Re: auto-split of commit. Was: [PATCH bpf-next 04/10] tools/bpf: add libbpf_prog_type_(from|to)_str helpers

["Toke Høiland-Jørgensen" <toke@redhat.com>]

Greg Kroah-Hartman <gregkh@linuxfoundation.org> writes:

> On Thu, Aug 29, 2019 at 10:16:56AM -0700, Alexei Starovoitov wrote:
>> On Thu, Aug 29, 2019 at 08:51:51AM +0200, Greg Kroah-Hartman wrote:
>> > On Wed, Aug 28, 2019 at 04:46:28PM -0700, Alexei Starovoitov wrote:
>> > > On Wed, Aug 28, 2019 at 04:34:22PM -0700, Jakub Kicinski wrote:
>> > > > 
>> > > > Greg, Thomas, libbpf is extracted from the kernel sources and
>> > > > maintained in a clone repo on GitHub for ease of packaging.
>> > > > 
>> > > > IIUC Alexei's concern is that since we are moving the commits from
>> > > > the kernel repo to the GitHub one we have to preserve the commits
>> > > > exactly as they are, otherwise SOB lines lose their power.
>> > > > 
>> > > > Can you provide some guidance on whether that's a valid concern, 
>> > > > or whether it's perfectly fine to apply a partial patch?
>> > > 
>> > > Right. That's exactly the concern.
>> > > 
>> > > Greg, Thomas,
>> > > could you please put your legal hat on and clarify the following.
>> > > Say some developer does a patch that modifies
>> > > include/uapi/linux/bpf.h
>> > > ..some other kernel code...and
>> > > tools/include/uapi/linux/bpf.h
>> > > 
>> > > That tools/include/uapi/linux/bpf.h is used by perf and by libbpf.
>> > > We have automatic mirror of tools/libbpf into github/libbpf/
>> > > so that external projects and can do git submodule of it,
>> > > can build packages out of it, etc.
>> > > 
>> > > The question is whether it's ok to split tools/* part out of
>> > > original commit, keep Author and SOB, create new commit out of it,
>> > > and automatically push that auto-generated commit into github mirror.
>> > 
>> > Note, I am not a laywer, and am not _your_ lawyer either, only _your_
>> > lawyer can answer questions as to what is best for you.
>> > 
>> > That being said, from a "are you keeping the correct authorship info",
>> > yes, it sounds like you are doing the correct thing here.
>> > 
>> > Look at what I do for stable kernels, I take the original commit and add
>> > it to "another tree" keeping the original author and s-o-b chain intact,
>> > and adding a "this is the original git commit id" type message to the
>> > changelog text so that people can link it back to the original.
>> 
>> I think you're describing 'git cherry-pick -x'.
>
> Well, my scripts don't use git, but yes, it's much the same thing :)
>
>> The question was about taking pieces of the original commit. Not the whole commit.
>> Author field obviously stays, but SOB is questionable.
>
> sob matters to the file the commit is touching, and if it is identical
> to the original file (including same license), then it should be fine.
>
>> If author meant to change X and Y and Z. Silently taking only Z chunk of the diff
>> doesn't quite seem right.
>
> It can be confusing, I agree.
>
>> If we document that such commit split happens in Documentation/bpf/bpf_devel_QA.rst
>> do you think it will be enough to properly inform developers?
>> The main concern is the surprise factor when people start seeing their commits
>> in the mirror, but not their full commits.
>
> Personally, I wouldn't care, but maybe you should just enforce the fact
> that the original patch should ONLY touch Z, and not X and Y in the same
> patch, to make this a lot more obvious.
>
> Patches should only be doing "one logical thing" in the first place, but
> maybe you also need to touch other things when doing a change that you
> can't do this, I really do not know, sorry.

As someone who has been asked to split otherwise logically consistent
patches to accommodate the sync, I would very much appreciate it if the
process could be set up so this was not necessary :)

-Toke