[Bug target/41930] New: [4.5 regression] cc1 SEGV compiling maxval_r16.c

Between 20091005 and 20091102, bootstrapping Tru64 UNIX V5.1B started to fail
in
libgfortan:

$ /vol/gcc/obj/gcc-4.5.0-20091102/5.1b-gcc/./gcc/xgcc
-B/vol/gcc/obj/gcc-4.5.0-20091102/5.1b-gcc/./gcc/
-B/vol/gcc/alpha-dec-osf5.1b/bin/ -B/vol/gcc/alpha-dec-osf5.1b/lib/ -isystem
/vol/gcc/alpha-dec-osf5.1b/include -isystem
/vol/gcc/alpha-dec-osf5.1b/sys-include -DHAVE_CONFIG_H -I.
-I/vol/gcc/src/gcc-dist/libgfortran -iquote/vol/gcc/src/gcc-dist/libgfortran/io
-I/vol/gcc/src/gcc-dist/libgfortran/../gcc
-I/vol/gcc/src/gcc-dist/libgfortran/../gcc/config -I../.././gcc -D_GNU_SOURCE
-std=gnu99 -Wall -Wstrict-prototypes -Wmissing-prototypes
-Wold-style-definition -Wextra -Wwrite-strings -fcx-fortran-rules -g -O2 -mieee
-MT maxval_r16.lo -MD -MP -MF .deps/maxval_r16.Tpo -c
/vol/gcc/src/gcc-dist/libgfortran/generated/maxval_r16.c  -DPIC -o
.libs/maxval_r16.o
/vol/gcc/src/gcc-dist/libgfortran/generated/maxval_r16.c: In function
'mmaxval_r16':
/vol/gcc/src/gcc-dist/libgfortran/generated/maxval_r16.c:413:1: internal
compiler error: Segmentation fault

Unfortunately, I cannot investigate the crash with gdb 6.6 since gdb crashes
itself when run on cc1.  I'll try with 7.0 instead.

Interestingly, it works ok on alpha-dec-osf4.0f: it seems to be due to the fact
that V5.1B supports 128-bit long double, while V4.0F does not.


--
           Summary: [4.5 regression] cc1 SEGV compiling maxval_r16.c
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: ro at gcc dot gnu dot org
 GCC build triplet: alpha-dec-osf5.1b
  GCC host triplet: alpha-dec-osf5.1b
GCC target triplet: alpha-dec-osf5.1b


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

------- Comment #1 from ro at CeBiTec dot Uni-Bielefeld dot DE  2009-11-04 17:52 -------
Subject:  [4.5 regression] cc1 SEGV compiling maxval_r16.c

Unfortunately, gdb 7.0 crashes as well, so I cannot investigate the
problem this way.  But I've been able to produce a minimized testcase,
which I'll attach.

I'll build a cross compiler from i386-pc-solaris2.10 and see if I can
reproduce the problem there.

          Rainer


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

------- Comment #2 from ro at gcc dot gnu dot org  2009-11-04 17:53 -------
Created an attachment (id=18968)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18968&action=view)
minimized testcase

Compile with gcc -std=c99 -c -O2 to produce SEGV.


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

------- Comment #3 from ro at CeBiTec dot Uni-Bielefeld dot DE  2009-11-04 18:15 -------
Subject: Re:  [4.5 regression] cc1 SEGV compiling maxval_r16.c


> I'll build a cross compiler from i386-pc-solaris2.10 and see if I can
> reproduce the problem there.

Unfortunately, the problem does not reproduce in such a cross compiler,
so I'll have to either start a native reghunt or find a way to debug the
native compiler (maybe compiling with -g without -O2 or try the stage1
compiler if it is also affected).

         Rainer


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

------- Comment #4 from ro at gcc dot gnu dot org  2009-11-04 18:32 -------
Richard, this may be due to your patch.


--

ro at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rsandifo at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

------- Comment #5 from ro at CeBiTec dot Uni-Bielefeld dot DE  2009-11-04 18:34 -------
Subject: Re:  [4.5 regression] cc1 SEGV compiling maxval_r16.c

> Unfortunately, the problem does not reproduce in such a cross compiler,
> so I'll have to either start a native reghunt or find a way to debug the
> native compiler (maybe compiling with -g without -O2 or try the stage1
> compiler if it is also affected).

