testcase for dwarf2 bootstrap ICE.

[Robert Lipe]

Compiling the attached program as C instead of C++ makes this problem go 
away.   

Linux/x86 configured for dwarf-2 also exhibits this problem so it isn't
really a SCO-specific thing.  I just happen to be the lucky one. :-)

# This directory was configured as follows:
/play/egcs/configure --with-gcc-version-trigger=/play/egcs/gcc/version.c --host=i686-pc-linux-gnu --with-dwarf2 --enable-languages=c,c++ --norecursion 
#  using "mt-frag"
(robertl) rjlinux:/home/robertl/tmp/egcs/gcc
$ ./xgcc -B./ -c  -gdwarf-2 /tmp/q.cc
/tmp/q.cc:1: Internal compiler error in `modified_type_die', at dwarf2out.c:6473




Robert Lipe wrote:
> I started hacking down the code to get a testcase for htis ICE.  Tests
> don't often resolve clear down to a one-liner. :-)
> 
> $ cat q.cc
> typedef volatile struct { } blah_t;
> (robertl) rjludi:/home3/tmp/x/gcc
> $ ./xgcc    -g  -c -B./ q.cc
> q.cc:1: Internal compiler error in `modified_type_die', at dwarf2out.c:6473
> 
> If either 'volatile' is removed or the struct is given an identifier the
> code doesn't ICE.
> 
> Hope this helps.
> 
> RJL
> 
> 
> 
> >   > The one that I have run into (ironically three times this very morning)
> >   > that I'd like to add is the dwarf2 bootstrap ICE that I just mentioned.
> >   > 
> >   >   /home3/tmp/x/gcc/xgcc -B/home3/tmp/x/gcc/ -B/usr/local/i686-UnixWare7.1.0
> >   > -sysv5/bin/ -c -g -O2 -fno-implicit-templates -I. -I/play/egcs/libio -nostd
> >   > inc++  -fpic /play/egcs/libio/filebuf.cc -o pic/filebuf.o
> >   > /play/egcs/libio/filebuf.cc:215: Internal compiler error in `modified_type_
> >   > die', at dwarf2out.c:6473
> >   > Please submit a full bug report to `egcs-bugs@egcs.cygnus.com'.
> >   > See <URL: http://egcs.cygnus.com/faq.html#bugreport > for details.
> > We'll see.  Depends on degree of difficulty :-)