zaks.anna added a comment. > However, the assert here has a reason: we clearly shouldn't be trying to > analyze synthesized bodies as top-level functions.
Yes, seems like we should update r264687 so that we use the available body when analyzing as top level. Another possible issue is that we will use the synthesized body if the function name **starts with** "OSAtomicCompareAndSwap" since we do not match the full function name. If the function body is available, there is a higher chance it is implementing something other than the standard compare and swap. We might want to start matching the full names of the functions are are synthesizing. @alexshap is that a problem for your codebase? Repository: rL LLVM https://reviews.llvm.org/D24792 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