I was misled by the fact that the hg mirror of trunk hasn't been updated
in almost two months, so I've been testing the wrong sources.  With svn
trunk, I can reproduce the problem:

run -std=gnu99 -O2 maxval_r16.i
Starting program: /vol/gcc/obj/gcc-reghunt/153800/gcc/cc1 -std=gnu99 -O2
maxval_r16.i
warning: Temporarily disabling breakpoints for unloaded shared library
"/usr/lib/ld.so.1"
Breakpoint 3 at 0xfeef4b00
Breakpoint 4 at 0xfef01652
 mmaxval_r16
Analyzing compilation unit
Performing interprocedural optimizations
 <visibility> <> <early_local_cleanups> <whole-program> <cp> <inline>
<static-var> <pure-const>Assembling functions:
 mmaxval_r16
Program received signal SIGSEGV, Segmentation fault.
simplify_replace_fn_rtx (x=0x0, old_rtx=0xfee55df0, fn=0, data=0xfee5fc08) at
/vol/gcc/src/gcc-dist/gcc/simplify-rtx.c:364
(gdb) where
#0  simplify_replace_fn_rtx (x=0x0, old_rtx=0xfee55df0, fn=0, data=0xfee5fc08)
at /vol/gcc/src/gcc-dist/gcc/simplify-rtx.c:364
#1  0x0836d367 in simplify_replace_fn_rtx (x=0xfee7f030, old_rtx=0xfee55df0,
fn=0, data=0xfee5fc08) at /vol/gcc/src/gcc-dist/gcc/simplify-rtx.c:492
#2  0x0836d367 in simplify_replace_fn_rtx (x=0xfee7f018, old_rtx=0xfee55df0,
fn=0, data=0xfee5fc08) at /vol/gcc/src/gcc-dist/gcc/simplify-rtx.c:492
#3  0x0836d367 in simplify_replace_fn_rtx (x=0xfee7f00c, old_rtx=0xfee55df0,
fn=0, data=0xfee5fc08) at /vol/gcc/src/gcc-dist/gcc/simplify-rtx.c:492
#4  0x0836d6e2 in simplify_replace_rtx (x=0xfee7f00c, old_rtx=0xfee55df0,
new_rtx=0xfee5fc08) at /vol/gcc/src/gcc-dist/gcc/simplify-rtx.c:510
#5  0x082507ef in try_replace_reg (from=0xfee55df0, to=0xfee5fc08,
insn=0xfee5369c) at /vol/gcc/src/gcc-dist/gcc/gcse.c:2278
#6  0x08250c46 in constprop_register (insn=0xfee5369c, from=0xfee55df0,
to=0xfee5fc08) at /vol/gcc/src/gcc-dist/gcc/gcse.c:2511
#7  0x08251e14 in execute_rtl_cprop () at /vol/gcc/src/gcc-dist/gcc/gcse.c:2701
#8  0x082ea4b0 in execute_one_pass (pass=0x87a3240) at
/vol/gcc/src/gcc-dist/gcc/passes.c:1518
#9  0x082ea577 in execute_pass_list (pass=0x87a3240) at
/vol/gcc/src/gcc-dist/gcc/passes.c:1567
#10 0x082ea58b in execute_pass_list (pass=0x87a38a0) at
/vol/gcc/src/gcc-dist/gcc/passes.c:1568
#11 0x083d6649 in tree_rest_of_compilation (fndecl=0xfee62000) at
/vol/gcc/src/gcc-dist/gcc/tree-optimize.c:392
#12 0x0852436c in cgraph_expand_function (node=0xfee64000) at
/vol/gcc/src/gcc-dist/gcc/cgraphunit.c:1160
#13 0x0852609d in cgraph_optimize () at
/vol/gcc/src/gcc-dist/gcc/cgraphunit.c:1219
#14 0x085263f2 in cgraph_finalize_compilation_unit () at
/vol/gcc/src/gcc-dist/gcc/cgraphunit.c:1089
#15 0x080beb4f in c_write_global_declarations () at
/vol/gcc/src/gcc-dist/gcc/c-decl.c:9489
#16 0x08381ed3 in toplev_main (argc=4, argv=0x80477bc) at
/vol/gcc/src/gcc-dist/gcc/toplev.c:1061
#17 0x08126fdf in main (argc=4, argv=0x80477bc) at
/vol/gcc/src/gcc-dist/gcc/main.c:35

simplify_replace_fn_rtx is called with x = NULL above, thus GET_MODE (x)
segfaults.  So it seems likely that the introduction of
simplify_replace_fn_rtx is the culprit.

