Jon Loeliger wrote at Tuesday, September 20, 2011 7:48 AM: > David Gibson wrote: > > Things get trickier when we want to extend this to macros or > > functions. The only problem I have with your patch at present is that > > I'd prefer not to implement a constant defining syntax, only to have > > it obsoleted by whatever we do for macros or functions. > > Exactly. > > > So, there are basically two approaches to macro or function support. > > > > A) Functions > > B) Macros > > B1) Use cpp itself > > B2) Make our own preprocessor, isomorphic to cpp > > To be thorough, there has been one other macro proposal: Use m4. > Suggesting that, however, has had the entertaining side effect > of causing internet-wide vomiting. > > > Our current impasse is roughly that Jon prefers approach (A), whereas > > I prefer (B1) on balance. (B1) would obsolete your suggested define > > syntax. (A) and (B2) could both potentially subsume it, instead. > > Right.
OK, that all makes sense. However, it leaves me wondering what the next steps are; I was hoping to get a quick and simple constant syntax into dtc that I could use for the pinmux initialization patches I was working on, so the data tables there would be integer-based for efficiency, yet named using constant names instead of seemingly random numbers. It sounds like the dtc issue isn't going to be resolved particularly quickly; should I just go back to using strings in my pinmux patches? But that'd tie us to using strings forever in order for the bindings to remain compatible... -- nvpublic _______________________________________________ devicetree-discuss mailing list [email protected] https://lists.ozlabs.org/listinfo/devicetree-discuss
