On 23/03/2023 18:56, Marc Zyngier wrote: > On 2023-03-23 17:38, Oliver Upton wrote: >> Hey Ryan, >> >> On Thu, Mar 23, 2023 at 12:56:18PM +0000, Ryan Roberts wrote: >>> Hi Oliver, >>> >>> Just a polite nudge on this: I was originally hoping to get these into 6.3 since >>> I thought they were fairly uncontroversial and clearly fixing bugs. What are my >>> chances? >> >> Yes, your changes are indeed uncontroversial :) At least for me, fixes to >> selftests take a strictly lower priority than fixes to the kernel outside of >> a merge window. AFAICT, only LPA systems are affected by the changes here and >> I'm not aware of any of those out in the wild. The first patch is not related to LPA, it adds a missing kernel config to the config fragment so that one of the tests (access_tracking_perf_test) is not skipped. This change will mean our CI system starts actually running the test. > > Agreed. My usual take on fixing tests is that unless the test has been > broken in the current cycle, we can safely delay merging the fix until > the following cycle. Thanks for the explanation. I have a slightly different opinion though (please bare with me through the rant): Being fairly new to Linux development, I'd like to be able to run (all) the selftests as a matter of course to be able to quickly answer the "did I obviously break anything?" question. But there is a lot of friction to even being able to compile, let alone run, the things - undocumented dependencies on libraries (even more difficult when needing to cross compile), undocumented dependencies on kernel configs, test code that is broken and fails to compile, tests that silently skip for difficult to determine reasons, tests that fail even when run against the unmodified kernel, and results buried in copious amounts of logs. These are all paper cuts that make them difficult to use and trust. Or perhaps I'm just doing it wrong... I would love to live in a world where I could confidently take a mainline release, compile and run tests at close-to-zero effort and see all tests running and passing... one day, perhaps. But only if we give more priority to the test code ;-) > > And yes, LPA-capable HW is essentially vapourware at this stage. > >> >> So, unless there is a burning issue, I'd like to defer these patches to the >> 6.4 merge window. Nonetheless, it all looks good to me: >> >> Reviewed-by: Oliver Upton <oliver.upton@xxxxxxxxx> Thanks for that. > > Thanks for that. I'll start queuing 6.4 material once I'm back to my > usual time zone, beginning of next week. Appreciated. Thanks for taking the patches. If you have a rule-of-thumb about the best time to post different types of patches (and what it's best to base them on), I'll try to follow it in future. Thanks, Ryan > > Cheers, > > M.
