Re: Fwd: sim bug

[Mike Frysinger <vapier at gentoo dot org>]

On Friday, August 05, 2011 20:03:04 Mike Stump wrote:
> I originally sent this onto the binutils people...  they directed me here.
> 
> I found a nasty sim bug:
> 
> Index: binutils/binutils/sim/common/sim-load.c
> ===================================================================
> --- binutils/binutils/sim/common/sim-load.c     (revision 1609)
> +++ binutils/binutils/sim/common/sim-load.c     (working copy)
> @@ -30,6 +30,7 @@ along with this program.  If not, see <h
> #endif
> #include <time.h>
> 
> +#include "sim-main.h"
> #include "sim-basics.h"
> #include "bfd.h"
> #include "sim-utils.h"
> 
> Now, the reason why this is a bug is obscure.  SIM_ADDR is defaulted to
> 32-bits, and in every other file, sim-main.h can define:
> 
>  #define CORE_ADDR_TYPE unsigned long
> 
> but this one file doesn't include that header, so, in this one file
> SIM_ADDR is 32-bits, and in the rest of the objects it is 64-bit, hence
> breaking 64-bit ports that use sim-main.h to define CORE_ADDR_TYPE.
> 
> If this isn't the right fix, let me know which header these are supposed to
> go in...  Thanks.

i dont think these headers are well (or at all?) documented.  for the code 
path you describe, we'll probably want to put this sim-main.h include into 
sim-basics.h before the gdb includes which setup the defaults (remote-sim.h).
-mike

Attachment: signature.asc
Description: This is a digitally signed message part.