        Rainer


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

--

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu dot
                   |                            |org
          Component|target                      |middle-end
           Keywords|                            |build, ice-on-valid-code
   Target Milestone|---                         |4.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

--

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P4


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

------- Comment #6 from ubizjak at gmail dot com  2009-11-11 10:01 -------
(In reply to comment #3)
> Subject: Re:  [4.5 regression] cc1 SEGV compiling maxval_r16.c
>
> > I'll build a cross compiler from i386-pc-solaris2.10 and see if I can
> > reproduce the problem there.
>
> Unfortunately, the problem does not reproduce in such a cross compiler,
> so I'll have to either start a native reghunt or find a way to debug the
> native compiler (maybe compiling with -g without -O2 or try the stage1
> compiler if it is also affected).

Add --with-long-double-128 to your configure flags and it will also trigger
with a cross compiler.

Confirmed with a cross from x86_64-pc-linux-gnu.


--

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2009-11-11 10:01:06
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

------- Comment #7 from ubizjak at gmail dot com  2009-11-11 10:37 -------
We recurse with:

simplify_replace_fn_rtx called, x =
(expr_list:REG_DEP_TRUE (symbol_ref:DI ("_OtsGeqX") [flags 0x41])
    (expr_list:REG_DEP_TRUE (mem:TF (reg/f:DI 75 [ array_3(D)->data ]) [3 S16
A128])
        (expr_list:REG_DEP_TRUE (reg:TF 76)
            (nil))))

simplify_replace_fn_rtx called, x =
(symbol_ref:DI ("_OtsGeqX") [flags 0x41])

simplify_replace_fn_rtx called, x =
(expr_list:REG_DEP_TRUE (mem:TF (reg/f:DI 75 [ array_3(D)->data ]) [3 S16
A128])
    (expr_list:REG_DEP_TRUE (reg:TF 76)
        (nil)))

simplify_replace_fn_rtx called, x =
(mem:TF (reg/f:DI 75 [ array_3(D)->data ]) [3 S16 A128])

simplify_replace_fn_rtx called, x =
(reg/f:DI 75 [ array_3(D)->data ])

simplify_replace_fn_rtx called, x =
(expr_list:REG_DEP_TRUE (reg:TF 76)
    (nil))

simplify_replace_fn_rtx called, x =
(reg:TF 76)

gcc dies in "case 'e':", recursing with (nil) from:

(expr_list:REG_DEP_TRUE (reg:TF 76)
    (nil))

The fix is then trivial, like [no whitespace diff]:

--cut here--
Index: simplify-rtx.c
===================================================================
--- simplify-rtx.c      (revision 154088)
+++ simplify-rtx.c      (working copy)
@@ -489,6 +489,8 @@ simplify_replace_fn_rtx (rtx x, const_rt
        break;

       case 'e':
+       if (XEXP (x, i))
+         {
        op = simplify_replace_fn_rtx (XEXP (x, i), old_rtx, fn, data);
        if (op != XEXP (x, i))
          {
@@ -496,6 +498,7 @@ simplify_replace_fn_rtx (rtx x, const_rt
              newx = shallow_copy_rtx (x);
            XEXP (newx, i) = op;
          }
+         }
        break;
       }
   return newx;
--cut here--


--

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |ubizjak at gmail dot com
                   |dot org                     |
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2009-11-11 10:01:06         |2009-11-11 10:37:52
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

------- Comment #8 from ubizjak at gmail dot com  2009-11-11 11:10 -------
Testing patch (it bootstraps OK on alpha-linux-gnu).


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

------- Comment #9 from uros at gcc dot gnu dot org  2009-11-12 07:30 -------
Subject: Bug 41930

Author: uros
Date: Thu Nov 12 07:30:05 2009
New Revision: 154116

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=154116
Log:
        PR middle-end/41930
        * simplify-rtx.c (simplify_replace_fn_rtx) <case 'e'>: Skip
        processing if XEXP (x, i) is NULL.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/simplify-rtx.c


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

------- Comment #10 from ubizjak at gmail dot com  2009-11-12 07:40 -------
Fixed.


--

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |http://gcc.gnu.org/ml/gcc-
                   |                            |patches/2009-
                   |                            |11/msg00562.html
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930

------- Comment #11 from ro at techfak dot uni-bielefeld dot de  2009-11-12 16:46 -------
Subject: Re:  [4.5 regression] cc1 SEGV compiling maxval_r16.c

Bootstrap finishes with your patch on alpha-dec-osf5.1b; the testsuite is
still running.


--


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41930