[Bug rtl-optimization/49474] [4.7 Regression] GCC 4.7-pre fails to bootstrap on stage 2: gcc/fortran/resolve.c:5360:1: internal compiler error: vector VEC(edge,base) index domain error, in find_implicit_sets at cprop.c:1346

["dougmencken at gmail dot com" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49474

--- Comment #2 from Douglas Mencken <dougmencken at gmail dot com> 2011-07-09 17:19:38 UTC ---
Tried with gcc-v4.7-20110702, now adding `--enable-stage1-languages=c,fortran'
to configure options.
Stage1 builds fine (now stage0 compiler is 4.6.1-release), even with fortran
enabled. On stage2, the process stops at the same moment as I reported
initially.

~/build-farm/_gcc_bootstrap # ls -l prev-gcc/fortran/resolve.o
-rw-r--r--    1 root     0           455536 Jul  9 00:15
prev-gcc/fortran/resolve.o
~/build-farm/_gcc_bootstrap # ls -l gcc/fortran/resolve.o
ls: gcc/fortran/resolve.o: No such file or directory

I also compared sizes of cprop.o object files on both stages, they do differ
very much (files are also uploaded to
http://ftp.osuosl.org/pub/manulix/logs+errors/gcc-v4.7-20110702/ ):

~/build-farm # ls -l stage*-cprop.o
-rw-r--r--    1 root     0           959512 Jul  9 10:59 stage1-cprop.o
-rw-r--r--    1 root     0           314088 Jul  9 10:58 stage2-cprop.o

I also figured out that gcc/cprop.c file does not exist in GCC < 4.7 source
tarballs.

So the possible reason for the error may be Steven Bosscher's efforts to split
gcse.c --> cprop.c and then "clean it up", which caused this regression.

I would be thankful for any response.