[Bug tree-optimization/21541] New: [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[Bug tree-optimization/21541] New: [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[jsm28 at gcc dot gnu dot org]

FAIL: gcc.c-torture/execute/20031215-1.c compilation,  -O2 
FAIL: gcc.c-torture/execute/20031215-1.c compilation,  -O3 -fomit-frame-pointer 
FAIL: gcc.c-torture/execute/20031215-1.c compilation,  -O3 -g 
FAIL: gcc.c-torture/execute/20031215-1.c compilation,  -Os 
UNRESOLVED: gcc.c-torture/execute/20031215-1.c execution,  -O2 
UNRESOLVED: gcc.c-torture/execute/20031215-1.c execution,  -O3 -fomit-frame-pointer 
UNRESOLVED: gcc.c-torture/execute/20031215-1.c execution,  -O3 -g 
UNRESOLVED: gcc.c-torture/execute/20031215-1.c execution,  -Os 

have appeared on mainline between 20050511 and 20050512 on platforms including
at least i686-pc-linux-gnu and hppa{2.0w,64}-hp-hpux11.{11,23}.  gcc-testresults
shows other platforms as well.

/home/gcc/nightlies/src-mainline-2005-05-12/gcc-mainline-2005-05-12/gcc/testsuite/gcc.c-torture/execute/20031215-1.c:
In function 'test1':
/home/gcc/nightlies/src-mainline-2005-05-12/gcc-mainline-2005-05-12/gcc/testsuite/gcc.c-torture/execute/20031215-1.c:11:
error: Statement makes a memory store, but has no V_MAY_DEFS nor V_MUST_DEFS
#   VUSE <ao_1>;
ao.ch[2] = 0;
/home/gcc/nightlies/src-mainline-2005-05-12/gcc-mainline-2005-05-12/gcc/testsuite/gcc.c-torture/execute/20031215-1.c:11:
internal compiler error: verify_ssa failed.
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.

-- 
           Summary: [4.1 Regression] gcc.c-torture/execute/20031215-1.c
                    compilation fails
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: jsm28 at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-05-12 21:57 -------
This is why I hate the type system of GCC and what is consider compatible and such for the tree 
optimizers

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
           Keywords|                            |ice-on-valid-code
   Last reconfirmed|0000-00-00 00:00:00         |2005-05-12 21:57:33
               date|                            |
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-05-13 12:46 -------
*** Bug 21545 has been marked as a duplicate of this bug. ***

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[rguenth at gcc dot gnu dot org]

------- Additional Comments From rguenth at gcc dot gnu dot org  2005-05-13 12:48 -------
See duplicate 21545 for (some) analysis.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[rguenth at gcc dot gnu dot org]

------- Additional Comments From rguenth at gcc dot gnu dot org  2005-05-14 16:45 -------
Reduced testcase:

typedef struct { char ch[3]; } pstr;
const pstr ao = { "OK" };
const pstr * const a = &ao;

void test1(void)
{
   ((char *)a->ch)[2] = 0;
}

aliasing chokes on gimplification stripping 02.original

 *((char *) (const char *) &ao.ch + 2B) = 0;

down to 03.gimple

  char * D.1238;

  D.1238 = &ao.ch[0] + 2B;
  *D.1238 = 0;

which has "confused" types for the rhs of
  D.1238 = &ao.ch[0] + 2B;


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[rguenth at gcc dot gnu dot org]

------- Additional Comments From rguenth at gcc dot gnu dot org  2005-05-14 16:48 -------
actually trees look like

3845		  ret = gimplify_modify_expr (expr_p, pre_p, post_p,
Value returned is $3 = GS_ALL_DONE
(gdb) call debug_tree(*expr_p)
 <modify_expr 0x401481d4
    type <pointer_type 0x40156bd0
        type <integer_type 0x401493cc char public QI
            size <integer_cst 0x40141210 constant invariant 8>
            unit size <integer_cst 0x40141228 constant invariant 1>
            align 8 symtab 0 alias set -1 precision 8 min <integer_cst
0x40141270 -128> max <integer_cst 0x401412d0 127>
            pointer_to_this <pointer_type 0x40156bd0>>
        unsigned SI
        size <integer_cst 0x40141408 constant invariant 32>
        unit size <integer_cst 0x40141198 constant invariant 4>
        align 32 symtab 0 alias set -1>
    side-effects
    arg 0 <var_decl 0x401be6c0 D.1238 type <pointer_type 0x40156bd0>
        used unsigned ignored SI file 20031215-1.c line 12 size <integer_cst
0x40141408 32> unit size <integer_cst 0x40141198 4>
        align 32 context <function_decl 0x401be438 test1>>
    arg 1 <plus_expr 0x4014818c type <pointer_type 0x40156bd0>
        constant invariant
        arg 0 <addr_expr 0x40146400 type <pointer_type 0x4015821c>
            constant invariant
            arg 0 <array_ref 0x401b91d4 type <integer_type 0x401581b0 char>
               
