[Bug debug/18067] New: ICE at loc_descriptor_from_tree_1 in dwarf2out.c

[Bug debug/18067] New: ICE at loc_descriptor_from_tree_1 in dwarf2out.c

[phython at gcc dot gnu dot org]

While compiling snow.c from ffmpeg loc_descriptor_from_tree_1 segfaults.

Here is the stack trace:
#0  fancy_abort (file=0x1048ec30 "/home/jim/cvs/gcc.head/gcc/gcc/dwarf2out.c",
    line=9190, function=0x105e5e98 "loc_descriptor_from_tree_1")
    at /home/jim/cvs/gcc.head/gcc/gcc/diagnostic.c:556
#1  0x1018ff14 in loc_descriptor_from_tree_1 (loc=0xa, want_address=0)
    at /home/jim/cvs/gcc.head/gcc/gcc/dwarf2out.c:9190
#2  0x1018f964 in loc_descriptor_from_tree_1 (loc=0x304fd988, want_address=0)
    at /home/jim/cvs/gcc.head/gcc/gcc/dwarf2out.c:9099
#3  0x1018f964 in loc_descriptor_from_tree_1 (loc=0x304fd9b0, want_address=2)
    at /home/jim/cvs/gcc.head/gcc/gcc/dwarf2out.c:9099
#4  0x10197168 in add_bound_info (subrange_die=0x3063bb70,
    bound_attr=DW_AT_upper_bound, bound=0x304fd9b0)
    at /home/jim/cvs/gcc.head/gcc/gcc/dwarf2out.c:9223
#5  0x1019b0dc in gen_array_type_die (type=0x304ff000, context_die=0x3063b120)
    at /home/jim/cvs/gcc.head/gcc/gcc/dwarf2out.c:10271
#6  0x10195394 in gen_type_die (type=0x304ff000, context_die=0x3063b120)
    at /home/jim/cvs/gcc.head/gcc/gcc/dwarf2out.c:12101
#7  0x10193b90 in gen_decl_die (decl=0x304ff100, context_die=0x3063b120)
    at /home/jim/cvs/gcc.head/gcc/gcc/dwarf2out.c:12643
#8  0x10199ff4 in decls_for_scope (stmt=0x3050c690, context_die=0x3063b120,
    depth=0) at /home/jim/cvs/gcc.head/gcc/gcc/dwarf2out.c:12330
#9  0x10198288 in gen_subprogram_die (decl=0x304f9200, context_die=Variable
"context_die" is not available.
    at /home/jim/cvs/gcc.head/gcc/gcc/dwarf2out.c:11400
#10 0x1037b9bc in rest_of_compilation ()
    at /home/jim/cvs/gcc.head/gcc/gcc/passes.c:340
#11 0x10089694 in execute_pass_list (pass=0x105ec0c8)
    at /home/jim/cvs/gcc.head/gcc/gcc/tree-optimize.c:503
#12 0x100899f8 in tree_rest_of_compilation (fndecl=0x304f9200)
    at /home/jim/cvs/gcc.head/gcc/gcc/tree-optimize.c:635
#13 0x1001a9c0 in c_expand_body (fndecl=0x304f9200)
    at /home/jim/cvs/gcc.head/gcc/gcc/c-decl.c:6371
#14 0x103ab764 in cgraph_expand_function (node=0x30517100)
    at /home/jim/cvs/gcc.head/gcc/gcc/cgraphunit.c:1045
#15 0x103afe90 in cgraph_optimize ()
    at /home/jim/cvs/gcc.head/gcc/gcc/cgraphunit.c:2726
#16 0x1001d2c8 in c_write_global_declarations ()
    at /home/jim/cvs/gcc.head/gcc/gcc/c-decl.c:7349
#17 0x1034b4fc in toplev_main (argc=Variable "argc" is not available.
)
    at /home/jim/cvs/gcc.head/gcc/gcc/toplev.c:999
#18 0x1006710c in main (argc=Variable "argc" is not available.
) at /home/jim/cvs/gcc.head/gcc/gcc/main.c:35

-- 
           Summary: ICE at loc_descriptor_from_tree_1 in dwarf2out.c
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: debug
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: phython at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: ppc-linux
  GCC host triplet: ppc-linux
GCC target triplet: ppc-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug debug/18067] ICE at loc_descriptor_from_tree_1 in dwarf2out.c

[phython at gcc dot gnu dot org]

------- Additional Comments From phython at gcc dot gnu dot org  2004-10-19 23:05 -------
Created an attachment (id=7379)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=7379&action=view)
preprocessed source for snow.c

 Compile with -O1 -g

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug debug/18067] ICE at loc_descriptor_from_tree_1 in dwarf2out.c

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-19 23:12 -------
Might be related to PR 14900 or 17076 but I cannot reproduce this on powerpc-darwin (even after 
supplying -gdwarf-2 since dwarf-2 is not default).

