[Bug middle-end/26565] New: Unaligned accesses with __attribute__(packed) and memcpy

[Bug middle-end/26565] New: Unaligned accesses with __attribute__(packed) and memcpy

[falk at debian dot org]

gcc version 4.2.0 20060304 (experimental)

falk@juist:/tmp% cat test.c
void *memcpy(void *dest, const void *src, __SIZE_TYPE__ n);

struct timeval {
    long tv_sec;
};

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

void send_probe(struct outdata *outdata, struct timeval *tp) {
    memcpy(&outdata->tv, tp, sizeof outdata->tv);
}

falk@juist:/tmp% gcc-3.4 -c -O3 test.c && objdump -dr test.o 
0000000000000000 <send_probe>:
   0:   00 00 51 a6     ldq     a2,0(a1)
   4:   08 00 10 2d     ldq_u   t7,8(a0)
   8:   01 00 70 20     lda     t2,1(a0)
   c:   01 00 f0 2c     ldq_u   t6,1(a0)
  10:   e5 0e 43 4a     insqh   a2,t2,t4
  14:   64 07 43 4a     insql   a2,t2,t3
  18:   46 0e 03 49     mskqh   t7,t2,t5
  1c:   40 06 e3 48     mskql   t6,t2,v0
  20:   01 04 c5 44     or      t5,t4,t0
  24:   08 00 30 3c     stq_u   t0,8(a0)
  28:   02 04 04 44     or      v0,t3,t1
  2c:   01 00 50 3c     stq_u   t1,1(a0)
  30:   01 80 fa 6b     ret

falk@juist:/tmp% gcc -c -O3 test.c && objdump -dr test.o     
0000000000000000 <send_probe>:
   0:   00 00 31 a4     ldq     t0,0(a1)
   4:   01 00 30 b4     stq     t0,1(a0)
   8:   01 80 fa 6b     ret

So 3.4 accounts for an unaligned access, but mainline doesn't.
This has been broken since 4.0. Just reading outdata->tv.tv_sec works fine.


-- 
           Summary: Unaligned accesses with __attribute__(packed) and memcpy
           Product: gcc
           Version: 4.2.0
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: falk at debian dot org
 GCC build triplet: alphaev68-linux-gnu
  GCC host triplet: alphaev68-linux-gnu
GCC target triplet: alphaev68-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] Unaligned accesses with __attribute__(packed) and memcpy

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2006-03-05 14:22 -------
Hmm:
&outdata->tv

isn't that semi wrong as the pointer just lost the alignment information?
A workaround is to do:
    memcpy((void*)&outdata->tv, tp, sizeof outdata->tv);


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] Unaligned accesses with __attribute__(packed) and memcpy

[falk at debian dot org]

------- Comment #2 from falk at debian dot org  2006-03-05 14:40 -------
(In reply to comment #1)
> Hmm:
> &outdata->tv
> 
> isn't that semi wrong as the pointer just lost the alignment information?

I'm not sure what you mean.

> A workaround is to do:
>     memcpy((void*)&outdata->tv, tp, sizeof outdata->tv);

That doesn't change anything.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] Unaligned accesses with __attribute__(packed) and memcpy

[rguenth at gcc dot gnu dot org]

------- Comment #3 from rguenth at gcc dot gnu dot org  2006-03-06 10:58 -------
We indeed lose alignment information of outdata->tv.  We start expanding

 memcpy (&outdataD.1529->tvD.1528, tpD.1530, 4) [tail call]

with

(gdb) print dest_align
$1 = 32

so, builtins.c:get_pointer_alignment returns wrong (non conservative) results.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2006-03-06 10:58:47
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] Unaligned accesses with __attribute__(packed) and memcpy

[rguenth at gcc dot gnu dot org]

------- Comment #4 from rguenth at gcc dot gnu dot org  2006-03-06 11:58 -------
I have a patch.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |rguenth at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2006-03-06 10:58:47         |2006-03-06 11:58:08
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] [4.1/4.2 Regression] Unaligned accesses with __attribute__(packed) and memcpy

[rguenth at gcc dot gnu dot org]

------- Comment #5 from rguenth at gcc dot gnu dot org  2006-03-06 12:17 -------
Works with 3.3.3, 3.4.2, fails with 4.1.0 and 4.2.0 (didn't check 4.0.x, but I
guess it's another tree-ssa fallout)


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|                            |4.1.0 4.2.0
      Known to work|                            |3.3.3 3.4.2
            Summary|Unaligned accesses with     |[4.1/4.2 Regression]
                   |__attribute__(packed) and   |Unaligned accesses with
                   |memcpy                      |__attribute__(packed) and
                   |                            |memcpy


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] [4.1/4.2 Regression] Unaligned accesses with __attribute__(packed) and memcpy

[rguenth at gcc dot gnu dot org]

------- Comment #6 from rguenth at gcc dot gnu dot org  2006-03-06 12:23 -------
It also affects ia64 and s390(x)


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  GCC build triplet|alphaev68-linux-gnu         |
   GCC host triplet|alphaev68-linux-gnu         |
 GCC target triplet|alphaev68-linux-gnu         |alphaev68-linux-gnu, ia64-*-
                   |                            |*, s390-*-*, s390x-*-*


