From mboxrd@z Thu Jan 1 00:00:00 1970 From: Rob Herring Subject: Re: [PATCH v2 14/34] dt-bindings: arm: Convert Amlogic board/soc bindings to json-schema Date: Tue, 4 Dec 2018 08:44:52 -0600 Message-ID: References: <20181203213223.16986-1-robh@kernel.org> <20181203213223.16986-15-robh@kernel.org> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Return-path: In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org To: Neil Armstrong Cc: devicetree@vger.kernel.org, "linux-kernel@vger.kernel.org" , Mark Rutland , Kumar Gala , ARM-SoC Maintainers , Sean Hudson , Frank Rowand , Grant Likely , Kevin Hilman , Carlo Caione , linuxppc-dev , "moderated list:ARM/FREESCALE IMX / MXC ARM ARCHITECTURE" , linux-amlogic@lists.infradead.org List-Id: devicetree@vger.kernel.org On Tue, Dec 4, 2018 at 2:39 AM Neil Armstrong wrote: > > Hi Rob, > > You forgot linux-amlogic in CC... > > On 03/12/2018 22:32, Rob Herring wrote: > > Convert Amlogic SoC bindings to DT schema format using json-schema. > > > > Cc: Carlo Caione > > Cc: Kevin Hilman > > Cc: Mark Rutland > > Cc: devicetree@vger.kernel.org > > Signed-off-by: Rob Herring > > --- [...] > > + - items: > > + - enum: > > + - amlogic,s400 > > + - const: amlogic,a113d > > + - const: amlogic,meson-axg > > + - items: > > + - enum: > > + - amlogic,u200 > > + - const: amlogic,g12a > > but all this feels wrong for me. > > First of all, this yaml description is not human friendly and not intuitive at all, > and secondly with this conversion we loose all the comments about the SoC family relationship > with the compatible strings ! > > I really understand the point to have automated verification, but really it's a pain to read > (I can't imagine newcomers... the actual DT bindings are already hard to read...) and > I feel it will be a real pain to write ! What do you suggest that would be easier? Is it the YAML itself or the json-schema vocabulary? For the former, we could use {} and [] to make things more json style. But I imagine it is the latter. There is some learning curve for json-schema and is certainly a concern I have, but there would be a learning curve for anything. Our choices are use some existing schema language or invent one. All the previous efforts (there's been about 5 since 2013) have been inventing one, and they've not gone far. There will be far few resources available to train people with if we do something custom. > Can't we mix an "humam text" with a "yaml" part on a same document ? we are in 2018 (nearly 2019), > and it should be easy to extract a yaml description from a text document without pain and > keep all the human description, no ? Yes. Please go look at the annotated example in patch 2. > What will be the case for all the bindings with ASCII art to describe the architecture of the > HW ? will you simply drop it to replace it with cold yaml description ? No, you can have literal blocks and/or comments with however much description you like. In fact, there's some notion to write the descriptions in sphinx and extract them to generate documentation, but that's a ways off. That's just the extent of what is possible. Rob