[Bug middle-end/56074] New: [4.8 regression] ICE compiling gcc.dg/vect/pr49093.c

["ro at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56074

             Bug #: 56074
           Summary: [4.8 regression] ICE compiling gcc.dg/vect/pr49093.c
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: ro@gcc.gnu.org
                CC: tejohnson@gcc.gnu.org
              Host: *-*-solaris2.{9,10}
            Target: *-*-solaris2.{9,10}
             Build: *-*-solaris2.{9,10}


Between 20130102 and 20130104, several gcc.dg/vect testcases started failing:

FAIL: gcc.dg/vect/pr49093.c (internal compiler error)
FAIL: gcc.dg/vect/pr49093.c (test for excess errors)

on both Solaris/SPARC and x86, many others only on Solaris/x86:

FAIL: gcc.dg/vect/bb-slp-1.c (internal compiler error)
FAIL: gcc.dg/vect/bb-slp-1.c (test for excess errors)
WARNING: gcc.dg/vect/bb-slp-1.c compilation failed to produce executable
FAIL: gcc.dg/vect/bb-slp-1.c scan-tree-dump-times slp "basic block vectorized
using SLP" 1

I'm always seeing a SEGV in strlen, probably from strlen(NULL):

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 1 (LWP 1)]
0xfee163ec in strlen () from /lib/libc.so.1
(gdb) where
#0  0xfee163ec in strlen () from /lib/libc.so.1
#1  0xfee7199e in _ndoprnt () from /lib/libc.so.1
#2  0xfee7454e in fprintf () from /lib/libc.so.1
#3  0x082f1fa8 in dump_loc (dump_kind=<optimized out>, 
    dfile=0x914b140 <_iob+64>, loc=2147483648)
    at /vol/gcc/src/hg/trunk/local/gcc/dumpfile.c:270
#4  0x082f261a in dump_printf_loc (dump_kind=268435456, loc=2147483648, 
    format=0x8dab89c <error reading variable>)
    at /vol/gcc/src/hg/trunk/local/gcc/dumpfile.c:370
#5  0x08798336 in vect_analyze_loop (loop=0xfea9a7d0)
    at /vol/gcc/src/hg/trunk/local/gcc/tree-vect-loop.c:1749
#6  0x087adf64 in vectorize_loops ()
    at /vol/gcc/src/hg/trunk/local/gcc/tree-vectorizer.c:113
#7  0x08520bdb in execute_one_pass (pass=pass@entry=0x90be020 <pass_vectorize>)
    at /vol/gcc/src/hg/trunk/local/gcc/passes.c:2333
#8  0x08520fed in execute_pass_list (pass=<optimized out>)
    at /vol/gcc/src/hg/trunk/local/gcc/passes.c:2381
#9  0x08521000 in execute_pass_list (pass=<optimized out>)
    at /vol/gcc/src/hg/trunk/local/gcc/passes.c:2382
#10 0x08521000 in execute_pass_list (pass=<optimized out>)
    at /vol/gcc/src/hg/trunk/local/gcc/passes.c:2382
#11 0x082b4630 in expand_function (node=0xfea0e3f0)
    at /vol/gcc/src/hg/trunk/local/gcc/cgraphunit.c:1640
#12 0x082b664d in expand_all_functions ()
    at /vol/gcc/src/hg/trunk/local/gcc/cgraphunit.c:1744
#13 compile () at /vol/gcc/src/hg/trunk/local/gcc/cgraphunit.c:2042
#14 0x082b6d2a in finalize_compilation_unit ()
    at /vol/gcc/src/hg/trunk/local/gcc/cgraphunit.c:2119
#15 0x0818aced in c_write_global_declarations ()
    at /vol/gcc/src/hg/trunk/local/gcc/c/c-decl.c:10118
#16 0x085c36cd in compile_file ()
    at /vol/gcc/src/hg/trunk/local/gcc/toplev.c:557
#17 0x085c56a5 in do_compile ()
    at /vol/gcc/src/hg/trunk/local/gcc/toplev.c:1876
#18 toplev_main (argc=9, argv=0x804751c)
    at /vol/gcc/src/hg/trunk/local/gcc/toplev.c:1952
#19 0x08cfcc2b in main (argc=9, argv=0x804751c)
    at /vol/gcc/src/hg/trunk/local/gcc/main.c:36

On Solaris 11 and Linux/x86_64 I don't see those SEGVs, but the dump file
contains

(null):0: note: ===== analyze_loop_nest =====
(null):0: note: === vect_analyze_loop_form ===
(null):0: note: not vectorized: control flow in loop.
(null):0: note: bad loop form.

which clearly indicates that NULL was passed as %s argument.

I strongly suppose this is from the following patch:

2013-01-02  Teresa Johnson  <tejohnson@google.com>

    * dumpfile.c (dump_loc): Print filename with location.
[...]

Please fix; this is a regression from 4.7.

  Rainer