sizeof(wchar_t) == 1 on GCC 6.5.0

[deadwood2]

For some reason, sizeof(wchar_t) is 1 on GCC 6.5.0. This is wrong as wchar_t should be 4 bytes long - otherwise it is no different than char. Also builtin define __ WCHAR_TYPE __ gets resolved for some reason to 'char'.

Need to check with GCC 10.5.0. If it was fixed, that it actually breaks binary compatibility with current "wrong" wc#? functions exposed in stdlib.library

[deadwood2]

Further testing shows that this is consistent between i386 and x86_64 GCC 6.5.0 cross & native as well as x86_64 10.5.0 cross and native:

echo "" | ./x86_64-aros-gcc -E - -dM | grep WCHAR

#define __WCHAR_MAX__ 0x7f
#define __WCHAR_MIN__ (-__WCHAR_MAX__ - 1)
#define __WCHAR_WIDTH__ 8
#define __GCC_ATOMIC_WCHAR_T_LOCK_FREE 2
#define __WCHAR_TYPE__ char
#define __SIZEOF_WCHAR_T__ 1

echo "" | ./x86_64-aros-gcc -E - -dM -fshort-wchar | grep WCHAR

Furthermore there is -fshort-wchar option which then changes size to 2 bytes:

#define __WCHAR_MAX__ 0xffff
#define __WCHAR_MIN__ 0
#define __WCHAR_WIDTH__ 16
#define __GCC_ATOMIC_WCHAR_T_LOCK_FREE 2
#define __WCHAR_TYPE__ short unsigned int
#define __SIZEOF_WCHAR_T__ 2

Which in default mode should be downgrading wchar_t from 4 to 2 bytes, but now it's actually upgrading it!

[deadwood2]

Quote of GCC docs:

"Override the underlying type for wchar_t to be short unsigned int instead of the default for the target. This option is useful for building programs to run under WINE.

Warning: the -fshort-wchar switch causes GCC to generate code that is not binary compatible with code generated without that switch. Use it to conform to a non-default application binary interface."

[deadwood2]

AROS patch for GCC 6.5.0 explicitly sets wchar as 8 bit char. Looking at history it has been this way since first version (2019)

gcc/config/aros.h

+#undef WCHAR_TYPE
+#define WCHAR_TYPE "char"
+
+#undef WCHAR_TYPE_SIZE
+#define WCHAR_TYPE_SIZE 8

This needs to be changed to either 'int' or 'unsigned int'. Seems like some systems use 'unsigned' and also python 3.x checks whether wchar_t is unsigned to decide if it is "usable"