Artur K.
Artur K.
Agreed, call graph should be a separate thing, and that tree should be a simple linear list.
I think we should adopt the llvm's way of modeling things and switch to thinking/converting various parts of decompilation as Passes.. We'll have to see if we can just 'borrow'...
For now I moved/converted it into a first prototype of FunctionPass. Will leave this open so that we'll remember to actually use it at some point :)
Agreed. although I think we should consider a few things first: 1. Some passes might fail - we need to handle that gracefully ( LLVM doesn't have that kind of...
Well, for now it's just a wish, since milestone would mean that we're actually going to implement it :smile:
The problem is not with the description of the AMD64, but with the cpu specification framework boomerang uses. It's a pretty obscure machine code parser generator using New Jersey Machine...
Thank You for the suggestion, Right now I'm using valgrind/clang scan-build/sanitizer for finding non-obvious bugs, but I've totally forgotten about Coverity :) I'll register boomerang on their site, and should...
The boomerang project has been analyzed and is available on Coverity's website https://scan.coverity.com/projects/2534
I'll tackle your comment in FILO order if You don't mind :) > When you suggest pattern matching, what is the mechanism for doing so other than hard-coding into the...
ya, might as well