Re: [PATCH v2 1/2] dt-bindings: usb: mt6360-tcpc: Drop interrupt-names

[Conor Dooley <conor@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 3175 bytes --]

On Thu, Jan 25, 2024 at 04:57:33PM +0000, Conor Dooley wrote:
> On Thu, Jan 25, 2024 at 12:41:57PM +0100, AngeloGioacchino Del Regno wrote:
> > Il 25/01/24 11:32, Krzysztof Kozlowski ha scritto:
> > > On 24/01/2024 09:48, AngeloGioacchino Del Regno wrote:
> > > > Il 23/01/24 18:14, Conor Dooley ha scritto:
> > > > > On Mon, Jan 22, 2024 at 11:32:30AM +0100, AngeloGioacchino Del Regno wrote:
> > > > > > Il 19/01/24 17:32, Conor Dooley ha scritto:
> > > > > > > On Fri, Jan 19, 2024 at 10:41:04AM +0100, AngeloGioacchino Del Regno wrote:
> > > > > > > > This IP has only one interrupt, hence interrupt-names is not necessary
> > > > > > > > to have.
> > > > > > > > Since there is no user yet, simply remove interrupt-names.
> > > > > > > 
> > > > > > > I'm a bit confused chief. Patch 2 in this series removes a user of this
> > > > > > > property from a driver, so can you explain how this statement is true?
> > > > > > > 
> > > > > > > Maybe I need to drink a few cans of Monster and revisit this patchset?
> > > > > > > 
> > > > > > 
> > > > > > What I mean with "there is no user" is that there's no device tree with any
> > > > > > mt6360-tcpc node upstream yet, so there is no meaningful ABI breakage.
> > > > > > Different story would be if there was a device tree using this already, in
> > > > > > which case, you can make a required property optional but not remove it.
> > > > > 
> > > > > Not every devicetree lives within the kernel.. If the driver is using
> > > > > it, I'm not inclined to agree that it should be removed.
> > > > 
> > > > I get the point, but as far as I remember, it's not the first time that this
> > > > kind of change is upstreamed.
> > > > 
> > > > I'm fine with keeping things as they are but, since my intention is to actually
> > > > introduce an actual user of this binding upstream, and that actually depends on
> > > > if this change is accepted or not (as I have to know whether I can omit adding
> > > > the interrupt-names property or not)....
> > > > 
> > > > ....may I ask for more feedback/opinions from Rob and/or Krzk?
> > > 
> > > Driver is the user and this is an old binding (released!), thus there
> > > can be out-of-kernel users already.
> > > 
> > > Minor cleanup is not really a reason to affect ABI. You could deprecate
> > > it, though. Driver change is fine.
> > > 
> > 
> > Thanks for the clarification. If USB maintainers want to take the driver part only
> > without me resending this, I'd appreciate that.
> > 
> 
> > The interrupt-names is not a required property in this binding anyway... :-)
> 
> Having -names properties that are not required when the base property is
> always seem so pointless to me, except in cases where they're not
> required for the case where there's one item but required when there are
> more than one. Ultimately they're pointless if not required since they
> can't be relied on. I think dropping it from the driver is required for
> correctness.

Actually, looking at the binding again:

| required:
|   - compatible
|   - interrupts
|   - interrupt-names

It looks like it is a required property after all!

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]