Reloc BUG in as

[Mark E.]

> Your coff-i386 result might well be correct, depending upon how the
> linker handles the reloc.  After the final link you should see a call
> to the absolute address 0.  If you see anything else, then something
> went wrong somewhere.

I changed the example to put the call into section "foo" so I could easily 
find it. Here's the result:

Disassembly of section foo:

0000ce00 <a>:
    ce00:	e8 fb 31 ff ff       	call   0 <start-0x10d0>
    ce05:	00 00                	add    %al,(%eax)

Looks good to me. Perhaps the solution isn't in generic code then?

Mark