Re: [Qemu-devel] [PATCH 2/4] virtio-input: core code & base class

[Michael S. Tsirkin]

On Thu, Mar 19, 2015 at 03:38:59PM +0100, Gerd Hoffmann wrote:
> On Do, 2015-03-19 at 12:37 +0100, Michael S. Tsirkin wrote:
> > On Thu, Mar 19, 2015 at 12:33:01PM +0100, Gerd Hoffmann wrote:
> > >   Hi,
> > > 
> > > > > Adding not-yet upstream files to the list of headers to sync over from
> > > > > linux isn't sane IMHO.
> > > > > 
> > > > > cheers,
> > > > >   Gerd
> > > > > 
> > > > 
> > > > Of course it isn't. But you can sync it manually. Once it's upstream,
> > > > add it to list for automated sync.
> > > 
> > > Hmm, well, not that simple, there is no whitelist for virtio headers but
> > > a "*virtio*.h" glob.  So it'll simply sync automatic once landed
> > > upstream, and may get deleted by accident before.
> > 
> > Well no - by design we only update headers, never delete them.
> 
> Hmm?
> 
> I'm talking about scripts/update-linux-headers.sh in the qemu tree.
> That does a "rm -rf" followed by copying over the bits from linux
> kernel.  So when switching from my virtio-input branch back to master
> (in the linux kernel source tree) and re-running the script the header
> file will disappear ...
> 
> Sounds like you are talking about something completely different.
> 
> cheers,
>   Gerd

What I referred to is that we don't do any of git add/git rm.
User must manually run these to add/remove new headers.

-- 
MST