Bootstrap of GCC mainline on powerpc64-unknown-linux-gnu fails on some
distributions with an ICE while building libiberty in stage3, using the stage2
compiler:

------------------------------------------------
/home/gccbuild/build/nightly/build-gcc-trunk/./prev-gcc/xgcc
-B/home/gccbuild/build/nightly/build-gcc-trunk/./prev-gcc/
-B/opt/gcc-nightly/trunk-20090402/powerpc64-linux/bin/ -c -DHAVE_CONFIG_H -O2
-g  -I. -I/home/gccbuild/gcc_trunk_anonsvn/gcc/libiberty/../include  -W -Wall
-Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic 
/home/gccbuild/gcc_trunk_anonsvn/gcc/libiberty/hashtab.c -o hashtab.o
/home/gccbuild/gcc_trunk_anonsvn/gcc/libiberty/hashtab.c: In function
‘htab_find’:
/home/gccbuild/gcc_trunk_anonsvn/gcc/libiberty/hashtab.c:597: error: abnormal
edges for no purpose in bb 2
/home/gccbuild/gcc_trunk_anonsvn/gcc/libiberty/hashtab.c:597: error: in basic
block 2:
/home/gccbuild/gcc_trunk_anonsvn/gcc/libiberty/hashtab.c:597: error: flow
control insn inside a basic block
(call_insn/j 14 11 32 2
/home/gccbuild/gcc_trunk_anonsvn/gcc/libiberty/hashtab.c:596 (parallel [
            (set (reg:SI 3 3)
                (call (mem:SI (symbol_ref:SI ("htab_find_with_hash") [flags
0x3] <function_decl 0xf7e35e00 htab_find_with_hash>) [0 S4 A8])
                    (const_int 0 [0x0])))
            (use (const_int 0 [0x0]))
            (use (reg:SI 65 lr))
            (return)
        ]) 469 {*sibcall_value_local32} (expr_list:REG_EH_REGION (const_int 0
[0x0])
        (nil))
    (expr_list:REG_DEP_TRUE (use (reg:SI 5 5))
        (expr_list:REG_DEP_TRUE (use (reg:SI 4 4))
            (expr_list:REG_DEP_TRUE (use (reg:SI 3 3))
                (nil)))))
/home/gccbuild/gcc_trunk_anonsvn/gcc/libiberty/hashtab.c:597: internal compiler
error: in rtl_verify_flow_info_1, at cfgrtl.c:1920
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
------------------------------------------------

The failures start with r145204, a merge from pretty-ipa, and go away when that
patch is reverted from current mainline.  Since the failures happen on some
sysems (two different machines running SLES10) but not others (SLES9, RHEL4.3)
it's possible that they are actually intermittent and not really related to
this revision at all, or are due to a latent bug.

I tried starting from different versions of GCC and it made no difference.  The
builds on all of these systems used binutils-2.19.

Configure options (other than prefix/with-gmp/with-mpfr/with-as/with-ld) were

  --with-cpu=default32
  --with-long-double-128
  --enable-secureplt
  --enable-languages=c

It also fails, in a shorter time, with 

  --build=powerpc-linux
  --host=powerpc-linux
  --target=powerpc-linux
  --with-long-double-128
  --enable-secureplt
  --enable-languages=c


-- 
           Summary: bootstrap failure with ICE building libiberty in stage3
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: major
          Priority: P3
         Component: bootstrap
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: janis at gcc dot gnu dot org
 GCC build triplet: powerpc-linux
  GCC host triplet: powerpc-linux
GCC target triplet: powerpc-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39617

Reply via email to