Anastasia added a comment. Just to summarize, it seems there are the following options to proceed, each has some benefits and disadvantages:
1. We can check the canonical type. This gives us possibility to accept the type name aliases, but reduces flexibility to implement this type in a custom way. 2. Identify the type by its name. The issue here is that any type alias via typedef will be treated as a different to ndrange_t (event though it's locally not). One solution to this issue could be to check the typedef chain? At the end I am not sure how critical this is at all to be able to create a typedef of ndrange_t. This approach generally allows vendors to define the type in any suitable way. 3. We can also drop semantical checks for this type completely and let it to a vendor implementation to define and manipulate the type. This offers the most of flexibility but moves more responsibilities to vendor implementations too. I am more in favor of either 1st or 2nd option, but the last one could work too. Repository: rL LLVM https://reviews.llvm.org/D23086 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
