Devicetree testing#2845
Conversation
|
What do you want to use device trees for? I generally find them hard to understand. |
|
I think the main thing I like about them compared to using toml, xml, json etc is that you can easily reference other nodes. For example One alternative would be to create some nice data structure in python? Then we don't need to parse it. The reason I'd like to use a system like this is to avoid having too many places to change when adding a new port/soc/board. |
| compatible = "arm,dtcm"; | ||
| reg = < 0x20000000 0x10000 >; | ||
| }; | ||
| pins { |
There was a problem hiding this comment.
I like this kind of syntax for defining the pins...what do you think?
There was a problem hiding this comment.
I don't like DTS's use of symbols for many things. I'm totally ok with how we do it now because it's much clearer to me.
I think that makes it harder to understand.
I think we'll end up with TOML in the long run but it's a very low priority. What we have now is well documented and well understood.
I don't think we have too many places now. New ports are rare and new boards are well documented and well defined. My main interest with the build system revamp I prototyped is to reduce the duplication of logic between ports and move it to a clearer system. I don't think that using device trees for configuration is clearer. So, I'm going to close this. Thanks for the proposal! |
I'd like to investigate the possibility of using device trees, I've copied in some code from the Zephyr project to test with. So far it's working to read in the flash-size of nucleo_f746zg into mpconfigboard.h.
I believe we can generate linker_files from this, and also replace mpconfigboard.h, mpconfigport.h, and maybe the .mk files as well. Ideally we'd be able to store everything that configures the board in the devicetree files.
This is very much not complete, I'd just like some feedback.
@hierophect @tannewt @ladyada @dhalbert @jepler