On 28/07/2017 14:13, Cornelia Huck wrote: >>> But would it not a better idea to add an entry for all headers touched >>> by update-linux-headers, then? >> >> One thing doesn't exclude the other. That entry would also list the >> script itself and linux-headers. KVM, VFIO and virtio patterns can only >> list the files they care about. Migration could list userfaultfd, >> possibly. And there's also include/standard-headers/, which includes >> PCI and input subsystem files... It has to be perfected, but it's a >> good idea. > > Personally, I'm not really a fan. If something is changed in one of the > headers, it implies that either it has already been changed in the > original headers in Linux (and I will have seen it then), or that > someone is sending a preliminary patch (and I should hope that I'm > already cc:ed for the changes that this headers change is for then). It > does not really hurt, but it feels wrong.
Sometime people pull changes too early, or otherwise I want to comment on the timing of the change... I don't know, I already get too much qemu-devel email. :) Paolo