Arnaldo Carvalho de Melo wrote on Tue, Mar 30, 2021 at 12:12:07PM -0300: > > I can understand reverting due to other pressure to get a release out > > but most distros (including fedora) frown upon vendoring code so I think > > it would be good to have back ultimately. > > > > Did you or someone else (Luca?) ever take the time to look at it? > > Not yet, I'm right now working on it, testing patches to support clang's > thin-LTO being used to build the kernel, so busy with it. Thanks > > I don't see what would be so different with fedora to make this > > unfixable, I'd be happy taking a look if nobody has so far. > > Please do. As said in my previous mail (in reply to Luca's), I cannot reproduce on fedora 33 so I'm not sure where to look at. Was the problem specific to one architecture (I only tried x86_64), or maybe the build dir was tainted by a previous run of cmake with another version of the patch? (if rpmbuild, build dir somehow included in the tarball?) Looking at the timing the libbpf version should have been identical but it's possible that system wasn't up to date or some other version change since Feb? Either way I would appreciate if you could confirm you still have the problem when you have a moment, and if so provide a bit more details on your setup. > > Would you take the patch back in if I somehow fix rpmbuild with a libbpf > > package installed on fedora33? > > Sure. I had it merged, as, IIRC, it was a opt-in procedure. Yes, strictly opt-in and shouldn't affect developers or people building from git without explicitly requiring to use the system's version. -- Dominique
