On Fri, Sep 6, 2013 at 1:56 PM, Delesley Hutchins <[email protected]> wrote: > > With respect to the string literals, the parser will currently accept only a > single unresolved identifier as an attribute. Chris wanted to get this > patch out now, under the expectation that we would update the parser, and > then change this back after the parser has been updated. While I do feel > that we should fix variadic enums in the parser, in the longer term we > probably want to transition to a full-fledged typestate system, and make > this a list of Exprs, which must be EnumConstantDecls. In other words, the > states would be declared in the code as an arbitrary enum, and would not be > limited to "consumed" or "unconsumed".
I think the longer term approach is a good idea, but in the short term, there's not that much holding us back from using identifiers instead of string literals. Using string literals is self-inconsistent, and makes the transition to the long-term goal harder. ~Aaron _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
