[Bug c/32628] New: bogus integer overflow warning

[Bug c/32628] New: bogus integer overflow warning

[ak at muc dot de]

gcc version 4.3.0 20070704 (experimental)

Following test case derived from the Linux ACPI code warns

toverflow.c: In function 'f':
toverflow.c:8: warning: integer overflow in expression

but the warning is not correct because 0 + ULONG_MAX doesn't really overflow.

typedef void *acpi_handle;
typedef unsigned long long UINT64;
typedef unsigned long long acpi_native_uint;
typedef unsigned char u8;

int f(acpi_handle device)
{
        if (device == ((acpi_handle *) (void *) ((((u8 *) (void *) ((((void
*)0)))) + (acpi_native_uint)((UINT64)(~((UINT64) 0)))))))
                return 1;
        return 0;
}


-- 
           Summary: bogus integer overflow warning
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: ak at muc dot de
 GCC build triplet: x86_64-linux
  GCC host triplet: x86_64-linux
GCC target triplet: x86_64-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32628

[Bug middle-end/32628] [4.3 Regression] bogus integer overflow warning

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2007-07-04 23:34 -------
A simple example is:
void *f(void)
{
  char *a = ((char*)0)+(unsigned long long)(-1);
  return a;
}

This is caused by pointer_plus so mine.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |pinskia at gcc dot gnu dot
                   |dot org                     |org
             Status|UNCONFIRMED                 |ASSIGNED
          Component|c                           |middle-end
     Ever Confirmed|0                           |1
           Keywords|                            |diagnostic
   Last reconfirmed|0000-00-00 00:00:00         |2007-07-04 23:34:35
               date|                            |
            Summary|bogus integer overflow      |[4.3 Regression] bogus
                   |warning                     |integer overflow warning
   Target Milestone|---                         |4.3.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32628

[Bug middle-end/32628] [4.3 Regression] bogus integer overflow warning

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2007-07-05 19:02 -------
This patch works for me:
Index: fold-const.c
===================================================================
--- fold-const.c        (revision 126338)
+++ fold-const.c        (working copy)
@@ -282,14 +282,8 @@
                       HOST_WIDE_INT high, int overflowable,
                       bool overflowed)
 {
-  int sign_extended_type;
   bool overflow;

-  /* Size types *are* sign extended.  */
-  sign_extended_type = (!TYPE_UNSIGNED (type)
-                       || (TREE_CODE (type) == INTEGER_TYPE
-                           && TYPE_IS_SIZETYPE (type)));
-
   overflow = fit_double_type (low, high, &low, &high, type);

   /* If we need to set overflow flags, return a new unshared node.  */
@@ -297,7 +291,7 @@
     {
       if (overflowed
          || overflowable < 0
-         || (overflowable > 0 && sign_extended_type))
+         || (overflowable > 0 && !TYPE_UNSIGNED (type)))
        {
           tree t = make_node (INTEGER_CST);
           TREE_INT_CST_LOW (t) = low;


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32628

[Bug middle-end/32628] [4.3 Regression] bogus integer overflow warning

[ebotcazou at gcc dot gnu dot org]

------- Comment #3 from ebotcazou at gcc dot gnu dot org  2007-07-05 21:57 -------
> This patch works for me:

Well, that sizetypes are sign-extended is assumed in other places too.


-- 

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=32628

[Bug middle-end/32628] [4.3 Regression] bogus integer overflow warning

[pinskia at gcc dot gnu dot org]

------- Comment #4 from pinskia at gcc dot gnu dot org  2007-07-05 22:57 -------
> Well, that sizetypes are sign-extended is assumed in other places too.

Yes this is known but this place is "special".  Though I have not figured out
why we are setting overflow in general here anyways.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32628

[Bug middle-end/32628] [4.3 Regression] bogus integer overflow warning

[pinskia at gcc dot gnu dot org]

------- Comment #5 from pinskia at gcc dot gnu dot org  2007-07-06 00:34 -------
Some more information, I don't think conversion should set the overflow bit any
more.
Roger started doing some of this work with:
http://gcc.gnu.org/ml/gcc-patches/2006-11/msg00812.html

But I don't know what happened after that as Roger disappeared.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32628

[Bug middle-end/32628] [4.3 Regression] bogus integer overflow warning

[pinskia at gcc dot gnu dot org]

------- