample that were part of the core fuzz rig and I missed them in review. Trying to decide whether it's best to move those back piecewise upstream, or just add them to SOF as that's the new architecture anyway. (The end result here is that the "outer loop" in the fuzzer becomes part of the app and not the architecture layer, in preparation for deprecating "native_posix" in favor of the equivalent "native_sim" arch which is based on an out of tree project).
platform/posix: Unbreak fuzzer support
andyross commentedon May 7, 2024
First cut at a fix up, @marc-hb I'm not 100% sure how to exercise all the CI cases (though... maybe CI will just run them on the PR, heh) so wouldn't hurt for you to give it a spin.
platform/posix: Unbreak fuzzer support
platform/posix: Unbreak fuzzer support
platform/posix: Unbreak fuzzer support
platform/posix: Unbreak fuzzer support