On 28 Jun 2010, at 10:46, Saso Kiselkov wrote:
> How can I make sure that sysv.S is being used to generate the necessary
> DWARF data?
This one looks slightly different. Presumably the error is in the the
ffi_closure functions. It would help debugging if, rather than throwing an
exception, you could tell the unwind library to print a backtrace, and then see
where it failed. You can do this by defining this function:
static _Unwind_Reason_Code trace(struct _Unwind_Context *context, void *c)
{
void *ip = (void*)_Unwind_GetIP(context);
Dl_info info;
dladdr(ip, &info);
printf("%p:%s() in %s\n", ip, info.dli_sname, info.dli_fname);
return _URC_CONTINUE_UNWIND;
}
And then calling this:
_Unwind_Backtrace(trace, 0);
You will need to include dlfcn.h and also the unwind.h header for your platform
(if you don't have one, grab the one from libunwind - the functions are in
libgcc_s or similar, so you don't need any extra libraries).
Did you compile libffi with -fexceptions? If not, then it may simply be that
the trampoline function did not get unwind tables generated for it.
David
-- Send from my Jacquard Loom
_______________________________________________
Discuss-gnustep mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/discuss-gnustep
