[Caml-list] disassembling cmo files

[Caml-list] disassembling cmo files

[Jonathan Roewen]

Hi,

How can I disassemble a .cmo into something useful? I think such a
tool would prove useful to me for understanding exactly how bytecode
works, esp wrt to interactions with the garbage collector.

Jonathan

Re: [Caml-list] disassembling cmo files

[basile]

> How can I disassemble a .cmo into something useful?

IIRC, the ocaml/tools directory (from Ocaml CVS)
contains a dumpobj.ml which should do it.

-- 
Basile Starynkevitch http://starynkevitch.net/Basile/

Re: [Caml-list] disassembling cmo files

[David Baelde]

Hi,

A simple and interesting way to learn about the bytecode is the
-dinstr option of ocamlc and ocaml toplevel. It dumps a human-readable
version of the produced bytecode. Similarly, there is -dlambda.

--
David

Re: [Caml-list] disassembling cmo files

[Jonathan Roewen]

On 12/7/05, David Baelde <david.baelde@gmail.com> wrote:
> Hi,
>
> A simple and interesting way to learn about the bytecode is the
> -dinstr option of ocamlc and ocaml toplevel. It dumps a human-readable
> version of the produced bytecode. Similarly, there is -dlambda.

Yes, I've used that before, but those are pseudo instructions .. not
what actually goes into a bytecode file I believe.

> IIRC, the ocaml/tools directory (from Ocaml CVS)
> contains a dumpobj.ml which should do it.

I tried compiling, but it complains about an undefined field in
debug_event, line 391 (ocaml3.09.0).

Jonathan

Re: [Caml-list] disassembling cmo files

[Oliver Bandel]

On Wed, Dec 07, 2005 at 10:52:25AM +0100, David Baelde wrote:
> Hi,
> 
> A simple and interesting way to learn about the bytecode is the
> -dinstr option of ocamlc and ocaml toplevel. It dumps a human-readable
> version of the produced bytecode. Similarly, there is -dlambda.

Both are very nice features but are a) undocumented and b) work
on the ml-file, not the cmo-files.

To have access to a cmo-file... well... IMHO maybe could be used to
combine Ocaml-stuff with other languages... e.g. dynamically loading
*.cmo's into C-programs...
(...should be possible to do that?!)



Ciao,
   Oliver