WIP on a small DMA abstraction

[Louis Opter]

Joa, here is my patch queue:

https://www.kalessin.fr/projs/hg/epitech/mq/rathaxes/

To reproduce our bug you only need to push the first three patches:

hg qgoto use_the_dma

Then, the part I'm currently working on is rtx_socket_skbuff_map in
socket.blt

I placed the fixes we made in the compiler in the first patch.

And to give everybody else a little bit of context:

I'm currently working on adding a small DMA abstraction so I can
refactor e1000::set_up_device which currently mixes platform and device
dependent code. Joa is helping me with a few blocker on the compiler to
make that happen.

--
Louis Opter

[Louis Opter]

On Jul 21, 2013, at 2:01 AM, Louis Opter <kale...@kalessin.fr> wrote:

> Joa, here is my patch queue:
>
> https://www.kalessin.fr/projs/hg/epitech/mq/rathaxes/
>
> To reproduce our bug you only need to push the first three patches:
>
> hg qgoto use_the_dma

For the record, the bug was that I was I calling a chunk ::CALL from a rvalue
(which, from my understanding, turns into an anonymous pointcut).

But, this wasn't supported, since you're only supposed to do that from a block
(a chunk –even a ::CALL chunk– is supposed to be a block of code anyway).
But in my use-case the ::CALL chunk translated into a simple function call.

I was actually already doing this trick with method calls, but the problem didn't
manifest because method calls are broken in the first place.

Anyway, Joa hacked support for this operation into the compiler, we still have
a few problems with things not being correctly expanded in the call, but he is
working on it.

--
Louis Opter