rengolin added a comment. Bradley, Jojo,
I'm a bit rusty in that part of the code, but that's not really what I was thinking either. The way we started this was to have the TargetParser have *all* parsing. So both check*ArchValid, if necessary, will have to use the TargetParser to parse the strings, and return valid or not. The best way to do this, so far, has been to just parse the Arch and return INVALID, as Bradley said. However, the uncertainty around AArch64/AArch32 vs ARMv8/ARMv8.1 (and possible future v9, v10...) does create an additional layer of complexity, which I, admittedly, haven't grasped fully. In this case, it may be simpler to just rely on the Triple object that already exists in the driver, or just the fact that this is in getAArch64ArchFeaturesFromMarch() and pass an extra flag to parseArch() forcing it to only recognise 32-bit or 64-bit arches as valid. cheers, --renato Repository: rL LLVM http://reviews.llvm.org/D20088 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
