On Fri, Mar 19, 2021 at 3:22 AM Jean-Philippe Brucker <jean-philippe@xxxxxxxxxx> wrote: > > On Thu, Mar 18, 2021 at 10:08:36AM -0700, Andrii Nakryiko wrote: > > Yeah, makes total sense. I missed that array forces a strong link > > between types. The fix looks good, but can you please add those two > > cases to selftests? There is progs/btf_dump_test_case_syntax.c that > > probably can be extended. Please think about a way to specify types > > such that the order of BTF types doesn't matter and the issue has to > > be handled always. > > Sure, I'll add those selftests. I didn't figure out a way to trigger the > error unconditionally, but given that the selftest is always built with > clang it's still a good regression test. Yes, that's fine. Thanks for the fix and tests, applied to the bpf tree. > > Thanks, > Jean >
