64 bit constants are truncated to 32 bit on windows

[rihi]

When decompiling functions on windows, 64 bit constants are truncated to 32 bit.

This is caused by the following code in the backend:

dewolf/decompiler/backend/cexpressiongenerator.py

Lines 83 to 94 in d4b9357

	SIGNED_FORMATS = {
	8: lambda x: c_byte(x).value,
	16: lambda x: c_short(x).value,
	32: lambda x: c_int(x).value,
	64: lambda x: c_long(x).value,
	}
	UNSIGNED_FORMATS = {
	8: lambda x: c_ubyte(x).value,
	16: lambda x: c_ushort(x).value,
	32: lambda x: c_uint(x).value,
	64: lambda x: c_ulong(x).value,

dewolf/decompiler/backend/cexpressiongenerator.py

Lines 296 to 307 in d4b9357

	def _get_integer_literal_value(self, literal: expressions.Constant) -> int:
	"""
	Return the right integer value for the given type, assuming that the
	re-compilation host has the same sizes as the decompilation host.
	"""
	if literal.type.is_signed:
	if handler := self.SIGNED_FORMATS.get(literal.type.size, None):
	return handler(literal.value)
	elif literal.value < 0:
	if handler := self.UNSIGNED_FORMATS.get(literal.type.size, None):
	return handler(literal.value)
	return literal.value

This code uses ctypes to normalize the constant values to their respective bit sizes, but erroneously assumes c_long/c_ulong to always be 64 bit in size. This is however platform dependent, as the c standard only guarantees the following minimum bit sizes: char: 8, short: 16, int: 16,, long: 32

[rihi]

This should be easy to fix as soon as #318 is merged, because that pull request adds a new method normalize_int(v: int, size: int, signed: bool) -> int which could be reused here to replace the ctypes usage.

[rihi]

/cib

[github-actions]

Branch issue-333-64_bit_constants_are_truncated_to_32_bit_on_windows created!