On Wed, 2 Apr 2025 18:55:29 +0100 Mark Brown wrote:
On Wed, Apr 02, 2025 at 10:36:55AM -0700, Jakub Kicinski wrote:
The Cirrus tests keep failing for me when run on x86 ./tools/testing/kunit/kunit.py run --alltests --json --arch=x86_64 https://netdev-3.bots.linux.dev/kunit/results/60103/stdout
You've not said what tree you're testing there but everything is fine in -next AFAICT, and there is one fix for cs_dsp on it's way to Linus at the minute specifically for KUnit.
It seems like new cases continue to appear and we have to keep adding them to the local ignored list. Is it possible to get these fixed or can we exclude the cirrus tests from KUNIT_ALL_TESTS?
This is the first report I've seen from you...
More of a test balloon that a real report as you alluded to. I was wondering if it's just me, and from your response it seems to be just me.
I did more digging and with newer compilers I don't see the issue either. So I did:
./tools/testing/kunit/kunit.py run --alltests --json --arch=x86_64 \ --kconfig_add CONFIG_INIT_STACK_ALL_ZERO=n
And the problem reproduces on recent compilers, too. Which leads me to thinking that the tests are broken but stack init covers it up.. ? Strangely tho
make W=1 LLVM=1 ...
doesn't report any uninitialized variable warnings.