                arg 0 <component_ref 0x4014e078 type <array_type 0x401b7e58>
                    readonly arg 0 <var_decl 0x401be0d8 ao> arg 1 <field_decl
0x401b7ec4 ch>>
                arg 1 <integer_cst 0x401411b0 constant invariant 0> arg 2
<integer_cst 0x401411b0 0> arg 3 <integer_cst 0x40141228 1>>>
        arg 1 <integer_cst 0x40141d50 constant invariant 2>>
    20031215-1.c:12>

which is mismatched types for the PLUS_EXPR.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[rguenth at gcc dot gnu dot org]

------- Additional Comments From rguenth at gcc dot gnu dot org  2005-05-14 17:51 -------
A little more simplification:

typedef struct { char ch; } pstr;
const pstr ao = { 'O' };

void test1(void)
{
   *((char *)&ao.ch) = 1;
}

.02.original:
  *(char *) &ao.ch = 1;

.02.gimple:
  ao.ch = 1;

which doesn't have the PLUS_EXPR but ends up with a modify expr
that has a const lhs.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[rguenth at gcc dot gnu dot org]

------- Additional Comments From rguenth at gcc dot gnu dot org  2005-05-14 18:28 -------
gimplify_expr calls fold_indirect_ref on *(char *)&ao.ch which happily folds
away the cast.  Disabling fold_indirect_ref leaves the stripping to
STRIP_USELESS_TYPE_CONVERSION, so we end up with

  const char * D.1237;

  D.1237 = &ao.ch;
  *D.1237 = 1;

resulting from gimplify_expr (
 <nop_expr 0x401462c0
    type <pointer_type 0x40156bd0
        type <integer_type 0x401493cc char public QI
            size <integer_cst 0x40141210 constant invariant 8>
            unit size <integer_cst 0x40141228 constant invariant 1>
            align 8 symtab 0 alias set -1 precision 8 min <integer_cst
0x40141270 -128> max <integer_cst 0x401412d0 127>
            pointer_to_this <pointer_type 0x40156bd0>>
        unsigned SI
        size <integer_cst 0x40141408 constant invariant 32>
        unit size <integer_cst 0x40141198 constant invariant 4>
        align 32 symtab 0 alias set -1>
    readonly constant invariant
    arg 0 <addr_expr 0x401462a0
        type <pointer_type 0x4015821c type <integer_type 0x401581b0 char>
            unsigned SI size <integer_cst 0x40141408 32> unit size <integer_cst
0x40141198 4>
            align 32 symtab 0 alias set -1
            pointer_to_this <pointer_type 0x4015857c>>
        readonly constant invariant
        arg 0 <component_ref 0x4014e078 type <integer_type 0x401493cc char>
            readonly arg 0 <var_decl 0x401be000 ao> arg 1 <field_decl 0x401b7dec
ch>>>>
)

and tree_ssa_useless_type_conversion_1 tells us that the conversion from

 <pointer_type 0x4015821c
    type <integer_type 0x401581b0 char readonly QI
        size <integer_cst 0x40141210 constant invariant 8>
        unit size <integer_cst 0x40141228 constant invariant 1>
        align 8 symtab 0 alias set -1 precision 8 min <integer_cst 0x40141270
-128> max <integer_cst 0x401412d0 127>
        pointer_to_this <pointer_type 0x4015821c>>
    unsigned SI
    size <integer_cst 0x40141408 type <integer_type 0x40149288 bit_size_type>
constant invariant 32>
    unit size <integer_cst 0x40141198 type <integer_type 0x4014921c unsigned
int> constant invariant 4>
    align 32 symtab 0 alias set -1
    pointer_to_this <pointer_type 0x4015857c>>

