[Bug target/18665] New: -ftrapv borks up simple integer arithmetic

[Bug target/18665] New: -ftrapv borks up simple integer arithmetic

[skunk at iskunk dot org]

A large program was failing mysteriously on our new Opteron box. Went through it
with a fine-toothed debugger, until I came to the line

    k -= i;  /* <k> and <i> are ints */

Before this line: k == 2, i == 1
 After this line: k == -14, i == 1

Compiling without -ftrapv makes this weirdness go away.

I will attach a tarball containing two .c files and a makefile, basically the
aforementioned big program with all irrelevant code ripped out---just enough to
reproduce the problem.

This bug first came up with 3.4.2, as shipped with Fedora Core 3 (x86-64). I
built 3.4.3 with the below configuration to confirm that the bug is still present.

Output from "gcc -v"
====================
Reading specs from
/mnt/freeport/opt/gcc-3.4.3--linux64/bin/../lib/gcc/x86_64-unknown-linux-gnu/3.4.3/specs
Configured with: /mnt/freeport/src/gcc/current/configure
--prefix=/mnt/freeport/arch/linux64 --disable-dependency-tracking
--disable-shared --disable-nls --with-libiconv-prefix=/mnt/freeport/arch/linux64
--disable-multilib --with-arch=opteron
Thread model: posix
gcc version 3.4.3

-- 
           Summary: -ftrapv borks up simple integer arithmetic
           Product: gcc
           Version: 3.4.3
            Status: UNCONFIRMED
          Severity: critical
          Priority: P2
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: skunk at iskunk dot org
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: x86_64-unknown-linux-gnu
  GCC host triplet: x86_64-unknown-linux-gnu
GCC target triplet: x86_64-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] -ftrapv borks up simple integer arithmetic

[skunk at iskunk dot org]

------- Additional Comments From skunk at iskunk dot org  2004-11-24 22:32 -------
Created an attachment (id=7600)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=7600&action=view)
Test case exhibiting subject bug

The bug occurs in the lib.c file. I have added some extra printf()s and an
assert() to make it stand out. The makefile's default target will compile the
program and execute it in the manner that raises the bug.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] -ftrapv borks up simple integer arithmetic

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|critical                    |normal
           Keywords|                            |wrong-code


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] -ftrapv borks up simple integer arithmetic

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2004-11-28 00:07 -------
I can reproduce this on x86_64-unknown-linux-gnu with 
"GNU C version 4.0.0 20041127 (experimental) (x86_64-unknown-linux-gnu)" 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2004-11-28 00:07:18
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] -ftrapv borks up simple integer arithmetic

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2004-11-28 00:22 -------
Much simpler test case: 
 
$ cat t.c 
extern int printf (__const char *__restrict __format, ...); 
 
int __attribute__ ((__noinline__)) 
foo (int a, int b) 
{ 
  return a - b; 
} 
 
int 
main (void) 
{ 
  printf ("%d\n", foo (2,3)); 
  return 0; 
} 
$ ../xgcc -B.. t.c 
$ ./a.out 
-1 
$ ../xgcc -B.. t.c -ftrapv 
$ ./a.out 
1073748906 
$ 
 
 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] -ftrapv borks up simple integer arithmetic

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2004-11-28 00:26 -------
Honza, looks like we're not setting up the libcall right in the -ftrapv 
case.  Can you with all your RTL and amd64-fu give this a look? 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jh at suse dot cz


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] [4.0 Regression] -ftrapv borks up simple integer arithmetic

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2004-11-28 00:31 -------
Works with SuSE 3.3.3. 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|                            |4.0.0
      Known to work|                            |3.3.3 3.3.1
            Summary|-ftrapv borks up simple     |[4.0 Regression] -ftrapv
                   |integer arithmetic          |borks up simple integer
                   |                            |arithmetic


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.0 Regression] -ftrapv    |[3.4/4.0 Regression] -ftrapv
                   |borks up simple integer     |borks up simple integer
                   |arithmetic                  |arithmetic
   Target Milestone|---                         |3.4.4


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2004-11-28 00:34 -------
Here's the difference between what gcc 3.3.3 produces and what 
gcc 4.0.0 manages to make of it (first 3.3.3, second 4.0.0): 
 
