David Gibson wrote:
Oh, just remembered the more precise reason why this is the case. By
the time we get to the parser, whitespace and comments have
disappeared entirely, which means they'd be ignored *in the midst* of
node property names under this proposal, which is just not nice.
It's technically possible to recognize whitespace at the parser level
instead, but it makes the grammar hideous.
Ah, right, that'd be unpleasant. I think it could be done with a little
less ugliness than fully recognizing whitespace in the parser (have the
lexer set a flag on relevant tokens if the previous character was
whitespace or a comment, and have the parser rule that glues the
propnodename together check the flag and raise an error if set), but
that's still uglier than the propnodename-after-opening-brace approach.
-Scott
_______________________________________________
devicetree-discuss mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/devicetree-discuss