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug debug/18067] ICE at loc_descriptor_from_tree_1 in dwarf2out.c

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-19 23:14 -------
I can produce with a cross to i686-pc-linux, reducing.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug tree-optimization/18067] [4.0 Regression] ICE at loc_descriptor_from_tree_1 in dwarf2out.c (VLA)

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-19 23:43 -------
Here is the reduced testcase:
typedef unsigned int size_t;
extern void *memcpy (void *__restrict __dest,
       __const void *__restrict __src, size_t __n) ;

int encode_q_branch(int level, unsigned char *t, int stride1)
 {
    const int block_w= level;
    const int stride= stride1;
    unsigned char current_mb[3][stride*block_w];
    memcpy(&current_mb[0][0], t, 1);
}


: Search converges between 2004-07-02-trunk (#477) and 2004-07-04-trunk (#478).

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
          Component|debug                       |tree-optimization
     Ever Confirmed|                            |1
           Keywords|                            |ice-on-valid-code
   Last reconfirmed|0000-00-00 00:00:00         |2004-10-19 23:43:07
               date|                            |
            Summary|ICE at                      |[4.0 Regression] ICE at
                   |loc_descriptor_from_tree_1  |loc_descriptor_from_tree_1
                   |in dwarf2out.c              |in dwarf2out.c (VLA)
   Target Milestone|---                         |4.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug tree-optimization/18067] [4.0 Regression] ICE at loc_descriptor_from_tree_1 in dwarf2out.c (VLA)

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-19 23:49 -------
Oh, I forgot to say the problem is that we are not taking the variables in the size out of SSA form for 
some reason.
Here is the tree level at .vars which shows the problem:
  saved_stack.3 = __builtin_stack_save ();
  D.1135 = __builtin_alloca ((unsigned int) (level * stride1) * 3);
  memcpy ((unsigned char *) (unsigned char[(unsigned int) (block_w_3 * stride_5 - 1)] *) (unsigned 
char[2][(unsigned int) (block_w_3 * stride_5 - 1)] *) D.1135, t, 1);
  __builtin_stack_restore (saved_stack.3);

Note the block_w_3 and stride_5.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug tree-optimization/18067] [4.0 Regression] ICE at loc_descriptor_from_tree_1 in dwarf2out.c (VLA)

[reichelt at gcc dot gnu dot org]

------- Additional Comments From reichelt at gcc dot gnu dot org  2004-10-20 00:01 -------
Here's an even shorter example (for i686-pc-linux-gnu):

=====================
void foo(int i)
{
    const int j=i+1;
    int a[1][j*j];
}
=====================

The regression seems to appear a little later with this example:
: Search converges between 2004-07-16-trunk (#487) and 2004-07-17-trunk (#488).

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |reichelt at gcc dot gnu dot
                   |                            |org
           Keywords|                            |monitored


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug tree-optimization/18067] [4.0 Regression] ICE at loc_descriptor_from_tree_1 in dwarf2out.c (VLA)

[phython at gcc dot gnu dot org]

------- Additional Comments From phython at gcc dot gnu dot org  2004-10-31 16:33 -------
I hacked around the ICE by doing nothing for SSA_NAMEs in
loc_descriptor_from_tree_1.  The optimized dump clearly shows that SSA_NAMES in
declarations are not rewritten out of SSA.
;; Function foo (foo)

foo (i)
{
  int a[D.1141][D.1138_9 * 4];
  const int j;
  void * saved_stack.1;
  void * D.1143;
  int[0][(unsigned int) (j_3 * j_3 - 1)] * a.0;
  unsigned int D.1141;
  bit_size_type D.1140;
  bit_size_type D.1139;
  unsigned int D.1138;
  int D.1137;

