On Wed, Aug 28, 2019 at 03:01:15PM -0500, Rob Herring wrote:
> On Tue, Aug 27, 2019 at 2:56 PM Mark Brown <broonie@kernel.org> wrote:
> > On Tue, Aug 27, 2019 at 10:50:05AM -0500, Rob Herring wrote:

> > Well, you have been pushing people to change over to using
> > dt-bindings: so I guess you do care :(

> Well, yes. In the absence of any sort of pattern, I have pushed for
> some consistency. And to get rid of subjects like this:

> Documentation/devicetree/bindings: Add the DT binding documentation for foo-bar

> If subsystems are consistent with their own standard as you are, then
> as a maintainer I don't really care. My point was in regard to what
> submitters need to know and follow.

I agree that things should be more consistent.

> > It really does cause me
> > to miss stuff, especially where people don't even include the
> > subsystem name in the header.  I get quite a lot of CCs for

> I can't imagine filtering on subjects will ever be that reliable
> unless we add subject prefixes to MAINTAINERS and have checkpatch
> check commits against those. Filtering on the diffstat is the only
> thing that's kept things to a sane list for me (MAINTAINERS for DT
> used to tag of_* functions which just meant getting copied on *every*
> driver). This is done on the patchwork server side for me, but I
> imagine one could do it on the client side too.

Part of the problem for me here is that stuff that's flagged as just a
binding has a very high chance of being misdirected, I'm unlikely to
have much input unless it's for a driver or subsystem I maintain and I
get a lot of bindings docs for things like other bits of MFDs that have
a regulator on them or where there was some interesting interaction with
one of my subsystems that hasn't yet filtered out of get_maintainers'
view.

The other trick here is that sometimes I am actually being asked about
the thing that I'm getting CCed on so I don't want to actually filter
stuff out of my inbox, it's more of a scoring system thing with lots of
guessing going on.  I say filtering but it's more a strong signal than
strictly a filter.