[Bug target/29111] New: FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[Bug target/29111] New: FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[danglin at gcc dot gnu dot org]

Executing on host: /home/dave/gcc-4.2/objdir/gcc/xgcc
-B/home/dave/gcc-4.2/objdi
r/gcc/ /home/dave/gcc-4.2/gcc/gcc/testsuite/gcc.dg/torture/pr26565.c   -O0  
-fn
o-show-column  -lm   -o ./pr26565.exe    (timeout = 300)
PASS: gcc.dg/torture/pr26565.c  -O0  (test for excess errors)
Setting LD_LIBRARY_PATH to
:/home/dave/gcc-4.2/objdir/gcc::/home/dave/gcc-4.2/ob
jdir/gcc:/home/dave/gcc-4.2/objdir/hppa-linux/libstdc++-v3/.libs:/home/dave/gcc-
4.2/objdir/hppa-linux/libmudflap/.libs:/home/dave/gcc-4.2/objdir/hppa-linux/libs
sp/.libs:/home/dave/gcc-4.2/objdir/hppa-linux/libgomp/.libs:/home/dave/gcc-4.2/o
bjdir/./gcc:/home/dave/gcc-4.2/objdir/./prev-gcc
FAIL: gcc.dg/torture/pr26565.c  -O0  execution test

(gdb) r
Starting program: /home/dave/gcc-4.2/objdir/gcc/testsuite/gcc/pr26565.xg

Program received signal SIGBUS, Bus error.
0x00010474 in send_probe (outdata=0x20a28, tp=0x20a24)
    at /home/dave/gcc-4.2/gcc/gcc/testsuite/gcc.dg/torture/pr26565.c:20
20          memcpy(&outdata->tv, tp, sizeof outdata->tv);
(gdb) bt
#0  0x00010474 in send_probe (outdata=0x20a28, tp=0x20a24)
    at /home/dave/gcc-4.2/gcc/gcc/testsuite/gcc.dg/torture/pr26565.c:20
#1  0x000104ac in main ()
    at /home/dave/gcc-4.2/gcc/gcc/testsuite/gcc.dg/torture/pr26565.c:28
(gdb) p/x $pc
$1 = 0x10474
(gdb) disass send_probe
Dump of assembler code for function send_probe:
0x00010450 <send_probe+0>:      copy r3,r1
0x00010454 <send_probe+4>:      copy sp,r3
0x00010458 <send_probe+8>:      stw,ma r1,40(sp)
0x0001045c <send_probe+12>:     stw r26,-24(r3)
0x00010460 <send_probe+16>:     stw r25,-28(r3)
0x00010464 <send_probe+20>:     ldw -24(r3),ret0
0x00010468 <send_probe+24>:     ldo 5(ret0),r19
0x0001046c <send_probe+28>:     ldw -28(r3),ret0
0x00010470 <send_probe+32>:     ldw 0(ret0),ret0
0x00010474 <send_probe+36>:     stw ret0,0(r19)
0x00010478 <send_probe+40>:     ldo 40(r3),sp
0x0001047c <send_probe+44>:     ldw,mb -40(sp),r3
0x00010480 <send_probe+48>:     bv,n r0(rp)
End of assembler dump.
(gdb) p/x $r19
$2 = 0x20a2d

The "stw" at 0x00010474 is unaligned.  I noticed this because I
disabled the default unaligned fixup support in the kernel.

This is a regression from 3.4 which called memcpy.  All 4.x branches
fail.  The test doesn't fail under hpux because memcpy is called.


-- 
           Summary: FAIL: gcc.dg/torture/pr26565.c  -O0  execution test
           Product: gcc
           Version: 4.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: danglin at gcc dot gnu dot org
 GCC build triplet: hppa-unknown-linux-gnu
  GCC host triplet: hppa-unknown-linux-gnu
GCC target triplet: hppa-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111

[Bug target/29111] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[danglin at gcc dot gnu dot org]

------- Comment #1 from danglin at gcc dot gnu dot org  2006-09-16 18:03 -------
The code is wrong in the initial RTL expansion:

(insn 10 9 0 (set (reg:SI 94 [ D.1473 ])
        (plus:SI (reg/f:SI 95)
            (const_int 5 [0x5]))) -1 (nil)
    (nil))

;; memcpy (D.1473, tp, 4)
(insn 11 10 12 (set (reg/f:SI 96)
        (mem/f/c/i:SI (plus:SI (reg/f:SI 89 virtual-incoming-args)
                (const_int -8 [0xfffffff8])) [0 tp+0 S4 A32])) -1 (nil)
    (nil))

(insn 12 11 13 (set (reg:SI 97)
        (mem/s:SI (reg/f:SI 96) [0 S4 A32])) -1 (nil)
    (nil))

