[haiku-development] Re: GSOC valgrind port

  • From: Danny Robson <danny@xxxxxxxxxxx>
  • To: haiku-development@xxxxxxxxxxxxx
  • Date: Wed, 25 Mar 2009 13:21:32 +1100

On Tue, 24 Mar 2009 18:15:40 +0100 CET
"Axel Dörfler" <axeld@xxxxxxxxxxxxxxxx> wrote:
> Valgrind would be pretty cool to have. However, I have no idea how
> hard it would be to port it.

I currently utilise Valgrind in a research capacity (we hack to a
limited extent on the coregrind internals and develop our own tools).

To port coregrind one would ideally be comfortable working with image
loading routines, some level of assembly, address space knowledge,
syscall knowledge, etc (though I could be reading more into this than
there really is).

> I would think it might be very well possible that it does not take 3 
> month to port. Maybe you could have some additional suggestions what 
> you might do if that proves to be correct :-)

IIRC a group from Sun estimated around 6-12 months to port Valgrind to
Solaris. This probably included a very thorough error checking and
porting a good deal of the tools across.

To my mind a better goal would be to get a couple of console apps
running through the null (no instrumentation) tool. Anything after that
would be a bonus. If there's a lot of time available to hack on it,
then you might be able to get it done, but I'd have a quick dive
through the coregrind code beforehand just to reassure yourself.

If the project is taken up I'd be happy to help wherever possible.

- Danny

Other related posts: