On Tue, 14 May 2019 at 17:46, Kees Cook keescook@chromium.org wrote:
On Tue, May 14, 2019 at 05:28:45PM +0200, Anders Roxell wrote:
On Sat, 11 May 2019 at 16:51, Kees Cook keescook@chromium.org wrote:
Test Suites
- boot-lkft-kselftests-master-519
What counts as a "pass" for this? I looked at the x86_64 log, and there are lots of kselftest failures, but the dashboard counts it as a "pass"?
The boot is passed when the auto-login-action in suite 'lava' is passed [1].
There's is two issues:
- We had a way that changed the run_kselftest.sh file in order to
skip tests that hanged (that results in no test output from the run) various boards. Since the new change to kselftest, the patch to run_kselftest.sh got added that breaks our way to skip the tests and kernel hangs.
aha, right that's awesome. So its only scripts that we have to skip now then.
There's a fix in -next that adds an "alarm" check now so no tests should ever hang for more than 30 seconds now: https://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest.git/co...
- The way we parse the output doesn't match the new output format.
We will turn off the "next-* kselftest restults" until we have resolved this issue.
I'd prefer to keep the tests running -- they're still valuable logs to have. I was just curious about the specific reporting. As long as the issue is known, that's fine. No need to disable it entirely. :)
We will still run the tests on next, the plan is just to stop sending reports until we have the parsing fixed.
Cheers, Anders
-- Kees Cook