to

 <pointer_type 0x40156bd0
    type <integer_type 0x401493cc char public QI
        size <integer_cst 0x40141210 constant invariant 8>
        unit size <integer_cst 0x40141228 constant invariant 1>
        align 8 symtab 0 alias set -1 precision 8 min <integer_cst 0x40141270
-128> max <integer_cst 0x401412d0 127>
        pointer_to_this <pointer_type 0x40156bd0>>
    unsigned SI
    size <integer_cst 0x40141408 type <integer_type 0x40149288 bit_size_type>
constant invariant 32>
    unit size <integer_cst 0x40141198 type <integer_type 0x4014921c unsigned
int> constant invariant 4>
    align 32 symtab 0 alias set -1>

is useless because c_types_compatible_p in turn thinks "char" and "const char"
are compatible because they share the same TYPE_MAIN_VARIANT "char".

So we finally strip the NOP_EXPR doing the cast to char* from const char*.

If we disable that part of tree_ssa_useless_type_conversion_1, we don't ICE
any longer, but alias1 still does not insert V_MAY_DEFs here and dce1 finally
deletes the store.  Which may be a valid optimization; .t17.alias1:

test1 ()
{
  char * D.1237;

<bb 0>:
  D.1237_1 = (char *) &ao.ch;
  #   VUSE <SFT.0_2>;
  *D.1237_1 = 1;
  return;
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[dberlin at gcc dot gnu dot org]

------- Additional Comments From dberlin at gcc dot gnu dot org  2005-05-15 17:21 -------
No, these should be getting may-defs in either form, so someting is not going
right :(

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-05-15 17:26 -------
*** Bug 21585 has been marked as a duplicate of this bug. ***

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[hp at gcc dot gnu dot org]

------- Additional Comments From hp at gcc dot gnu dot org  2005-05-16 19:48 -------
I don't see a blame for this regression, so here goes.
Passed with LAST_UPDATED "Wed May 11 04:48:16 UTC 2005" on cris-elf and
mmix-knuth-mmixware, failed with "Wed May 11 12:21:26 UTC 2005" and since.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[rguenth at gcc dot gnu dot org]

------- Additional Comments From rguenth at gcc dot gnu dot org  2005-05-16 19:57 -------
2005-05-11  Richard Guenther  <rguenth@gcc.gnu.org>

        PR middle-end/19807
        PR tree-optimization/19639
        * fold-const.c (try_move_mult_to_index): Handle INTEGER_CST
        and generic summands for char* as s * delta, too, folding &a[i]
        CODE x to &a[i CODE x/s].  Use tree_int_cst_equal
        for comparison of steps.  Convert types for index addition.
        (fold_binary): Adjust the callers to always dispatch to
        try_move_mult_to_index.
        * tree-ssa-propagate.c (set_rhs): Avoid setting rhs to
        expr with non-gimple ARRAY_REF offset.


It exposed a latent bug in conjunction with


2005-05-10  Diego Novillo  <dnovillo@redhat.com>

        * tree-optimize.c (init_tree_optimization_passes): Re-organize
        optimization passes to do an initial batch of scalar cleanups.


which I had not in during testing.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[rth at gcc dot gnu dot org]

------- Additional Comments From rth at gcc dot gnu dot org  2005-05-19 01:58 -------
The root cause of this has nothing to do with the type system per-se.
The original program skirts the line of legality, and get_expr_operands
is falling over.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |rth at gcc dot gnu dot org
                   |dot org                     |
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2005-05-12 21:57:33         |2005-05-19 01:58:36
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[rth at gcc dot gnu dot org]

------- Additional Comments From rth at gcc dot gnu dot org  2005-05-19 04:23 -------
Part 1 of a fix: http://gcc.gnu.org/ml/gcc-patches/2005-05/msg01947.html

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-05-19 06:30 -------
Subject: Bug 21541

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	rth@gcc.gnu.org	2005-05-19 06:29:47

Modified files:
	gcc            : ChangeLog ifcvt.c 

Log message:
	PR 21541
	* ifcvt.c (noce_process_if_block): Avoid conversion when the
	memory destination is readonly.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.8858&r2=2.8859
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ifcvt.c.diff?cvsroot=gcc&r1=1.186&r2=1.187



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541

[Bug tree-optimization/21541] [4.1 Regression] gcc.c-torture/execute/20031215-1.c compilation fails

[rth at gcc dot gnu dot org]

------- Additional Comments From rth at gcc dot gnu dot org  2005-05-19 06:31 -------
Fixed.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21541