Fix printf specifier handling: invalid specifier should not be marked as "consuming data arguments"

[bruno]

LGTM

clang/include/clang/Analysis/Analyses/FormatString.h

public:
 
   bool consumesDataArgument() const {
     switch (kind) {
       case PrintErrno:
         assert(IsPrintf);
         return false;
       case PercentArg:
         return false;
+      case InvalidSpecifier:
+        return false;
       default:
         return true;
     }
   }
 
   Kind getKind() const { return kind; }
   void setKind(Kind k) { kind = k; }
   unsigned getLength() const {

clang/test/CodeGen/builtins.c

void test_builtin_os_log_precision_width(void *buf, const char *data,
   // CHECK: [[ARG3:%.*]] = getelementptr i8, i8* [[BUF2]], i64 16
   // CHECK: [[ARG3_PTR:%.*]] = bitcast i8* [[ARG3]] to i8**
   // CHECK: [[DATA2:%.*]] = load i8*, i8** [[DATA_ADDR]]
   // CHECK: store i8* [[DATA2]], i8** [[ARG3_PTR]]
 
   __builtin_os_log_format(buf, "Hello %*.*s World", precision, width, data);
 }
 
+// CHECK-LABEL: define void @test_builtin_os_log_invalid
+// CHECK: (i8* [[BUF:%.*]], i32 [[DATA:%.*]])
+void test_builtin_os_log_invalid(void *buf, int data) {
+  volatile int len;
+  // CHECK: store i8* [[BUF]], i8** [[BUF_ADDR:%.*]], align 8
+  // CHECK: store i32 [[DATA]], i32* [[DATA_ADDR:%.*]]
+
+  // CHECK: store volatile i32 8,
+  len = __builtin_os_log_format_buffer_size("invalid specifier %: %d even a trailing one%", data);
+
+  // CHECK: [[BUF2:%.*]] = load i8*, i8** [[BUF_ADDR]]
+  // CHECK: [[SUMMARY:%.*]] = getelementptr i8, i8* [[BUF2]], i64 0
+  // CHECK: store i8 0, i8* [[SUMMARY]]
+  // CHECK: [[NUM_ARGS:%.*]] = getelementptr i8, i8* [[BUF2]], i64 1
+  // CHECK: store i8 1, i8* [[NUM_ARGS]]
+
+  // CHECK: [[ARG1_DESC:%.*]] = getelementptr i8, i8* [[BUF2]], i64 2
+  // CHECK: store i8 0, i8* [[ARG1_DESC]]
+  // CHECK: [[ARG1_SIZE:%.*]] = getelementptr i8, i8* [[BUF2]], i64 3
+  // CHECK: store i8 4, i8* [[ARG1_SIZE]]
+  // CHECK: [[ARG1:%.*]] = getelementptr i8, i8* [[BUF2]], i64 4
+  // CHECK: [[ARG1_INT:%.*]] = bitcast i8* [[ARG1]] to i32*
+  // CHECK: [[ARG1_VAL:%.*]] = load i32, i32* [[DATA_ADDR]]
+  // CHECK: store i32 [[ARG1_VAL]], i32* [[ARG1_INT]]
+
+  __builtin_os_log_format(buf, "invalid specifier %: %d even a trailing one%", data);
+}
+
 // CHECK-LABEL: define void @test_builtin_os_log_percent
 // CHECK: (i8* [[BUF:%.*]], i8* [[DATA1:%.*]], i8* [[DATA2:%.*]])
 // Check that the %% which does not consume any argument is correctly handled
 void test_builtin_os_log_percent(void *buf, const char *data1, const char *data2) {
   volatile int len;
   // CHECK: store i8* [[BUF]], i8** [[BUF_ADDR:%.*]], align 8
   // CHECK: store i8* [[DATA1]], i8** [[DATA1_ADDR:%.*]], align 8
   // CHECK: store i8* [[DATA2]], i8** [[DATA2_ADDR:%.*]], align 8
void test_builtin_os_log_percent(void *buf, const char *data1, const char *data2) {
   // CHECK: [[ARG2:%.*]] = getelementptr i8, i8* [[BUF2]], i64 14
   // CHECK: [[ARG2_PTR:%.*]] = bitcast i8* [[ARG2]] to i8**
   // CHECK: [[DATA2:%.*]] = load i8*, i8** [[DATA2_ADDR]]
   // CHECK: store i8* [[DATA2]], i8** [[ARG2_PTR]]
   __builtin_os_log_format(buf, "%s %% %s", data1, data2);
 }
 
 #endif
 No newline at end of file

clang/test/Sema/format-strings.c

void check_writeback_specifier()
   printf("%qn", (int*)0); // expected-warning{{format specifies type 'long long *' but the argument has type 'int *'}}
 
   printf("%Ln", 0); // expected-warning{{length modifier 'L' results in undefined behavior or no effect with 'n' conversion specifier}}
   // expected-note@-1{{did you mean to use 'll'?}}
 }
 
