On Wed, Jul 13, 2022 at 08:18:10PM +0000, Daniel Müller wrote: > On Wed, Jul 13, 2022 at 09:18:43AM -0700, sdf@xxxxxxxxxx wrote: > > On 07/13, Zeng, Oak wrote: > > > Thank you sdf for the reply. > > > > > It is news to me that samples/bpf tend to go stale. When I looked the > > > samples/bpf folder git history, the last update is from only 2 months > > > ago. And yes I can see samples/bfp is not actively updated recently. We > > > are from Intel's GPU group and we are working on some bpf tools for GPU > > > profiling purpose. We made our work based on the structure of > > > samples/bpf because we can conveniently use libbpf. We chose bpf c > > > frontend (vs python frontend) because python bpf program seems can't > > > execute under non-root leading to some security concerns. This work is > > > not yet upstream but we planned to upstream it. > > > > I'm mostly talking from the following perspective: > > http://vger.kernel.org/bpfconf2022_material/lsfmmbpf2022-bpf-ci.pdf > > > > BPF maintainers have a CI that continuously builds and runs > > tools/testing/selftests/bpf. I don't think it includes samples/bpf; > > that's what I mean by "go stale". Eventually, people fix samples, but > > there is no continuous system to verify they are healthy. > > I am looking into some aspects of the CI. I will see what it would take to > build and/or test samples/bpf in an automated fashion there. For what it's worth, we now [0] have samples/bpf built in BPF CI. Here is an example run: https://github.com/kernel-patches/bpf/runs/7527372083 (it may not be possible for folks not part of the organization to see detailed logs, but you should be able to see the "Build samples" step) Thanks, Daniel [0] https://github.com/kernel-patches/vmtest/pull/98
