binutils and cross linking alpha to i960

[Stephen Williams <steve@icarus.icarus.com>]

I'm a little pressed to track this down right now (AIIM deadline) so
I though I'd throw this out there and see what happens. My setup:

	Linux/alpha (RedHat 5.0)
	binutils 2.8.1 --target==i960-ucr (a form of i960-coff)
	egcs 1.0 --target=i960-ucr
	gdb 4.16.97 --target=i960-ucr

	I also have all this working on Linux/intel (but my alpha is
	*so* much faster:-) The gdb connection to the target is done
	myself to communicate across the PCI bus to the target board.

I have egcs-1.0 based cross compilers on linux/alpha generating i960-coff
code. 64bit to 32bit, I know, but the generated code is correct as far as
I can tell. Trouble is the debugging information seems to be messed up. It
is not clear to me where the problem lies.

If I use objdump on a Linux/intel host I get a lot of 

Warning: numeric overflow: 1;001000000000000000000000;000777777777777777777777;

messages. I also notice that I may get different results from objdump.
Big surprise, given the messages above.

Also, gdb-4.16.97 has difficulty with matching addresses to line numbers.
It seems to be off by a few lines, although a stack trace gets the right
function names, and gdb can display variables, including nasty C++ classes,
just fine. It works perfectly well except for matching ip to source line
number.

This rather contorted combination worked fine with the same binutils 2.8.1
but gcc 2.7.3 instead of egcs 1.0. I haven't tried binutils 2.9(beta) yet.

If there is a fairly complete description of the stabs stuff that the
compiler generates (and gdb interprets) I would appreciate a pointer.
-- 
Steve Williams                "The woods are lovely, dark and deep.
steve@icarus.com              But I have promises to keep,
steve@picturel.com            and lines to code before I sleep,
http://www.picturel.com       And lines to code before I sleep."