On 10/23/19 4:45 PM, Rob Herring wrote: > Yes, but would like to understand the use better. > > What problem are you trying to solve that fdtget doesn't work? Perhaps > it is too slow? Loss of type information (the little we have)? Unfortunately I don't know how much I'm allowed to tell about my main use case. I don't think it's applicable outside the company I work for. I thought that someone might find tool like this useful, because while working with dts files, from time to time, I used to find myself lacking convenience of a tool that would print a node(maybe recursively) or property from dts. I don't find grep and vim as convenient, especially if the inclusion graph isn't small. On 10/24/19 5:13 AM, David Gibson wrote: > It would be pretty easy to write a script that runs dtc to compile the > dts, then fdtget to pull out the property you want. > > The difficulty with doing something equivalent for dts directly is > that it essentially needs to include the whole dts parser. That then > needs to be kept in sync with the version in dtc, which is kind of a > pain. On 10/23/19 4:45 PM, Rob Herring wrote: > With C preprocessor defines and include files, you still need to run > the input through something right? Or you want to ignore include > files? If not, you're still going to need the front end of dtc at > least. Then you need either some output format or to tap into dtc's > livetree C structure. The checks code in dtc does the latter. I've > hacked up dtc checks before to dump out things like all compatible > strings when building a tree of dts files. Something I'd like to do is > make that plug-able such that we could have multiple checks modules > including ones in python. This would make it easier to add in the DT > schema checking to dtc. Or maybe another way to approach it is making > the dts->livetree parsing a library. Yes, preprocessing is needed for this application to be useful. At first I've just included dtc.h (and intended to split dtc's front end into a library and refactor when upstreaming) but approach with a script is much simpler and would still work for me. To me either is fine. I can separate out dts->livetree into a lib and make an app that uses it, or write a script that uses dtc and fdtget as You suggest. Let me know if this is compelling enough for inclusion in the repo and if so, which implementation approach You'd find more preferable. -- Łukasz Dobrowolski