(insn 13 12 0 (set (mem/s:SI (reg:SI 94 [ D.1473 ]) [0 S4 A32])
        (reg:SI 97)) -1 (nil)
    (nil))


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111

[Bug target/29111] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[danglin at gcc dot gnu dot org]

------- Comment #2 from danglin at gcc dot gnu dot org  2006-09-16 19:18 -------
(gdb) step
get_pointer_alignment (exp=0x40013108, max_align=64)
    at ../../gcc/gcc/builtins.c:236
236       if (! POINTER_TYPE_P (TREE_TYPE (exp)))
(gdb) p debug_tree (exp)
 <var_decl 0x40013108 D.1473
    type <pointer_type 0x40089c00
        type <record_type 0x40089960 timeval type_0 SI
            size <integer_cst 0x400092b8 constant invariant 32>
            unit size <integer_cst 0x40009048 constant invariant 4>
            align 32 symtab 0 alias set -1 fields <field_decl 0x400899c0
tv_sec>
            pointer_to_this <pointer_type 0x40089c00> chain <type_decl
0x4001dc98 D.1460>>
        unsigned SI size <integer_cst 0x400092b8 32> unit size <integer_cst
0x40009048 4>
        align 32 symtab 0 alias set -1>
    used unsigned ignored SI file
/home/dave/gcc-4.2/gcc/gcc/testsuite/gcc.dg/torture/pr26565.c line 20 size
<integer_cst 0x400092b8 32> unit size <integer_cst 0x40009048 4>
    align 32 context <function_decl 0x400889a0 send_probe>
    (reg:SI 94 [ D.1473 ])>
$7 = void

get_pointer_alignment returns an alignment of 32 which appears wrong
because of the packed attribute:

struct outdata {
    long align;
    char seq;
    struct timeval tv __attribute__((packed));
};


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111

[Bug middle-end/29111] [4.2 Regression] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|target                      |middle-end
      Known to work|                            |3.4.0
            Summary|FAIL:                       |[4.2 Regression] FAIL:
                   |gcc.dg/torture/pr26565.c  - |gcc.dg/torture/pr26565.c  -
                   |O0  execution test          |O0  execution test
   Target Milestone|---                         |4.2.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111

[Bug middle-end/29111] [4.2 Regression] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[danglin at gcc dot gnu dot org]

------- Comment #3 from danglin at gcc dot gnu dot org  2006-09-17 20:05 -------
There's a question in my mind as to whether we should always be
emitting a library call:

  /* When not optimizing, generate calls to library functions for a certain
     set of builtins.  */
  if (!optimize
      && !called_as_built_in (fndecl)
      && DECL_ASSEMBLER_NAME_SET_P (fndecl)
      && fcode != BUILT_IN_ALLOCA)
    return expand_call (exp, target, ignore);

The setting of DECL_ASSEMBLER_NAME doesn't seem to be entirely reliable.
We have in builtin_function:

  if (library_name)
    SET_DECL_ASSEMBLER_NAME (decl, get_identifier (library_name));

library_name is set for __builtin_memcpy but not for memcpy.

The difference in behavior between the 32-bit hpux targets and the
32-bit linux target arises because the hpux target defines ASM_OUTPUT_EXTERNAL.
So, the assembler name for memcpy gets set indirectly by assemble_external_real
and this causes expand_builtin to use the library routine.  Is this the
intended behavior?

Ok, back to looking at the linux SEGV.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111

[Bug middle-end/29111] [4.2 Regression] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[danglin at gcc dot gnu dot org]

------- Comment #4 from danglin at gcc dot gnu dot org  2006-09-17 20:21 -------
Hmmm, seems this is a known issue that never got fixed.  See this thread:
http://gcc.gnu.org/ml/gcc-patches/2006-03/msg01558.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111

[Bug middle-end/29111] [4.2 Regression] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[ebotcazou at gcc dot gnu dot org]

------- Comment #5 from ebotcazou at gcc dot gnu dot org  2006-09-18 11:36 -------
> Hmmm, seems this is a known issue that never got fixed.  See this thread:
> http://gcc.gnu.org/ml/gcc-patches/2006-03/msg01558.html

And this one: http://gcc.gnu.org/ml/gcc-patches/2006-03/msg01569.html


-- 

ebotcazou at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111

[Bug middle-end/29111] [4.2 Regression] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[dave at hiauly1 dot hia dot nrc dot ca]

------- Comment #6 from dave at hiauly1 dot hia dot nrc dot ca  2006-09-18 13:39 -------
Subject: Re:  [4.2 Regression] FAIL: gcc.dg/torture/pr26565.c  -O0  execution
test

> And this one: http://gcc.gnu.org/ml/gcc-patches/2006-03/msg01569.html

Actually, I was thinking that the library name should be set by
def_builtin_1 if fallback_p is set for the builtin.

