Differences between format version 1 and version 0?

[matthilde]

Hey! I've been looking into your project's source because it is interesting.

I looked into v0.1.2 and v0.2.1 and the format version changes I can't see the difference.

Is there a subtility behind that that I haven't checked? I looked into the headers and they're pretty much identical (magic + version + 16-bit length). Are there new opcode pairs? If so that could mean that a version 1 parser can still extract version 0 data.

[woodruffw]

Hi @matthilde, thanks for poking at the codebase. I'm happy to answer any questions you have about it.

Are there new opcode pairs?

Yeah, version 1 adds the TEST opcodes as encoding pairs. It's been a few years but I believe that was the only change; the length encoding itself wasn't changed between the two versions.

If so that could mean that a version 1 parser can still extract version 0 data.

Could you say more? I don't think this is true, since a version 1 decoder would interpret any TEST opcodes in the executable as data, which would be incorrect for a version 0 message (since the TEST opcodes would be in whatever form the compiler selected). But I might be misunderstanding what you mean.

[matthilde]

Oh I see, I haven't noticed this.
Maybe I have been lucky and the outdated executable did not have TEST opcodes. Yea it probably wouldn't decode then, unless one is lucky to not have the new opcodes in their executable.

I might try to make a pull request when I have some time to make the extract command backwards compatible. Thank you for your answer!

[woodruffw]

Maybe I have been lucky and the outdated executable did not have TEST opcodes. Yea it probably wouldn't decode then, unless one is lucky to not have the new opcodes in their executable.

Yeah, it wouldn't surprise me -- I think TEST isn't used very much on default x86 targets.