On Wed, Nov 13, 2013 at 05:47:07PM +0100, Tomasz Figa wrote:
On Wednesday 13 of November 2013 16:43:28 Mark Brown wrote:
The best practice on this one seems to vary somewhat randomly - sometimes it's a requirement, sometimes it isn't.
AFAIR we decided on ARM Mini Summit to mandate this, but maybe my memory is misleading me.
OK, I hadn't heard that - it's good that folks made their mind up.
Is this a constructive thing from a style point of view? We're not allowed to actually do anything useful with the value at runtime so people may as well choose what they like.
This is what ePAPR says and I believe this is reasonable, because looking at device tree sources you don't need to think what kind of hardware a cs42l52 is. The information that it's a cs42l52 is still contained inside compatible string.
Given that a meaningful name was already specified for the handle it's really not going to help anything - it's just going to duplicate that most likely. Given that it can't be actually used for anything it seems better to just let people write whatever they feel like in there (even if it's just a single letter to keep the parser happy) rather than nitpick over their choices.
Really it looks like one of those silly things standards makers do where they start specifying a feature (in this case device classes) but don't define any useful behaviour for it.