[libc] Fix printf long double inf, bitcast in msan (#70067)

These bugs were found with the new printf long double fuzzing. The long
double inf vs nan bug was introduced when we changed to
get_explicit_exponent. The bitcast msan issue hadn't come up previously,
but isn't a real bug, just a poisoning confusion.

Author: michaelrj-google

libc/src/__support/CPP/CMakeLists.txt

@@ -15,7 +15,10 @@ add_header_library(
   HDRS
     bit.h
   DEPENDS
-    libc.src.__support.macros.properties.compiler
+    .type_traits
+    libc.src.__support.macros.attributes
+    libc.src.__support.macros.config
+    libc.src.__support.macros.sanitizer
 )
 
 add_header_library(

libc/src/__support/CPP/bit.h

@@ -12,6 +12,7 @@
 #include "src/__support/CPP/type_traits.h"
 #include "src/__support/macros/attributes.h"
 #include "src/__support/macros/config.h" // LIBC_HAS_BUILTIN
+#include "src/__support/macros/sanitizer.h"
 
 namespace LIBC_NAMESPACE::cpp {
 
@@ -31,6 +32,7 @@ LIBC_INLINE constexpr To bit_cast(const From &from) {
   static_assert(cpp::is_trivially_copyable<To>::value &&
                     cpp::is_trivially_copyable<From>::value,
                 "Cannot bit-cast instances of non-trivially copyable classes.");
+  MSAN_UNPOISON(&from, sizeof(From));
 #if defined(LLVM_LIBC_HAS_BUILTIN_BIT_CAST)
   return __builtin_bit_cast(To, from);
 #else

libc/src/stdio/printf_core/float_inf_nan_converter.h

@@ -34,13 +34,13 @@ LIBC_INLINE int convert_inf_nan(Writer *writer, const FormatSection &to_conv) {
     fputil::FPBits<long double>::UIntType float_raw = to_conv.conv_val_raw;
     fputil::FPBits<long double> float_bits(float_raw);
     is_negative = float_bits.get_sign();
-    mantissa = float_bits.get_explicit_mantissa();
+    mantissa = float_bits.get_mantissa();
   } else {
     fputil::FPBits<double>::UIntType float_raw =
         static_cast<fputil::FPBits<double>::UIntType>(to_conv.conv_val_raw);
     fputil::FPBits<double> float_bits(float_raw);
     is_negative = float_bits.get_sign();
-    mantissa = float_bits.get_explicit_mantissa();
+    mantissa = float_bits.get_mantissa();
   }
 
   char sign_char = 0;

libc/test/src/stdio/sprintf_test.cpp

@@ -572,7 +572,6 @@ TEST_F(LlvmLibcSPrintfTest, FloatHexExpConv) {
   ForceRoundingMode r(RoundingMode::Nearest);
   double inf = LIBC_NAMESPACE::fputil::FPBits<double>::inf().get_val();
   double nan = LIBC_NAMESPACE::fputil::FPBits<double>::build_nan(1);
-
   written = LIBC_NAMESPACE::sprintf(buff, "%a", 1.0);
   ASSERT_STREQ_LEN(written, buff, "0x1p+0");
 
@@ -937,6 +936,10 @@ TEST_F(LlvmLibcSPrintfTest, FloatDecimalConv) {
   ForceRoundingMode r(RoundingMode::Nearest);
   double inf = LIBC_NAMESPACE::fputil::FPBits<double>::inf().get_val();
   double nan = LIBC_NAMESPACE::fputil::FPBits<double>::build_nan(1);
+  long double ld_inf =
+      LIBC_NAMESPACE::fputil::FPBits<long double>::inf().get_val();
+  long double ld_nan =
+      LIBC_NAMESPACE::fputil::FPBits<long double>::build_nan(1);
 
   char big_buff[10000]; // Used for long doubles and other extremely wide
                         // numbers.
@@ -996,6 +999,18 @@ TEST_F(LlvmLibcSPrintfTest, FloatDecimalConv) {
   written = LIBC_NAMESPACE::sprintf(buff, "%F", -nan);
   ASSERT_STREQ_LEN(written, buff, "-NAN");
 
+  written = LIBC_NAMESPACE::sprintf(buff, "%Lf", ld_inf);
+  ASSERT_STREQ_LEN(written, buff, "inf");
+
+  written = LIBC_NAMESPACE::sprintf(buff, "%LF", -ld_inf);
+  ASSERT_STREQ_LEN(written, buff, "-INF");
+
+  written = LIBC_NAMESPACE::sprintf(buff, "%Lf", ld_nan);
+  ASSERT_STREQ_LEN(written, buff, "nan");
+
+  written = LIBC_NAMESPACE::sprintf(buff, "%LF", -ld_nan);
+  ASSERT_STREQ_LEN(written, buff, "-NAN");
+
   // Length Modifier Tests.
 
   // TODO(michaelrj): Add tests for LONG_DOUBLE_IS_DOUBLE and 128 bit long

utils/bazel/llvm-project-overlay/libc/BUILD.bazel

@@ -170,6 +170,7 @@ libc_support_library(
         ":__support_cpp_type_traits",
         ":__support_macros_attributes",
         ":__support_macros_config",
+        ":__support_macros_sanitizer",
         ":libc_root",
     ],
 )