16,18c16,20 
<       movl    -8(%rbp), %esi 
<       movl    -4(%rbp), %edi 
<       call    __subvsi3 
--- 
>       movl    -4(%rbp), %eax 
>       movslq  %eax,%rdi 
>       movl    -8(%rbp), %eax 
>       movslq  %eax,%rsi 
>       call    __subvdi3 
 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2004-11-28 12:59 -------
This may be related to (or in fact be the same as) Bug 17603. 
 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-12-06 18:45 -------
The (useless?) mode promotion from SImode to DImode comes from:
http://gcc.gnu.org/ml/gcc-patches/2003-02/msg00625.html


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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-12-06 18:47 -------
Richard, why do we not register optabs for modes smaller than word_mode now? 
This appears to hurt 64-bit targets.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rth at redhat dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[rth at gcc dot gnu dot org]

------- Additional Comments From rth at gcc dot gnu dot org  2004-12-06 20:51 -------
Look at the routines we implement in libgcc and ask again.  We never did
implement sub-word routines.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-12-07 06:59 -------
> Look at the routines we implement in libgcc and ask again.  We never did
> implement sub-word routines.

Not sure what you meant with "never", but take a look at comment #6...


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[schlie at comcast dot net]

------- Additional Comments From schlie at comcast dot net  2004-12-07 14:36 -------
Subject:  [3.4/4.0 Regression] -ftrapv borks up simple
 integer arithmetic

> ------- Additional Comment  #8 From  Eric Botcazou  2004-12-06 18:45
> The (useless?) mode promotion from SImode to DImode comes from:
> http://gcc.gnu.org/ml/gcc-patches/2003-02/msg00625.html
>

The c4X inspired logic seems basically wrong; built-in function type sizes
should likely be determined by the mode/size (QI/HI/SI/etc.) specified by
the target, not words-per-"anything" (where 8-bit targets may define a short
as being 1 word-wide, and long as 4-words-wide; 64-bit targets may specify a
short as being 1/4-words-wide, and long as 1-words-wide).





-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-12-10 07:35 -------
I'll try to come up with something.


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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-12-10 09:53 -------
This fails on all 64-bit targets.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
  GCC build triplet|x86_64-unknown-linux-gnu    |*-*-*
   GCC host triplet|x86_64-unknown-linux-gnu    |*-*-*
 GCC target triplet|x86_64-unknown-linux-gnu    |*-*-*
      Known to fail|4.0.0                       |3.4.3 4.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug target/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-12-10 10:11 -------
OK, I understand what happens: the 5 symbols __absvdi2, __addvdi3, __subvdi3,
__mulvdi3 and __negvdi2 are not #defined in libgcc2.h, hence not tweaked based
on the word size.  They are, therefore, double-word symbols everywhere, which
doesn't work on 64-bit targets.

The fix is straightforward, however there is a hitch: on 64-bit targets, libgcc
will contain 5 new symbols (__absvti2, __addvti3, __mulvti3, __negvti2,
__subvti3)  and will not contain (__absvsi2, __addvsi3, __mulvsi3, __negvsi2,
__subvsi3) anymore.  While the former is OK, the latter is probably not so we
might need to provide stubs that will redirect them to the equivalent mode-sized
routines for backwards compatibility.

Richard, does that sound plausible?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[rth at gcc dot gnu dot org]

------- Additional Comments From rth at gcc dot gnu dot org  2004-12-10 18:58 -------
I'm not sure redirecting __addvsi3 to __addvdi3 is correct.  If we provide that
symbol at all, I think it should do what it advertises.

Just so I'm clear, are we invoking __addvsi3 right now, or is it just that
__addvdi3 has the wrong implementation?

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
  GCC build triplet|*-*-*                       |
   GCC host triplet|*-*-*                       |
 GCC target triplet|*-*-*                       |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-12-10 19:24 -------
> I'm not sure redirecting __addvsi3 to __addvdi3 is correct.  If we provide that
> symbol at all, I think it should do what it advertises.

But what we export now as __addvsi3 will be exported as __addvdi3 after the fix.

> Just so I'm clear, are we invoking __addvsi3 right now, or is it just that
> __addvdi3 has the wrong implementation?

