What A 3.0 Might Look Like

[iximeow]

idle wishlist of things i'd like to break for a 2.0 version of the disassembler.. yaxpeax-x86 was one of the first crates i'd published and has some poor API choices as a result.

• {long,protected,real}_mode need better names
  • years ago i'd had the mistaken impression that you could only run 16-bit code in real mode, 32-bit in protected mode, 64-bit in long mode. the actual hardware-allowed behavior is more subtle and controlled by a few different bits. these really should just be x86_64/x86_32/x86_16 or... something, talking about the bit width of the instruction set.
• multiple-register operands should be structure-style with named fields
  • Operand docs speak for themselves. which RegSpec is the base, or first register in the address? which is second? which RegSpec is the mask register in RegIndexBaseScaleDispMasked? this would be a little more reasonable as RegIndexBase { base: RegSpec, index: RegSpec } and so on. this will change in 2.0
• instruction encoding
  • doesn't necessarily need a breaking change, may happen before or after 2.0, but by 2.0 i'd like to have an idea of if there need to be changes here or in yaxpeax-arch..
• address the open issues re. register names and RegSpec relationships
  • again probably won't need breaking changes. i just want to address these wrinkles before worrying about larger breaking changes.

[iximeow]

• ... real_mode, or whatever name it should be, should not have eip probably... or should have an ip. both? unsure.
  • nothing actually constructs eipin either protected_mode or real_mode. i think this RegSpec constructor and variant will get deleted.

[iximeow]

• fix the mess with fn cs(&mut self) and fn selects_cs(&self).