[PATCH] "Fancy" ARBs

Richard Smith richard at metafoo.co.uk
Tue May 12 13:27:59 PDT 2015


First off: I'm not happy about having this extension in upstream clang until we have a strong indication that this is the direction that will end up standardized. For now, I'd recommend maintaining this as a clang fork on github or similar. With that said, I'm going to review this as if for upstream clang.

I don't like that you create two variables here. We try to maintain as much source fidelity as we can, and I think we can do better here -- how about instead introducing a new form of expression that represents "stack-allocate a certain amount of memory" (with a subexpression for the initialization, if you allow these variables to have an initializer), much like MaterializeTemporaryExpr does for a SD_Automatic temporary, but parameterized by an expression specifying the array size? Then create just a single expression of the std::arb type, initialized by that expression.

You should also introduce a Type subclass to represent type sugar for the ARB type, so that we can model that int[n] desugars to std::arb<int> but should be pretty-printed as an array type.


http://reviews.llvm.org/D9714

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/






More information about the cfe-commits mailing list