[okl4-developer] vserial_serverloop.c generates bad code for (u)int64_t variables in okl4 2.1.

[Gabi Voiculescu]

Hi.

I am using the original okl4 2.1.

I will see what are the changes in the fixed version.

Thank you,
Gabi

--- On Tue, 1/20/09, Josh Matthews <jmatthews at ok-labs.com> wrote:
From: Josh Matthews <jmatthews at ok-labs.com>
Subject: Re: [okl4-developer] vserial_serverloop.c generates bad code for      (u)int64_t variables in okl4 2.1.
To: "Gabi Voiculescu" <boy3dfx2 at yahoo.com>
Cc: developer at okl4.org
Date: Tuesday, January 20, 2009, 1:01 AM

Hi Gabi,

On Mon, January 19, 2009 4:17 pm, Gabi Voiculescu wrote:

> When using okl4 2.1. vtimer on an arm based platform (armv5, armv6) when
> running drivers iguana example I saw an overflow in the displayed time,
> every 6th frame or so.
>
> I traced this back to buld/iguana/include/interfaces/vtimer_serverloop.c
> where for the handle_virtual_timer_request() and
> handle_virtual_timer_current_time() the autogenerated code
>
> The magpie_(un)marshal_int32 is used instead of it's 64bit
counterpart. Is
> there an easy fix for the magpie stuff to make it generate
> magpie_(un)marshal_int64 instead of ..._int32?

Are you using the original release of 2.1, or the later-released
2.1.1-fix.7 (available at http://wiki.ok-labs.com/PreviousReleases)?

There was a known problem with Magpie and 64-bit data types, but it was
fixed in that version - I think this may be related.

> For now I will attempt to manually fix after each new build lines 68-70
> and 99-101 from:
>
>
> magpie_marshal_int32((uint32_t)__return, &mr[4]);  ----------->
> magpie_marshal_int64((uint64_t)__return, &mr[4]);
> /* Return number of return MRs used */
> *((L4_Word_t *)(void *)(&mr[0])) = 1 + 1;   ----------> 
*((L4_Word_t
> *)(void *)(&mr[0])) = 1 + 2;
>

Best regards,
Josh




      
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.okl4.org/pipermail/developer/attachments/20090120/b2b67ca2/attachment.htm