Path: utzoo!telly!ddsw1!lll-winken!killer!osu-cis!tut.cis.ohio-state.edu!csl.sony.jp!diamond
From: diamond@csl.sony.jp (Norman Diamond)
Newsgroups: gnu.gcc.bug
Subject: (none)
Message-ID: <8901070927.AA02051@diamond.csl.sony.junet>
Date: 7 Jan 89 09:26:53 GMT
Sender: root@tut.cis.ohio-state.edu
Distribution: gnu
Organization: GNUs Not Usenet
Lines: 62

To:    GNU gcc bug department
       bug-gcc@prep.ai.mit.edu

From:  Norman Diamond, Sony Computer Science Laboratory
       diamond@csl.sony.jp
       diamond%csl.sony.jp@relay.cs.net

Date:  1989.01.07


Installation instruction number 7 is as follows:
   make CC=stage1/gcc CFLAGS="-g -O -Bstage1/"

One of its generated commands is as follows:
   stage1/gcc -g -O -Bstage1/ -c parse.tab.c

The result is:
   stage1/gcc: Program cc1 got fatal signal 10.
(bus error) with a core dump.

When I try:
   dbx stage1/cc1 core
dbx starts reading the symbol table, gets a segmentation error, and  ...
er ... "overdumps" my core file.  (Sure, the second time, I renamed it
at a suitable time.  Nonetheless, I can't debug it.)

Symbolic links have been made as follows:
   config.h     -->     config-m68k.h
   tm.h         -->     tm-news800.h
   md           -->     m68k.md
   aux-output.c -->     output-m68k.c

Bison had no trouble at all.

I successfully generated the GNU ASSEMBLER by adding a compiler flag to
its makefile:
   -DNO_VARARGS
and copied the assembler (filename "a") to pathname /usr2/gnu/bin/as
and directory /usr2/gnu/bin is very early in my search path.
(Yes, Sony News's native assembler barfed all over me before I did
that.)

Oh yes, before reaching this illustrious point, I also had to:
   rm gnulib     # it had a bad magic number, wherever it came from
   make clean
and start all over.  I did start all over (more than once) and this pair
of faults is reproducible.

Actually I am required to have g++ working by Monday (Sunday night from
your point of view, and it's now Saturday).  So, I will now to try to
make it without using gcc.

Nonetheless, please kindly advise regarding the gcc bug.

If you wish a copy of its core dump (and/or dbx's own core dump), then
if a command should be executed to encode the binary into a form that
will not be garbled by transmission (e.g. jis / shift-jis translation
of what might look like kanji), please specify the full command line.

Thank you very much.

-- Norman Diamond