Re: Lots of "Warning (unit_address_vs_reg): Node /blah/blah/blah has a unit name, but no reg property"

[Ian Campbell <ijc-KcIKpvwj1kUDXYZnReoRVg@public.gmane.org>]

On Mon, 2017-01-30 at 10:46 -0600, Rob Herring wrote:
> +dtc list
> 
> On Mon, Jan 30, 2017 at 3:55 AM, Ian Campbell <ijc-KcIKpvwj1kUDXYZnReoRVg@public.gmane.org>
> wrote:
> > I recently updated the device-tree-compiler package on the system
> which
> > builds the split device tree repo[0] from 1.4.0+dfsg-2 to 1.4.2-1
> > (debian package versions) and now during the test build I see many
> > (over 1,000) of these:
> >
> >     Warning (unit_address_vs_reg): Node /soc@ff700000/ethernet@b000
> 0 has a unit name, but no reg property
> >
> > Is this something which is known? Maybe Linux hasn't pulled in the
> > newer dtc yet so you aren't hitting it yet, or maybe you've done
> what I
> > did as a quick fix and added:
> >     DTC_FLAGS="-W no-unit_address_vs_reg"
> > ?
> 
> Linux has this, but the warnings are only enabled with "W=1" builds.
> I'm trying to not have to review the trivial crap.

Specifically Linux adds the DTC_FLAGS="-W no-unit_address_vs_reg" only
if W=0 (rather than enabling it if W=1).

I'll disable that one in the split tree too then.

> It's going to spew even more soon. I've posted some more dtc checks
> last week.

Fun!

> > (a second issue is the lack of the offending filename in the
> > warning message, I'll figure out the right venue for that next)
> 
> Yeah, I don't think the dtc checks infrastructure has the information
> as they run on the live tree after all the parsing is done.

I also started a thread on -compiler. I've replied there already (to
whit: printing the output filename in the warning would be a useful
improvement).

Ian.