<bb 0>:
  saved_stack.1 = __builtin_stack_save ();
  j = i + 1;
  D.1143 = __builtin_alloca ((unsigned int) (j * j) * 4);
  __builtin_stack_restore (saved_stack.1);
  return;

}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug tree-optimization/18067] [4.0 Regression] ICE at loc_descriptor_from_tree_1 in dwarf2out.c (VLA)

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-31 17:05 -------
*** Bug 18253 has been marked as a duplicate of this bug. ***

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |atmosfear at users dot
                   |                            |sourceforge dot net


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug tree-optimization/18067] [4.0 Regression] ICE at loc_descriptor_from_tree_1 in dwarf2out.c (VLA)

[phython at gcc dot gnu dot org]

------- Additional Comments From phython at gcc dot gnu dot org  2004-11-17 15:13 -------
 As Andrew mentioned, this bug affects more than powerpc.  An enable-checking
build is needed to reproduce this bug.  With a disable-checking build the
debugging code ignores SSA_NAMEs.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
  GCC build triplet|ppc-linux                   |
   GCC host triplet|ppc-linux                   |
 GCC target triplet|ppc-linux                   |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug tree-optimization/18067] [4.0 Regression] ICE at loc_descriptor_from_tree_1 in dwarf2out.c (VLA) with const int.

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-12-19 05:13 -------
For some reason the const makes a difference.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.0 Regression] ICE at     |[4.0 Regression] ICE at
                   |loc_descriptor_from_tree_1  |loc_descriptor_from_tree_1
                   |in dwarf2out.c (VLA)        |in dwarf2out.c (VLA) with
                   |                            |const int.


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug tree-optimization/18067] [4.0 Regression] ICE at loc_descriptor_from_tree_1 in dwarf2out.c (VLA) with const int.

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-12-19 05:22 -------
This patch fixes it but I don't know if it is the correct fix, I have not bootstrapped it yet but it works as it 
removes the constraint when not to add SAVE_EXPR around the expression:
Index: tree.c
===============================================================
====
RCS file: /cvs/gcc/gcc/gcc/tree.c,v
retrieving revision 1.457
diff -u -p -r1.457 tree.c
--- tree.c      17 Dec 2004 08:17:01 -0000      1.457
+++ tree.c      19 Dec 2004 05:21:23 -0000
@@ -1669,7 +1669,6 @@ save_expr (tree expr)
   inner = skip_simple_arithmetic (t);
 
   if (TREE_INVARIANT (inner)
-      || (TREE_READONLY (inner) && ! TREE_SIDE_EFFECTS (inner))
       || TREE_CODE (inner) == SAVE_EXPR
       || TREE_CODE (inner) == ERROR_MARK)
     return t;

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug tree-optimization/18067] [4.0 Regression] ICE at loc_descriptor_from_tree_1 in dwarf2out.c (VLA) with const int.

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |pinskia at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug tree-optimization/18067] [4.0 Regression] ICE at loc_descriptor_from_tree_1 in dwarf2out.c (VLA) with const int.

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-12-19 05:51 -------
I decided after all this was the correct fix and it had been running the bootstrap/testing already and 
just finished so I decided to post it.

Patch here: <http://gcc.gnu.org/ml/gcc-patches/2004-12/msg01376.html>.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug tree-optimization/18067] [4.0 Regression] ICE at loc_descriptor_from_tree_1 in dwarf2out.c (VLA) with const int.

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-12-19 21:08 -------
Subject: Bug 18067

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	rth@gcc.gnu.org	2004-12-19 21:08:34

Modified files:
	gcc            : ChangeLog stor-layout.c 
Added files:
	gcc/testsuite/gcc.dg/debug: 20041219-1.c 

Log message:
	PR 18067
	* stor-layout.c (variable_size): Force creation of a SAVE_EXPR.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.6891&r2=2.6892
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/stor-layout.c.diff?cvsroot=gcc&r1=1.220&r2=1.221
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/debug/20041219-1.c.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067

[Bug tree-optimization/18067] [4.0 Regression] ICE at loc_descriptor_from_tree_1 in dwarf2out.c (VLA) with const int.

[rth at gcc dot gnu dot org]

------- Additional Comments From rth at gcc dot gnu dot org  2004-12-19 21:10 -------
fixed: http://gcc.gnu.org/ml/gcc-patches/2004-12/msg01377.html

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18067