On Fri, May 06, 2022 at 03:46:19PM -0700, Andrii Nakryiko wrote: > You meant > > - [ARG_PTR_TO_UNINIT_MEM] = &mem_types, > > parts as stand-alone patch? That would be invalid on its own without > adding MEM_UNINT, so would potentially break bisection. So no, it > shouldn't be a stand-alone patch. Each patch has to be logically > separate but not causing any regressions in behavior, compilation, > selftest, etc. So, for example, while we normally put selftests into > separate tests, if kernel change breaks selftests, selftests have to > be fixed in the same patch to avoid having any point where bisection > can detect the breakage. Thanks for clarifying, Andrii. I misunderstood the existing verifier code while I was reviewing the diff and mistakenly thought it was safe to remove these lines without MEM_UNINIT. Apologies for the noise / confusion.