The latter.  __addvdi3 uses double words everywhere, so is actually __addvti3 on
64-bit targets.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[rth at gcc dot gnu dot org]

------- Additional Comments From rth at gcc dot gnu dot org  2004-12-10 19:40 -------
If addvsi3 actually performs addvdi3, that will be most surprising to
everyone.  I think that's a very bad idea.  Since we are not invoking
addvsi3, and I suspect it's impossible to convince the compiler to do
so, I think there's no reason to provide the backward compatibility
function at all; the symbol was unused and unusable.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-12-10 20:04 -------
> If addvsi3 actually performs addvdi3, that will be most surprising to
> everyone.  I think that's a very bad idea.  Since we are not invoking addvsi3, 
> and I suspect it's impossible to convince the compiler to do so,

After your patch, that's indeed the case.

> I think there's no reason to provide the backward compatibility function at
> all; the symbol was unused and unusable.

Except that, before your patch, the 64-bit compiler was emitting addvsi3 (see
comment #6 by Steven).  So I think we need to provide the symbol.  The question
would then be: should we provide real SImode addition or word_mode addition as
currently?  Strict backwards compatibility would imply the latter.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[rth at gcc dot gnu dot org]

------- Additional Comments From rth at gcc dot gnu dot org  2004-12-10 20:17 -------
Since the symbol exists in 3.3 (which I'd forgotten), we have to keep it.
But I don't see any reason we can't fix the bug in its implementation.
So let's do proper simode arithmetic.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-12-11 09:39 -------
> Since the symbol exists in 3.3 (which I'd forgotten), we have to keep it.
> But I don't see any reason we can't fix the bug in its implementation.
> So let's do proper simode arithmetic.

Agreed.  It turns out that the same problem exists for DImode arithmetic on
targets for which the word_mode is smaller than SImode: since the di variants
were always emitted, should we still emit them now?  Can the big __mulvdi3
routine contained in libgcc2.c be compiled with real DImode on those targets?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[rth at gcc dot gnu dot org]

------- Additional Comments From rth at gcc dot gnu dot org  2004-12-11 23:12 -------
Subject: Re:  [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

No, we won't have access to DImode on those platforms.  But then,
the user won't have access to DImode inside the compiler either,
so it *should* be impossible for that routine to be invoked.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-12-12 07:19 -------
> No, we won't have access to DImode on those platforms.  But then,
> the user won't have access to DImode inside the compiler either,
> so it *should* be impossible for that routine to be invoked.

Your "should" is a bit worrying, but the alternative would probably be a
compilation failure, which is not really better.  And I presume it's the same
dilemna with SImode for targets whose word_mode is QImode, so I'm not going to
include si variants for them either.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-12-13 17:11 -------
Patch here: <http://gcc.gnu.org/ml/gcc-patches/2004-12/msg00907.html>.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-12-14 14:15 -------
A slight different approach is here : <http://gcc.gnu.org/ml/gcc-patches/2004-12/msg01025.html>. 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-12-14 14:36 -------
> A slight different approach is here :
> <http://gcc.gnu.org/ml/gcc-patches/2004-12/msg01025.html>. 

AFAICS it's the same patch, without the bugfix for subsi3 and the compat stuff.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-12-15 12:31 -------
Subject: Bug 18665

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	ebotcazou@gcc.gnu.org	2004-12-15 12:30:47

Modified files:
	gcc            : libgcc2.c libgcc2.h libgcc-std.ver 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.dg: ftrapv-2.c 

