disassembler

[0x100001060]> pia 4
str z21, [x2, 0x4a, mul vl]
???
???
???
[0x100001060]> pia 8
str z21, [x2, 0x4a, mul vl]
???
???
???
invalid
???
???
???
[0x100001060]> p8 8
554889e541574156
[0x100001060]> e asm.arch;e asm.bits
arm
64
[0x100001060]>

Tl;dr: Use the workaround from https://github.com/pwndbg/pwndbg/pull/322/files only for broken gdb versions

It's currently always using mov rAX,[moffs] but if it fits in an i32, it's shorter to use mov rAX,[m] with a modrm byte.

The generated mov() methods would need to be updated. This is what the generated C# code looks like:

			if (dst.Value == Register.RAX && Bitness == 64 && src.IsDisplacementOnly) {
				code = Code.Mov_RAX_moffs64;

An extra check would need to be adde

pdgsd does not seem to process PUSH EBP very well, the p-code semantics show up as

    (unique,0x12f0,4) = COPY EBP
    ESP = INT_SUB ESP, 0x4
    (unique,0x12f0,4) = STORE ram[ESP]

which doesn't make sense. Last line should be STORE ram[ESP] = (unique,0x12f0,4)

the raw p-code will show you:

STORE (const,0x55fef11636d0,8) (register,ESP,4) (unique,0x1b50,4)

and the

We should create a corpus of tests to verify that CFG recovery is accurate and complete, and there are no regressions.