bootstrap failure, linux, today's mainline

[Paul Koning <pkoning at equallogic dot com>]

I just copied over the current CVS (rsync the CVS database, then a
clean checkout of the result).  Did a bootstrap, it fails in stage 2.

This is RedHat linux 7.2, kernel 2.4.18, base gcc is 3.0.2, binutils
from the current binutils cvs server.

Configure as follows:
#!/bin/sh
# This file was generated automatically by configure.  Do not edit.
# This directory was configured as follows:
/lhome/pkoning/gcccvs/gcc-20020605/gcc/configure --with-gcc-version-trigger=/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/version.c --host=i686-pc-linux-gnu --prefix=/usr/test --with-gnu-as --with-as=/usr/test/bin/as --with-gnu-ld --with-ld=/usr/test/bin/ld --norecursion 
#  using "mt-frag"

The failure looks like this:
 ...
make[2]: Leaving directory `/lhome/pkoning/builds/bootgcc2/gcc'
echo timestamp > stage1_copy
echo stage2_build > stage_last
make CC="stage1/xgcc -Bstage1/ -B/usr/test/i686-pc-linux-gnu/bin/" \
	 STAGE_PREFIX=stage1/ \
	 ADAC="\$(CC)" CFLAGS="-g -O2" LDFLAGS="" WARN_CFLAGS="\$(GCC_WARN_CFLAGS)" STRICT_WARN="-Wtraditional -pedantic -Wno-long-long" libdir=/usr/test/lib LANGUAGES="c gcov c++ f77 java objc" MAKEOVERRIDES= OUTPUT_OPTION="-o \$@"
make[2]: Entering directory `/lhome/pkoning/builds/bootgcc2/gcc'
stage1/xgcc -Bstage1/ -B/usr/test/i686-pc-linux-gnu/bin/ -c -DIN_GCC    -g -O2 -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -Wtraditional -pedantic -Wno-long-long  -DHAVE_CONFIG_H -DGENERATOR_FILE    -I. -I. -I/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc -I/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/. -I/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/config -I/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/../include /lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/gengenrtl.c -o gengenrtl.o
In file included from /lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/gengenrtl.c:29:/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/real.h:122:8: #error "REAL_WIDTH > 5 not supported"
In file included from /lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/gengenrtl.c:40:/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/rtl.def:727: `CONST_DOUBLE_FORMAT' undeclared here (not in a function)
/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/rtl.def:727: initializer element is not constant
/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/rtl.def:727: (near initialization for `defs[65].format')
/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/rtl.def:727: warning: missing initializer
/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/rtl.def:727: warning: (near initialization for `defs[65].format')
/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/rtl.def:727: initializer element is not constant
/lhome/pkoning/gcccvs/gcc-20020605/gcc/gcc/rtl.def:727: (near
initialization for `defs[65]') ...

and lots more error messages, probably all caused by the initial
error.

I ran into this yesterday with a slightly older CVS checkout.  In that
one I did a --save-temps -dD and looked at the resulting rtl.i to see
whether that explains the #error.  It doesn't; the REAL_WIDTH I see in
rtl.i evaluates to 3.  So it looks like something wrong in expression
evaluation in the preprocessor...

     paul