Description
Due to the way the code for m68k disassembler is written it outputs broken numbers when disassembling float immediate numbers, e.g. following output is created:
fmul.s #0.00000E-322, fp5
instead of
fmul.s #0.69314E0, fp5
This is due to the union (include/capstone/m68k.h):
/// Instruction operand
typedef struct cs_m68k_op {
union {
uint64_t imm; ///< immediate value for IMM operand
double dimm; ///< double imm
float simm; ///< float imm
m68k_reg reg; ///< register value for REG operand
cs_m68k_op_reg_pair reg_pair; ///< register pair in one operand
};
...
The layout of the union members on both little and big endian is as follows:
x x x x x x x x uint64_t imm
x x x x x x x x double dimm;
x x x x . . . . float simm;
Therefore, storing little endian integer, decoded from instruction stream, into 64-bit field imm allows one to read 32-bit float from the very same position. There, the least significant bytes come first. On big endian however, storing 32-bit number in 64-bit variable imm will zero out the top 32 bits, which are the location where float is read from.
I have a fix ready, but it assumes existence of __BYTE_ORDER__ define (supported by gcc) as well as __ORDER_LITTLE_ENDIAN__ or __ORDER_BIG_ENDIAN__.