Re: [PATCH] selftests: Deescalate error reporting

[Shuah Khan]

On 4/16/26 07:19, Mark Brown wrote:
> Commit 7e47389142b8 ("selftests: Preserve subtarget failures in
> all/install") updated the propagation of errors from indivdual kselftest
> targets to be similar to that seen with FORCE_TARGETS.  While it would
> be really nice to be in a position to do this currently it is premature
> to do this as the default behaviour.
>
> At present we default to trying to build all selftests but a combination
> of code quality issues and build dependencies mean that it is almost
> certain that at least one of them will fail to build (for example,
> several depend on clang so don't work in a GCC container) and a top
> level failure in the kselftest build reported.  Further, the resulting
> failures mean that the install target does not run at all so any build
> problem is escallated to a complete failure to produce a kselftest
> tarball so CI systems that run into issues loose all selftests coverage.
> This has been causing disruption to a range of CI systems including
> KernelCI, mine and Arm's internal one.
>
> Revert the commit, users who need this behaviour should be able to use
> FORCE_TARGETS for the time being.  At present users that do this (such
> as linux-next) are most likely building a subset of targets known to
> succeed in their environments.
>
> Fixes: 7e47389142b8 ("selftests: Preserve subtarget failures in all/install")
> Signed-off-by: Mark Brown <broonie@xxxxxxxxxx>
> ---
>   tools/testing/selftests/Makefile | 8 ++++----
>   1 file changed, 4 insertions(+), 4 deletions(-)

Thank you. I applied it to linux-kselftest next and will send pr tomorrow.

thanks,
-- Shuah