Log message:
	PR other/18665
	* libgcc-std.ver (GCC_3.4.4): Inherit from GCC_3.4.2.
	Export __absvti2, __addvti3, __mulvti3, __negvti2 and __subvti3.
	* libgcc2.c (__addvsi3): Rename to __addvSI3.
	New version if COMPAT_SIMODE_TRAPPING_ARITHMETIC.
	(__addvdi3): Rename to __addvDI3.
	(__subvsi3): Rename to __subvSI3.  Use word type for the result.
	New version if COMPAT_SIMODE_TRAPPING_ARITHMETIC.
	(__subvdi3): Rename to __subvDI3.
	(_mulvsi3): Rename to _mulvSI3.
	New version if COMPAT_SIMODE_TRAPPING_ARITHMETIC.
	(_mulvdi3): Rename to _mulvDI3.
	(__negvsi2): Rename to __negvSI2.
	New version if COMPAT_SIMODE_TRAPPING_ARITHMETIC.
	(__negvdi2): Rename to __negvDI2.
	(__absvsi2): Rename to __absvSI2.
	New version if COMPAT_SIMODE_TRAPPING_ARITHMETIC.
	(__absvdi2): Rename to __absvDI2.
	* libgcc2.h (64-bit targets): Define COMPAT_SIMODE_TRAPPING_ARITHMETIC.
	(__absvSI2, __addvSI3, __subvSI3, __mulvSI3, __negvSI2, __absvDI2,
	__addvDI3, __subvDI3, __mulvDI3, __negvDI2): Define to the appropriate
	symbol and declare.
	(__absvsi2, __addvsi3, __subvsi3, __mulvsi3, __negvsi2): Declare if
	COMPAT_SIMODE_TRAPPING_ARITHMETIC.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/libgcc2.c.diff?cvsroot=gcc&r1=1.178&r2=1.179
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/libgcc2.h.diff?cvsroot=gcc&r1=1.32&r2=1.33
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/libgcc-std.ver.diff?cvsroot=gcc&r1=1.26&r2=1.27
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.4759&r2=1.4760
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/ftrapv-2.c.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-12-15 12:35 -------
Subject: Bug 18665

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-3_4-branch
Changes by:	ebotcazou@gcc.gnu.org	2004-12-15 12:34:41

Modified files:
	gcc            : libgcc2.c libgcc2.h libgcc-darwin.ver 
	                 libgcc-std.ver 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.dg: ftrapv-2.c 

Log message:
	PR other/18665
	* libgcc-std.ver (GCC_3.4.4): Inherit from GCC_3.4.2.
	Export __absvti2, __addvti3, __mulvti3, __negvti2 and __subvti3.
	* libgcc-darwin.ver (GCC_3.4.4): Inherit from GCC_3.4.
	Export __absvti2, __addvti3, __mulvti3, __negvti2 and __subvti3.
	* libgcc2.c (__addvsi3): Rename to __addvSI3.
	New version if COMPAT_SIMODE_TRAPPING_ARITHMETIC.
	(__addvdi3): Rename to __addvDI3.
	(__subvsi3): Rename to __subvSI3.  Use word type for the result.
	New version if COMPAT_SIMODE_TRAPPING_ARITHMETIC.
	(__subvdi3): Rename to __subvDI3.
	(_mulvsi3): Rename to _mulvSI3.
	New version if COMPAT_SIMODE_TRAPPING_ARITHMETIC.
	(_mulvdi3): Rename to _mulvDI3.
	(__negvsi2): Rename to __negvSI2.
	New version if COMPAT_SIMODE_TRAPPING_ARITHMETIC.
	(__negvdi2): Rename to __negvDI2.
	(__absvsi2): Rename to __absvSI2.
	New version if COMPAT_SIMODE_TRAPPING_ARITHMETIC.
	(__absvdi2): Rename to __absvDI2.
	* libgcc2.h (64-bit targets): Define COMPAT_SIMODE_TRAPPING_ARITHMETIC.
	(__absvSI2, __addvSI3, __subvSI3, __mulvSI3, __negvSI2, __absvDI2,
	__addvDI3, __subvDI3, __mulvDI3, __negvDI2): Define to the appropriate
	symbol and declare.
	(__absvsi2, __addvsi3, __subvsi3, __mulvsi3, __negvsi2): Declare if
	COMPAT_SIMODE_TRAPPING_ARITHMETIC.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/libgcc2.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.170.6.2&r2=1.170.6.3
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/libgcc2.h.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.28&r2=1.28.4.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/libgcc-darwin.ver.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=2.1.10.1&r2=2.1.10.2
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/libgcc-std.ver.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.23.10.3&r2=1.23.10.4
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.3389.2.324&r2=1.3389.2.325
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/ftrapv-2.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=NONE&r2=1.1.2.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665

[Bug other/18665] [3.4/4.0 Regression] -ftrapv borks up simple integer arithmetic

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-12-15 12:39 -------
See http://gcc.gnu.org/ml/gcc-patches/2004-12/msg00907.html


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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18665