Also, based on comments in the thread:

Index: builtins.c
===================================================================
--- builtins.c  (revision 117005)
+++ builtins.c  (working copy)
@@ -233,6 +233,11 @@
 {
   unsigned int align, inner;

+  /* TER is not run at -O0, so our representation of alignment
+     information and its propagation is non-existant.  */
+  if (!optimize)
+    return 0;
+
   if (! POINTER_TYPE_P (TREE_TYPE (exp)))
     return 0;

Dave


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111

[Bug middle-end/29111] [4.2 Regression] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[ebotcazou at gcc dot gnu dot org]

------- Comment #7 from ebotcazou at gcc dot gnu dot org  2006-09-18 13:48 -------
> +  /* TER is not run at -O0, so our representation of alignment
> +     information and its propagation is non-existant.  */
> +  if (!optimize)
> +    return 0;

Perhaps

  if (!flag_tree_ter)
    return 0;

would be more accurate?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111

Re: [Bug middle-end/29111] [4.2 Regression] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[Andrew Pinski]

On Mon, 2006-09-18 at 13:48 +0000, ebotcazou at gcc dot gnu dot org
wrote:
> 
> ------- Comment #7 from ebotcazou at gcc dot gnu dot org  2006-09-18 13:48 -------
> > +  /* TER is not run at -O0, so our representation of alignment
> > +     information and its propagation is non-existant.  */
> > +  if (!optimize)
> > +    return 0;
> 
> Perhaps
> 
>   if (!flag_tree_ter)
>     return 0;
> 
> would be more accurate?
Well -f{no-,}tree-ter can be still passed and you will get it wrong,
what about:
if (!optimize || !flag_tree_ter)
  return 0;

Thanks,
Andrew Pinski

[Bug middle-end/29111] [4.2 Regression] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[pinskia at physics dot uc dot edu]

------- Comment #8 from pinskia at physics dot uc dot edu  2006-09-18 17:19 -------
Subject: Re:  [4.2 Regression] FAIL:
        gcc.dg/torture/pr26565.c  -O0  execution test

On Mon, 2006-09-18 at 13:48 +0000, ebotcazou at gcc dot gnu dot org
wrote:
> 
> ------- Comment #7 from ebotcazou at gcc dot gnu dot org  2006-09-18 13:48 -------
> > +  /* TER is not run at -O0, so our representation of alignment
> > +     information and its propagation is non-existant.  */
> > +  if (!optimize)
> > +    return 0;
> 
> Perhaps
> 
>   if (!flag_tree_ter)
>     return 0;
> 
> would be more accurate?
Well -f{no-,}tree-ter can be still passed and you will get it wrong,
what about:
if (!optimize || !flag_tree_ter)
  return 0;

Thanks,
Andrew Pinski


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111

[Bug middle-end/29111] [4.2 Regression] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[mmitchel at gcc dot gnu dot org]

------- Comment #9 from mmitchel at gcc dot gnu dot org  2006-09-21 02:03 -------
PA-RISC GNU/Linux is not a primary platform, so I've marked this P5.  However,
PA-RISC HP-UX is a primary platform, so if this bug manifests there, please set
this back to P3 with an explanatory comment.


-- 

mmitchel at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P5


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111

[Bug middle-end/29111] [4.2 Regression] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[ebotcazou at gcc dot gnu dot org]

------- Comment #10 from ebotcazou at gcc dot gnu dot org  2006-10-14 10:59 -------
> PA-RISC GNU/Linux is not a primary platform, so I've marked this P5.  However,
> PA-RISC HP-UX is a primary platform, so if this bug manifests there, please set
> this back to P3 with an explanatory comment.

Present on SPARC/Solaris.


-- 

ebotcazou at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
           Priority|P5                          |P3
   Last reconfirmed|0000-00-00 00:00:00         |2006-10-14 10:59:32
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111

[Bug middle-end/29111] [4.2 Regression] FAIL: gcc.dg/torture/pr26565.c -O0 execution test

[ebotcazou at gcc dot gnu dot org]

-- 

ebotcazou at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|ebotcazou at gcc dot gnu dot|
                   |org                         |
         AssignedTo|unassigned at gcc dot gnu   |ebotcazou at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED
  GCC build triplet|hppa-unknown-linux-gnu      |
   GCC host triplet|hppa-unknown-linux-gnu      |
 GCC target triplet|hppa-unknown-linux-gnu      |hppa-*-*, sparc-*-*
   Last reconfirmed|2006-10-14 10:59:32         |2006-10-14 11:00:23
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29111