[PATCH] D15139: [IR] Reformulate LLVM's EH funclet IR

[Andy Kaylor via llvm-commits]

andrew.w.kaylor added inline comments.

================
Comment at: lib/AsmParser/LLParser.cpp:5189-5190
@@ -5273,5 +5188,4 @@
   BasicBlock *UnwindBB = nullptr;
-  if (Lex.getKind() == lltok::kw_to) {
-    Lex.Lex();
-    if (ParseToken(lltok::kw_caller, "expected 'caller' in terminatepad"))
+  if (EatIfPresent(lltok::kw_to)) {
+    if (ParseToken(lltok::kw_caller, "expected 'caller' in catchswitch"))
       return true;
----------------
rnk wrote:
> I like staying consistent with invoke, which does "unwind label %foo".
I'm happy with "unwind label %foo" as the preferred and documented syntax.  I was only suggesting that the parser tolerate the presence of a 'to' token here.  I suppose that's too imprecise for the philosophy of the IR.  I can live with that.  The number of people who would ever care is likely to be less than five and may only be one.


http://reviews.llvm.org/D15139