✗ CI.xeBAT: failure for lib/kunit: Support writable filter* parameters of kunit module (rev2)

[Lucas De Marchi]

On Thu, Feb 01, 2024 at 12:06:35PM +0100, Janusz Krzysztofik wrote:
>On Wednesday, 31 January 2024 21:51:15 CET Lucas De Marchi wrote:
>> On Wed, Jan 31, 2024 at 08:22:35PM +0100, Janusz Krzysztofik wrote:
>> >On Wednesday, 31 January 2024 19:36:57 CET Patchwork wrote:
>> >> == Series Details ==
>> >>
>> >> Series: lib/kunit: Support writable filter* parameters of kunit module
>> >(rev2)
>> >> URL   : https://patchwork.freedesktop.org/series/129174/
>> >> State : failure
>> >>
>> >> == Summary ==
>> >>
>> >> CI Bug Log - changes from XEIGT_7699_BAT -> XEIGTPW_10613_BAT
>> >> ====================================================
>> >>
>> >> Summary
>> >> -------
>> >>
>> >>   **FAILURE**
>> >>
>> >>   Serious unknown changes coming with XEIGTPW_10613_BAT absolutely need to
>> >be
>> >>   verified manually.
>> >>
>> >>   If you think the reported changes have nothing to do with the changes
>> >>   introduced in XEIGTPW_10613_BAT, please notify your bug team (I915-ci-
>> >infra at lists.freedesktop.org) to allow them
>> >>   to document this new failure mode, which will reduce false positives in
>> >CI.
>> >>
>> >>
>> >>
>> >> Participating hosts (4 -> 4)
>> >> ------------------------------
>> >>
>> >>   No changes in participating hosts
>> >>
>> >> Possible new issues
>> >> -------------------
>> >>
>> >>   Here are the unknown changes that may have been introduced in
>> >XEIGTPW_10613_BAT:
>> >>
>> >> ### IGT changes ###
>> >>
>> >> #### Possible regressions ####
>> >>
>> >>   * igt at xe_live_ktest@bo:
>> >>     - bat-adlp-7:         [PASS][1] -> [SKIP][2] +2 other tests skip
>> >>    [1]: https://intel-gfx-ci.01.org/tree/intel-xe/IGT_7699/bat-adlp-7/
>> >igt at xe_live_ktest@bo.html
>> >>    [2]: https://intel-gfx-ci.01.org/tree/intel-xe/IGTPW_10613/bat-adlp-7/
>> >igt at xe_live_ktest@bo.html
>> >
>> >Hi Lucas,
>> >
>> >We've got an empty list of test cases because the current algorithm of IGT
>> >KUnit implementation expects subtest name provided via struct kunit_tests
>> >to be the same as name of a test suite provided by the KUnit test module.  As
>> >a special case, if the module provides only one test suite (most do) and the
>> >name of it is the same as the module name with the _test or _kunit suffix
>> >truncated (also the most common case) then subtest name may be omitted (i.e.,
>> >.name field may be NULL).
>> >
>> >I have to document the above, but we also have to use the same names on both
>> >sides (KUnit test module and its corresponding IGT test).  Now we don't, e.g.
>> >test suite named "xe_bo" is provided by xe_bo_test module while subtest name
>> >"bo" is specified for that module in IGT xe_live_ktest.
>>
>> in the end we want to have only one module: xe_live_test.ko and then
>> filter out by the testunit. From what I've seen previously in igt, it
>> would be a matter of always reloading the same module, but passing a
>> different name.
>
>As long as the module will provide several test suites, each of those test
>suites can be mapped to a separate subtest with the same name as the test
>suite, and with its test cases discovered and executed automatically by IGT as
>dynamic sub-subtests, still selectable individually via IGT --dynamic command
>line option.  For that to work this way, populating .name fields of the struct
>kunit_tests table with subtest names that are the same as test suite names is
>mandatory, otherwise IGT will expose one subtest, named after the module name,
>with dynamic sub-subtests named <suite_name>-<case_name>.

that's what I did when I submitted the module rename:
https://lore.kernel.org/igt-dev/20231205224926.2188996-1-lucas.demarchi@intel.com/

This one is to be applied on top of these fixes to kunit handling.

Lucas De Marchi