gdb vs. gcc. was: IA32 bootstrap error

[Robert Lipe <robertlipe at usa dot net>]

Richard Henderson wrote:
> On Sun, Dec 30, 2001 at 04:55:51PM -0600, Robert Lipe wrote:
> > My results weren't nearly as positive.  Should I start chasing GDB problems
> > or GCC problems?
> [...]
> > fancy_abort (file=warning: Cannot convert floating-point register
> > value to non-floating-point type.
> 
> Dunno.  The symptom resembles a register numbering mismatch between
> gcc and gdb.  It's possible that the debug info is screwed up enough
> that gdb is actually reading garbage though.

It's GDB.  It spews that warning and generally fails to cope with output
from four different compilers emitting a mixture of dwarf-1 and dwarf-2.
But the other debuggers I tried weren't happy with the current GCC
input, either.  (Of course, they may also share the producer/consumer
codependence I mentioned earlier...)

But the GDB of today behaves substantially worse in all combinations
than gdb 20000204 so we've at least narrowed the search domain down to a
specific 30Mb tree of source.  Joy. :-(

RJL