Fix the rest of PR2279:
a) correct rejection of ',' in pp expressions.
b) the precedence of ',' was wrong w.r.t. ?:.
Thanks again to Neil for finding these and providing testcases.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@50625 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Lex/PPExpressions.cpp b/lib/Lex/PPExpressions.cpp
index 6343a3c..3c0bf5c 100644
--- a/lib/Lex/PPExpressions.cpp
+++ b/lib/Lex/PPExpressions.cpp
@@ -328,9 +328,9 @@
case tok::pipe: return 7;
case tok::ampamp: return 6;
case tok::pipepipe: return 5;
- case tok::question: return 4;
- case tok::colon: return 3;
- case tok::comma: return 2;
+ case tok::comma: return 4;
+ case tok::question: return 3;
+ case tok::colon: return 2;
case tok::r_paren: return 0; // Lowest priority, end of expr.
case tok::eom: return 0; // Lowest priority, end of macro.
}
@@ -543,7 +543,10 @@
Res.setIsUnsigned(false); // C99 6.5.14p3, result is always int (signed)
break;
case tok::comma:
- PP.Diag(OpToken, diag::ext_pp_comma_expr);
+ // Comma is invalid in pp expressions in c89/c++ mode, but is valid in C99
+ // if not being evaluated.
+ if (!PP.getLangOptions().C99 || ValueLive)
+ PP.Diag(OpToken, diag::ext_pp_comma_expr);
Res = RHS; // LHS = LHS,RHS -> RHS.
break;
case tok::question: {