Fix AST printing of ascii char literals above 127.

[steven_watanabe]

Summary

Currently literals like '\xff' may be sign extended, resulting in output like '\Uffffffff', which is not valid.

[rsmith]

LGTM

lib/AST/StmtPrinter.cpp

1253–1256

Hmm, this is not correct for multicharacter char literals:

auto n = '\xff\xff\xff\xff';

Here, the value of the literal is -1, but printing it as '\xff' would give a literal of a different type. However, we get that case wrong here in general (printing invalid \U... escapes), so this is no worse than today. Can you add a FIXME?

lib/AST/StmtPrinter.cpp

case '\r':
     break;
   case '\t':
     OS << "'\\t'";
     break;
   case '\v':
     OS << "'\\v'";
     break;
   default:
+    // A character literal might be sign-extended, which
+    // would result in an invalid \U escape sequence.
+    if ((value & ~0xFFu) == ~0xFFu && Node->getKind() == CharacterLiteral::Ascii)
+      value &= 0xFFu;
     if (value < 256 && isPrintable((unsigned char)value))
       OS << "'" << (char)value << "'";
     else if (value < 256)
       OS << "'\\x" << llvm::format("%02x", value) << "'";
     else if (value <= 0xFFFF)
       OS << "'\\u" << llvm::format("%04x", value) << "'";
     else
       OS << "'\\U" << llvm::format("%08x", value) << "'";

test/Misc/ast-print-char-literal.cpp

 // RUN: %clang_cc1 -ast-print -std=c++1z %s -o - | FileCheck %s
 
 char c = u8'1';
 char d = '1';
 char e = U'1';
 char f = L'1';
 char g = u'1';
 
 template <char c = u8'1'>
 void h();
 
 void i() {
   h<u8'2'>();
 }
 
+char j = '\xFF';
+
 // CHECK: char c = u8'1';
 // CHECK-NEXT: char d = '1';
 // CHECK-NEXT: char e = U'1';
 // CHECK-NEXT: char f = L'1';
 // CHECK-NEXT: char g = u'1';
 
 // CHECK: template <char c = u8'1'>
 
 // CHECK: h<u8'2'>();
+// CHECK: char j = '\xff';