================
Comment at: lib/Sema/JumpDiagnostics.cpp:496
@@ -487,1 +495,3 @@
void JumpScopeChecker::VerifyJumps() {
+ std::map<StringRef, MSAsmStmt*> LabelProvider;
+ for (Stmt *J : Jumps) {
----------------
ehsan wrote:
> rnk wrote:
> > At a high level, I feel like we shouldn't need this map. It basically
> > reinvents the label name lookup that sema is already performing for us.
> >
> > What if we changed LabelDecl to hold an MSAsmStmt*? Conceptually, we could
> > make TheStmt be a Stmt* and getStmt() could be 'return
> > dyn_cast_or_null<LabelStmt>(TheStmt);'.
> How do we get the corresponding MSAsmStmt* though? LookupInlineAsmLabel
> doesn't have access to that information.
Right, but it could record them in the ClangAsmParserCallback and forward it on
to ActOnMSAsmStmt. At that point if there are any label decls that weren't
provided by the current asm statement, we can diagnose them.
http://reviews.llvm.org/D5516
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