 void check_invalid_specifier(FILE* fp, char *buf)
 {
-  printf("%s%lb%d","unix",10,20); // expected-warning {{invalid conversion specifier 'b'}}
+  printf("%s%lb%d","unix",10,20); // expected-warning {{invalid conversion specifier 'b'}} expected-warning {{data argument not used by format string}}
   fprintf(fp,"%%%l"); // expected-warning {{incomplete format specifier}}
   sprintf(buf,"%%%%%ld%d%d", 1, 2, 3); // expected-warning{{format specifies type 'long' but the argument has type 'int'}}
-  snprintf(buf, 2, "%%%%%ld%;%d", 1, 2, 3); // expected-warning{{format specifies type 'long' but the argument has type 'int'}} expected-warning {{invalid conversion specifier ';'}}
+  snprintf(buf, 2, "%%%%%ld%;%d", 1, 2, 3); // expected-warning{{format specifies type 'long' but the argument has type 'int'}} expected-warning {{invalid conversion specifier ';'}} expected-warning {{data argument not used by format string}}
 }
 
 void check_null_char_string(char* b)
 {
   printf("\0this is bogus%d",1); // expected-warning {{string contains '\0'}}
   snprintf(b,10,"%%%%%d\0%d",1,2); // expected-warning {{string contains '\0'}}
   printf("%\0d",1); // expected-warning {{string contains '\0'}}
 }
void test10(int x, float f, int i, long long lli) {
   printf("xs\0"); // expected-warning{{format string contains '\0' within the string body}}
   printf("%*d\n"); // expected-warning{{'*' specified field width is missing a matching 'int' argument}}
   printf("%*.*d\n", x); // expected-warning{{'.*' specified field precision is missing a matching 'int' argument}}
   printf("%*d\n", f, x); // expected-warning{{field width should have type 'int', but argument has type 'double'}}
   printf("%*.*d\n", x, f, x); // expected-warning{{field precision should have type 'int', but argument has type 'double'}}
   printf("%**\n"); // expected-warning{{invalid conversion specifier '*'}}
   printf("%d%d\n", x); // expected-warning{{more '%' conversions than data arguments}}
   printf("%d\n", x, x); // expected-warning{{data argument not used by format string}}
-  printf("%W%d\n", x, x); // expected-warning{{invalid conversion specifier 'W'}}
+  printf("%W%d\n", x, x); // expected-warning{{invalid conversion specifier 'W'}}  expected-warning {{data argument not used by format string}}
   printf("%"); // expected-warning{{incomplete format specifier}}
   printf("%.d", x); // no-warning
   printf("%.", x);  // expected-warning{{incomplete format specifier}}
   printf("%f", 4); // expected-warning{{format specifies type 'double' but the argument has type 'int'}}
   printf("%qd", lli); // no-warning
   printf("%qd", x); // expected-warning{{format specifies type 'long long' but the argument has type 'int'}}
   printf("%qp", (void *)0); // expected-warning{{length modifier 'q' results in undefined behavior or no effect with 'p' conversion specifier}}
   printf("hhX %hhX", (unsigned char)10); // no-warning
   printf("llX %llX", (long long) 10); // no-warning
   // This is fine, because there is an implicit conversion to an int.
   printf("%d", (unsigned char) 10); // no-warning
   printf("%d", (long long) 10); // expected-warning{{format specifies type 'int' but the argument has type 'long long'}}
   printf("%Lf\n", (long double) 1.0); // no-warning
   printf("%f\n", (long double) 1.0); // expected-warning{{format specifies type 'double' but the argument has type 'long double'}}
   // The man page says that a zero precision is okay.
   printf("%.0Lf", (long double) 1.0); // no-warning
   printf("%c\n", "x"); // expected-warning{{format specifies type 'int' but the argument has type 'char *'}}
   printf("%c\n", 1.23); // expected-warning{{format specifies type 'int' but the argument has type 'double'}}
-  printf("Format %d, is %! %f", 1, 2, 4.4); // expected-warning{{invalid conversion specifier '!'}}
+  printf("Format %d, is %! %f", 1, 4.4); // expected-warning{{invalid conversion specifier '!'}}
 }
 
 typedef unsigned char uint8_t;
 
 void should_understand_small_integers() {
   printf("%hhu", (short) 10); // expected-warning{{format specifies type 'unsigned char' but the argument has type 'short'}}
   printf("%hu\n", (unsigned char) 1); // expected-warning{{format specifies type 'unsigned short' but the argument has type 'unsigned char'}}
   printf("%hu\n", (uint8_t)1); // expected-warning{{format specifies type 'unsigned short' but the argument has type 'uint8_t'}}

clang/test/SemaObjC/format-strings-objc.m

 int printf(const char * restrict, ...) ;
 
 //===----------------------------------------------------------------------===//
 // Test cases.
 //===----------------------------------------------------------------------===//
 
 void check_nslog(unsigned k) {
   NSLog(@"%d%%", k); // no-warning
-  NSLog(@"%s%lb%d", "unix", 10,20); // expected-warning {{invalid conversion specifier 'b'}}
+  NSLog(@"%s%lb%d", "unix", 10, 20); // expected-warning {{invalid conversion specifier 'b'}} expected-warning {{data argument not used by format string}}
 }
 
 // Check type validation
 extern void NSLog2(int format, ...) __attribute__((format(__NSString__, 1, 2))); // expected-error {{format argument not an NSString}}
 extern void CFStringCreateWithFormat2(int *format, ...) __attribute__((format(CFString, 1, 2))); // expected-error {{format argument not a CFString}}
 
 // <rdar://problem/7068334> - Catch use of long long with int arguments.
 void rdar_7068334() {