Still problem bootstrapping 3.0

[Carlo Wood <carlo at alinoe dot com>]

Hiya,

I still have a problem with bootstrapping 3.0 (see PR c++/2088) on gnats.

Because bootstrapping worked 2 days ago, and a patch by Richard Kenner
has broken that (see PR c++/2087), it seemed probably to me that although
he fixed it for the mainline (3.1) it was still his patch/change (the ONLY
changes since the last check-out that worked for me) that still breaks boot-
strapping for me right now.

Since I am the only one that seems to have this problem, according to Richard,
he can't spend time on investigating this which very likely is a problem on my
side.  I fully understand that (I am a maintainer myself ;).

Hence my request to others with a i686-pc-linux-gnu to try the following
and tell me if it works for you - or otherwise report this problem too.


Here is the procedure I followed that should reproduce the problem;
note that I use '/usr/src/gcc/gcc-cvs-3.0' as cvs source tree directory
and '/usr/src/gcc/gcc-objdir-3.0' as object directory.

I thought that when it was a problem on my machine alone, then probably my
cvs check-out was screwed, so I started with a completely fresh check-out
as follows:


Move away the old cvs source tree:

  /usr/src/gcc>mv gcc-cvs-3.0 gcc-cvs-3.0.old

Check out of a fresh source tree, using branch tag "gcc-3_0-branch":

  /usr/src/gcc>cvs -z 9 -d:pserver:anoncvs@anoncvs.cygnus.com:/cvs/gcc \
                   co -dgcc-cvs-3.0 -rgcc-3_0-branch gcc

Clean up of my object directory:

  /usr/src/gcc>rm -rf gcc-objdir-3.0
  /usr/src/gcc>mkdir gcc-objdir-3.0

Configuration (please use the same flags!):

  /usr/src/gcc>env CC=gcc CFLAGS="-O9 -march=pentiumpro -fomit-frame-pointer" \
                   CXXFLAGS="-O9 -march=pentiumpro -fomit-frame-pointer" \
		   ./gcc-cvs-3.0/contrib/gcc_build \
		       -d /usr/src/gcc/gcc-cvs-3.0 \
		       -o ../gcc-objdir-3.0 \
		       -c "--prefix=/usr/local/gcc-3.0 --enable-shared --with-gnu-as --with-gnu-ld --enable-languages=c++" \
		       -m "CFLAGS='-O' LIBCFLAGS='-g -O2' LIBCXXFLAGS='-g -O2 -fno-implicit-templates'" \
		       configure

Finally, the bootstrap (same line as above but 'configure' replaced with 'bootstrap'):

  /usr/src/gcc>env CC=gcc CFLAGS="-O9 -march=pentiumpro -fomit-frame-pointer" \
                   CXXFLAGS="-O9 -march=pentiumpro -fomit-frame-pointer" \
		   ./gcc-cvs-3.0/contrib/gcc_build \
		       -d /usr/src/gcc/gcc-cvs-3.0 \
		       -o ../gcc-objdir-3.0 \
		       -c "--prefix=/usr/local/gcc-3.0 --enable-shared --with-gnu-as --with-gnu-ld --enable-languages=c++" \
		       -m "CFLAGS='-O' LIBCFLAGS='-g -O2' LIBCXXFLAGS='-g -O2 -fno-implicit-templates'" \
		       bootstrap

This will result in the error:

./xgcc -B./ -B/usr/local/gcc-3.0/i686-pc-linux-gnu/bin/ -isystem /usr/local/gcc-3.0/i686-pc-linux-gnu/include -O2   -DIN_GCC    -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -isystem ./include  -fPIC -g1  -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED  -I. -I. -I/usr/src/gcc/gcc-cvs-3.0/gcc -I/usr/src/gcc/gcc-cvs-3.0/gcc/. -I/usr/src/gcc/gcc-cvs-3.0/gcc/config -I/usr/src/gcc/gcc-cvs-3.0/gcc/../include  -fexceptions -DL_eh -c /usr/src/gcc/gcc-cvs-3.0/gcc/libgcc2.c -o libgcc/./_eh.o
/usr/src/gcc/gcc-cvs-3.0/gcc/libgcc2.c: In function `eh_context_static':
/usr/src/gcc/gcc-cvs-3.0/gcc/libgcc2.c:3335: Internal error: Segmentation fault
Please submit a full bug report, with preprocessed source if appropriate.
See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.
make[3]: *** [libgcc/./_eh.o] Error 1
make[3]: Leaving directory `/usr/src/gcc/gcc-objdir-3.0/gcc'
make[2]: *** [libgcc.a] Error 2
make[2]: Leaving directory `/usr/src/gcc/gcc-objdir-3.0/gcc'
make[1]: *** [stage1_build] Error 2
make[1]: Leaving directory `/usr/src/gcc/gcc-objdir-3.0/gcc'
make: *** [bootstrap] Error 2
gcc_build: error: Could not bootstrap the compiler


As this was a totally fresh check out, I really think this is a bug in the
compiler - perhaps one that only occurs with the flags that I use (ie, the
"-O9 -march=pentiumpro -fomit-frame-pointer" or something).

-- 
Carlo Wood <carlo@alinoe.com>