Datalog Disassembly (github.com)

HN does not seem to have discussed ddisasm enough.

When I discovered it, I was amazed, astonished, stunned, thrilled, and in disbelief.

It turns out that it actually works.

ooh souffle! wonder how they avoid saturation explosion. that's doomed my souffle projects in the past. but obviously they've found a way to partition (maybe by block?)

How does this actually work in practice?

I suppose altering instruction count or size still dooms the executable to become non-functional?

Or do all absolute & relative code & data references all get magically updated?