So, I want to use CMU's BAP to do some internal processing for a task that I have been assigned. One of the pertinent parts is transforming assembler representations of CPU instructions into the BAP Intermediate Language, or BIL. It's kind of difficult, because there's only so much documentation that is really anything more than just the MLI interface and the OCaml Doc generated stuff. I have a lot of questions about how to proceed, but before I begin eliciting the problem and all, let me explain about how I got where I am.

You can install BAP through opam, but you don't get the documentation I don't think. So,

git clone github.com/BinaryAnalysisPlatform/bap/

and then just follow the instructions on how to build it, it's not hard at all, I got it going on Ubuntu 14.04. The only thing I ran into was an error on a llvm dependency, which required that I edit the opam file so that I do "--with-llvm-version=3.4" on the configure command line as an option. After that everything ran smoothly.

Once you run bapbuild and make and all that, if you read the Makefile you can see that you can generate all the documentation with:

make doc

which will place the HTML files at:

_build/bap.docdir

Opening up the index file at _build/bap.docdir/index.html, you can see that the documentation starts off with a note about using Bap.Std as everything else is interface files. What confused me is the seeming repetition of the documentation that is generated. It seems that some of the documentation on some of the very same pages is duplicated for certain sections. Why does it do so much duplication?

The next question I have has to do with code organization. It seems that ocamldoc derives the documentation from MLI files, and I know, I know-you can limit or edit the exposure and type definition of your ml files so that they form modules that control access from the outside. But I don't see how to combine usage of the modules that the BAP author has done (I haven't read the BAP code itself in terms of the ml, just searched around, consumed the documentation, the examples directory contents (meager) and read over readbin.ml and bap_mc.ml

Most important regarding proper consumption of the code that I have is are the two avenues that I think would have made it the easiest for me to make use of the library.

First, using the toplevel I tried to construct a BIL set of statements. But the way the code works, you actually have to compose a disassembler that is specific to your architecture (x32/64 and ARM vs Intel or whatever). You then have to construct memory, and from that memory construct an Insn type, which is meant to be the canonical, cross disassembler type representation of an instruction. I can see how module use makes for great reusability of code. Problem is, the type definitions that the toplevel reports (baptop) and those of which are reported in the documentation seem to differ often. TL;DR here, I tried to get as close to the front page mention of how to use module Disasm, which meant Disasm.insn_at_mem function, but I had a hard time navigating the modules to create what I wanted. It seems like each one thing depends on some other portion of the library, and at one point I hit a dead end. The documentation mentions the same functions being exposed copiously, but that's when the type definitions wouldn't match up or something.

Lastly, and ultimately even more confusing is that of bap_mc.ml, which I saw as my second easiest avenue for usage of the BAP library. I saw bap_mc.ml line 55 as my chance;

https://github.com/BinaryAnalysisPlatform/bap/blob/master/src/bap_mc/bap_mc.ml#L55

If I just were to modify it so that it, instead of watering down the string constructed, were to just pipe the insn object to a BIL constructor, and then use the sexp_of_bil transformer, then I could just drop it from there to be printed or converted to string and then printed.

Naturally, I tried with several different module's bil constructor. But most notably I think that the Std bil constructor blew up, so here's what I replaced that line with:

  let s = (* Sexp.to_string @@ Sexp.List (List.rev res) in*)

    sexp_of_bil (Bap.Std.Insn.bil insn) in

But then even that blew up with:

Error: This expression has type ('a, 'b) Insn.t

       but an expression was expected of type insn

Anyway that describes the past two days of jumping into the biggest ocaml project that I've even been into. I'm really excited to become more proficient, but I think there's something I'm missing about OCaml library organization strategies writ large.