[coyotos-dev] BigNum.cxx buglet?

[Jonathan S. Shapiro]

Thomas:

This last change is very disturbing, and we need to understand why it is
necessary. If the problem is that the literals coming out of capidl are
too large, we *really* need to fix it there. The concern is that these
literals get used in assembly code, so we *cannot* resolve the problem
by casting them into submission.

Can you give me an example of a header file line that uses CAPIDL_U32
that is complaining? Just copy that one line of the header file and send
it to the list. Hopefully that will be enough to let me find the problem
in capidl.

sahp

On Mon, 2007-10-01 at 04:50 +0200, Jeroen Visser wrote:
> Thomas,
> 
> I'm about to call it a day but... The images run on my 64-bit Ubuntu
> with the following patches against the latest current:
> 
> diff -r d3dc47fbe797 src/ccs/mkimage/Makefile
> --- a/src/ccs/mkimage/Makefile  Sun Sep 30 20:54:41 2007 -0400
> +++ b/src/ccs/mkimage/Makefile  Sun Sep 30 14:25:14 2007 +0200
> @@ -41,7 +41,7 @@ LIBCOYIMAGE=../libcoyimage/BUILD/libcoyi
> 
>  LIBS+= $(LIBCOYIMAGE)
>  LIBS+= $(LIBSHERPA)
> -LIBS+= -licuuc -lstdc++
> +LIBS+= -lssl -licuuc -lstdc++
> 
>  OBJECTS=\
>         $(BUILDDIR)/mkimage.o \
> diff -r d3dc47fbe797 src/sys/coyotos/capidl.h
> --- a/src/sys/coyotos/capidl.h  Sun Sep 30 20:54:41 2007 -0400
> +++ b/src/sys/coyotos/capidl.h  Mon Oct 01 00:39:16 2007 +0200
> @@ -33,7 +33,7 @@ typedef uint64_t errcode_t;
> 
>  #define CAPIDL_U8(x)  x##u
>  #define CAPIDL_U16(x) x##u
> -#define CAPIDL_U32(x) x##ul
> +#define CAPIDL_U32(x) (uint32_t)((long long)(x##ull & 0xFFFFFFFF))
>  #define CAPIDL_U64(x) x##ull
> 
>  #define CAPIDL_I8(x)  x
-- 
Jonathan S. Shapiro
Managing Director
The EROS Group, LLC
www.coyotos.org, www.eros-os.org