On Fri, Sep 05, 2025 at 05:03:26PM +0530, Donet Tom wrote:
[...]
>> +FIXTURE(split_pmd_thp_to_order) {
>> +};
>> +
>> +FIXTURE_VARIANT(split_pmd_thp_to_order) {
>> + int order;
>> +};
>> +
>> +FIXTURE_SETUP(split_pmd_thp_to_order)
>> +{
>> + if (variant->order >= pmd_order)
>> + SKIP(return, "order %d is not supported", variant->order);
>
>
>On a 64 KB page size system, the page order is 5, so the remaining tests
>are getting skipped. On such systems, these tests are effectively invalid.
>Instead of skipping them, would it be better to simply not account those
>tests?
>
> # PASSED: 64 / 64 tests passed.
> # 30 skipped test(s) detected. Consider enabling relevant config options to
>improve coverage.
> # Totals: pass:34 fail:0 xfail:0 xpass:0 skip:30 error:0
>
>
>Here, the 30 skipped tests are invalid, and the total valid tests are 34.
>Would it make sense to exclude the 30 invalid tests entirely for 64KB page
>size systems?
>
Yes, this is a little annoying. But I don't find a way to achieve what you
mentioned with kselftest_harness. :-(
--
Wei Yang
Help you, Help me
