Re: [Cbor] [Suit] Code generation from CDDL descriptions

Hi Øyvind,

This is very interesting!

What is the license on this code?

Grüße, Carsten

> On 2020-02-11, at 20:26, Rønningstad, Øyvind <Oyvind.Ronningstad@nordicsemi.no> wrote:
> 
> Hi everyone.
> 
> I’ve created a python script that parses CDDL descriptions and generates C code that decodes and validates CBOR against the description.
> 
> The repo is here: https://github.com/oyvindronningstad/cddl_gen
> 
> I created it with SUIT in mind, and I’ve added the current (not updated for -03 yet) and earlier manifest formats as test cases in https://github.com/oyvindronningstad/cddl_gen/tree/master/tests/cbor_decode.
> 
> If you want to see it in the wild, I opened a PR to MCUboot for using it in their serial recovery functionality here: https://github..com/JuulLabs-OSS/mcuboot/pull/660                                                                                                                                                                                                                                           
> 
>  
> 
> Some numbers:
> 
> The MCUboot PR reduces the code size by 1100+ bytes compared to using TinyCBOR.
> 
> The current SUIT format can be completely decoded with ~7700 bytes of code, where 652 is from the handwritten library, while the rest comes from the generated code. These numbers are taken from the test case “test2_suit”.
> 
> The generated code decodes the data into a struct. The OuterWrapper (containing everything, except SUIT_Command_Sequences) struct is 808 bytes, while a SUIT_Command_Sequence struct is 3684 bytes.
> 
> Remember that this is for the entire format. My thought is that a SUIT implementation will strip the CDDL to match whatever it supports.
> 
> I have not benchmarked runtime.
> 
>  
> 
> Any feedback, ideas, thoughts, questions are appreciated.
> 
> Thanks,
> 
> Øyvind
> 
> _______________________________________________
> Suit mailing list
> Suit@ietf.org
> https://www.ietf.org/mailman/listinfo/suit