dcoughlin added a comment.
@danielmarjamaki I see what you mean -- thanks for providing the patch.
I don't think this is the right approach. It should be sufficient to reason
about blocks and not individual statements; further some blocks may be
non-empty but not have any statements.
Instead, I think the better approach is make sure the optimized and unoptimized
CFGs match block IDs. It looks like the only situation where there is currently
a mismatch is `do { } while (0);`.
I sketched out an approach on cfe-dev for hoisting block creation (but not edge
creation) in CFGBuilder::VisitDoStmt() during CFG construction to match the
other control flow constructs. That is the approach I would recommend taking.
https://reviews.llvm.org/D24759
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
