[C++] Fix -Wnarrowing and -Woverflow due to signed bitfields on G++ ARM (#6163)

Older versions of GCC (at least on ARM) complain about narrowing conversions and
overflows when setting the reference index (defined as an 11-bit signed integer
bitfield) to -1:

```
error: narrowing conversion of '-1' from 'int' to 'short unsigned int' inside { } [-Wnarrowing]
error: conversion from 'short unsigned int' to 'short unsigned int:11' changes value from '65535' to '2047'
```

This is due to the signedness of integer bitfields (not explicitly signed or
unsigned) being implementation-defined. This addresses this issue by explicitly
defining the signedness of all the bitfields in `flatbuffers::TypeCode` in C++.

1 files changed, 6 insertions, 3 deletions

diff --git a/include/flatbuffers/flatbuffers.h b/include/flatbuffers/flatbuffers.h
index 279afd54..69c8d71b 100644
--- a/include/flatbuffers/flatbuffers.h
+++ b/include/flatbuffers/flatbuffers.h
@@ -2745,10 +2745,13 @@ inline const char * const *ElementaryTypeNames() {
 // clang-format on
 
 // Basic type info cost just 16bits per field!
+// We're explicitly defining the signedness since the signedness of integer
+// bitfields is otherwise implementation-defined and causes warnings on older
+// GCC compilers.
 struct TypeCode {
-  uint16_t base_type : 4;  // ElementaryType
-  uint16_t is_repeating : 1;  // Either vector (in table) or array (in struct)
-  int16_t sequence_ref : 11;  // Index into type_refs below, or -1 for none.
+  unsigned short base_type : 4;  // ElementaryType
+  unsigned short is_repeating : 1;  // Either vector (in table) or array (in struct)
+  signed short sequence_ref : 11;  // Index into type_refs below, or -1 for none.
 };
 
 static_assert(sizeof(TypeCode) == 2, "TypeCode");