This approach will also introduce false negatives.
A better approach would be to check if the null statement is the result of folding an if constexpr.
The current AST API does not expose this information.
A better fix might require AST patches on the clang side.
The fix is proposed here: https://reviews.llvm.org/D46234
Richard suggested to replace null stmts with nulls, but I have no time to deal with all the fallouts for now. If someone is willing to pick that up feel free to, but until then it might be a good idea to land this quick fix.