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] [4.1/4.2 Regression] Unaligned accesses with __attribute__(packed) and memcpy

[rguenth at gcc dot gnu dot org]

------- Comment #7 from rguenth at gcc dot gnu dot org  2006-03-06 12:49 -------
A workaround is to do

  memcpy ((char*)outdata + 1, ...);


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] [4.0/4.1/4.2 Regression] Unaligned accesses with __attribute__(packed) and memcpy

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu dot
                   |                            |org
 GCC target triplet|alphaev68-linux-gnu, ia64-*-|STRICT_ALIGNMENT
                   |*, s390-*-*, s390x-*-*      |
            Summary|[4.1/4.2 Regression]        |[4.0/4.1/4.2 Regression]
                   |Unaligned accesses with     |Unaligned accesses with
                   |__attribute__(packed) and   |__attribute__(packed) and
                   |memcpy                      |memcpy
   Target Milestone|---                         |4.0.4


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] [4.0/4.1/4.2 Regression] Unaligned accesses with __attribute__(packed) and memcpy

[patchapp at dberlin dot org]

------- Comment #8 from patchapp at dberlin dot org  2006-03-06 14:55 -------
Subject: Bug number PR middle-end/26565

A patch for this bug has been added to the patch tracker.
The mailing list url for the patch is
http://gcc.gnu.org/ml/gcc-patches/2006-03/msg00324.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] [4.0/4.1 Regression] Unaligned accesses with __attribute__(packed) and memcpy

[rguenth at gcc dot gnu dot org]

------- Comment #10 from rguenth at gcc dot gnu dot org  2006-03-10 16:44 -------
Fixed on the mainline.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|3.3.3 3.4.2                 |3.3.3 3.4.2 4.2.0
            Summary|[4.0/4.1/4.2 Regression]    |[4.0/4.1 Regression]
                   |Unaligned accesses with     |Unaligned accesses with
                   |__attribute__(packed) and   |__attribute__(packed) and
                   |memcpy                      |memcpy


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] [4.0/4.1/4.2 Regression] Unaligned accesses with __attribute__(packed) and memcpy

[rguenth at gcc dot gnu dot org]

------- Comment #9 from rguenth at gcc dot gnu dot org  2006-03-10 16:44 -------
Subject: Bug 26565

Author: rguenth
Date: Fri Mar 10 16:44:01 2006
New Revision: 111934

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=111934
Log:
2006-03-10  Richard Guenther  <rguenther@suse.de>

        PR middle-end/26565
        * builtins.c (get_pointer_alignment): Handle component
        references for field alignment.

        * gcc.dg/torture/pr26565.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/torture/pr26565.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/builtins.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] [4.0/4.1 Regression] Unaligned accesses with __attribute__(packed) and memcpy

[rguenth at gcc dot gnu dot org]

------- Comment #11 from rguenth at gcc dot gnu dot org  2006-03-11 20:17 -------
*** Bug 26639 has been marked as a duplicate of this bug. ***


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hjl at lucon dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] [4.0/4.1 Regression] Unaligned accesses with __attribute__(packed) and memcpy

[patchapp at dberlin dot org]

------- Comment #12 from patchapp at dberlin dot org  2006-04-05 14:00 -------
Subject: Bug number PR26565

A patch for this bug has been added to the patch tracker.
The mailing list url for the patch is
http://gcc.gnu.org/ml/gcc-patches/2006-04/msg00192.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] [4.0/4.1 Regression] Unaligned accesses with __attribute__(packed) and memcpy

[rguenth at gcc dot gnu dot org]

------- Comment #13 from rguenth at gcc dot gnu dot org  2006-05-01 11:36 -------
Subject: Bug 26565

Author: rguenth
Date: Mon May  1 11:36:27 2006
New Revision: 113410

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=113410
Log:
2006-05-01  Richard Guenther  <rguenther@suse.de>

        PR middle-end/26565
        * builtins.c (get_pointer_alignment): Handle component
        references for field alignment.

Modified:
    branches/gcc-4_1-branch/gcc/ChangeLog
    branches/gcc-4_1-branch/gcc/builtins.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] [4.0/4.1 Regression] Unaligned accesses with __attribute__(packed) and memcpy

[rguenth at gcc dot gnu dot org]

------- Comment #14 from rguenth at gcc dot gnu dot org  2006-05-01 12:04 -------
Subject: Bug 26565

Author: rguenth
Date: Mon May  1 12:04:13 2006
New Revision: 113411

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=113411
Log:
2006-05-01  Richard Guenther  <rguenther@suse.de>

        PR middle-end/26565
        * builtins.c (get_pointer_alignment): Handle component
        references for field alignment.

Modified:
    branches/gcc-4_0-branch/gcc/ChangeLog
    branches/gcc-4_0-branch/gcc/builtins.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565

[Bug middle-end/26565] [4.0/4.1 Regression] Unaligned accesses with __attribute__(packed) and memcpy

[rguenth at gcc dot gnu dot org]

------- Comment #15 from rguenth at gcc dot gnu dot org  2006-05-01 12:04 -------
Fixed.


-- 

rguenth at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26565