M68K: Floating point loads and stores generate the same op_str

[michalsc]

I'm working on Emu68 project and since yesterday use capstone to print the translation results to the console. I have noticed that in case of FPU loads and stores exactly the same op_str is generated. This is not the case in case of integer family of instructions.

Steps to reproduce. The following code:

        fmove.d	fp0, -(a7)
	fmove.d (a7)+, fp1
	rts

Generates a binary stream:

f2 27 74 00 f2 1f 54 80 4e 75

Disassembled or checked in other tools (ODA for example) one may notice that the code is correct. The disassembly from capstone is however broken in op_str:

[JIT] 36c00000: fmove.d -(a7), fp0           -> 00000000: str     d8, [x29, #-8]!
[JIT] 36c00004: fmove.d (a7)+, fp1           -> 00000004: ldr     d9, [x29], #8

Since pre-decrement and post-increment modes are allowed for both loads and stores on m68k it cannot be distinguished based on op_str what kind of instruction it actually is. The current op_str suggest that both fmoved instructions are loads from memory.

[michalsc]

Fixed in PR #1709