19981025: Build failure making _bb.o (Internal error), underm68k-next-nextstep3

[Melissa O'Neill]

The problems building the 19981019 snapshot on m68k-next-nextstep3 are
still there in the 19981025 snapshot. :-(

   Melissa.

> Trying to build the latest snapshot, I get:
>
> xgcc: Internal compiler error: program cc1 got fatal signal 10
>
> from the stage2 compiler when compiling libgcc2.c -> _bb.o.
>
> Running under gdb, I get:
> Program generated(1): Memory access exception on address 0x8 (protection failure).
> Reading in symbols for /Rock/Build/egcs-snapshot/gcc/calls.c...debug info mismatch between compiler and debugger...debug info mismatch between compiler and debugger...done.
> 0x4f9a4 in emit_library_call (orgfun=0x141606, no_queue=?, outmode=VOIDmode, nargs=?) at /Rock/Build/egcs-snapshot/gcc/calls.c:2316
> 2316          if (mode == BLKmode
>
> But I suspect that this function probably isn't the culprit. Since stage1
> compiled okay, and it's just starting to use the cc1 built by the
> egcs-19981019 compiler (itself built by the host gcc, egcs-19981012),
> the problem could be a codegen bug anywhere in egcs-19981019.
>
> Melissa.
>
> P.S. I don't know whether `the debug info mismatch between compiler and
> debugger' is relevant or not. I suspect not. I'm stuck with GDB 4.7,
> which may not understand some things...