x86 --enable-checking bootstrap failure

[Marek Michalkiewicz <marekm at t19 dot ds dot pwr dot wroc dot pl>]

Trying to bootstrap CVS gcc on i686-pc-linux-gnu configured
with --enable-checking gives me this error:

./xgcc -B/usr/local/i686-pc-linux-gnu/bin/ -B./ -isystem /usr/local/i686-pc-linux-gnu/include -O2   -DIN_GCC    -g -W -Wall   -Wno-long-long -Wwrite-strings -isystem ./include  -fPIC -g1  -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED  -I. -I../../../egcs/gcc -I../../../egcs/gcc/config -I../../../egcs/gcc/../include  -DL_bb -c ../../../egcs/gcc/libgcc2.c -o libgcc/./_bb.o
../../../egcs/gcc/libgcc2.c: In function `__bb_exit_trace_func':
../../../egcs/gcc/libgcc2.c:2030: RTL check: access of elt 5 of `note' with last elt 4
../../../egcs/gcc/libgcc2.c:2030: Internal compiler error in `final_scan_insn', at final.c:2248
Please submit a full bug report.
See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.
make[3]: *** [libgcc/./_bb.o] Error 1

I started getting this a few days ago.  Is this a known problem?
The source file it crashes on is part of gcc itself so I guess I don't
have to send the preprocessed one with the bug report, but I can do it
if necessary, of course.

(I think it might be a good idea to have --enable-checking by default
in experimental versions.  It makes the compiler larger and slower, so
it should still be disabled by default for the stable releases.)

Thanks,
Marek