Hi,
I've been working with colleagues at Collabora to improve the testing quality overall for kselftests since several months. We had identified following key points to improve: * Make non-TAP conformant tests, TAP conformant to catch the bugs/failures in the CI and non-CI environment. Without clear success or failure message, it is difficult and at times impossible to notice the bugs and which part of the test or sub-test has failed. * Add config fragment if it is already isn't present. * Improve ancient tests to be robust. KernelCI helps a lot in finding flakiness or non-robust pieces of code by building it in range of configurations and running it on range of hardware.
As new KernelCI is in works, we are identifying which kselftest suites could be most suitable to be enabled in the start to keep the noise to a minimum. The criteria to enable a suite on KernelCI is: * The test suite is TAP compliant * The test suite passes in preliminary testing on x86_64 and ARM64 platforms * The test suite builds fine with Clang in-addition to gcc
To facilitate transparency and track our progress, I am diligently maintaining a spreadsheet [1] with detailed information about each suite's status and requirements.
Furthermore, I have commenced exploration into KUnit testing, and initial results are promising. I plan to delve deeper into KUnit testing and will provide further updates in the coming days.
Your feedback, insights and collaboration on any aspect of these initiatives would be highly valuable. We greatly appreciate your collaboration and support as we continue to enhance the testing quality of kselftests.
[1] https://docs.google.com/spreadsheets/d/1XRCgxlY1b74aIOIXQ7qDR-62KTJmuV1ow9st...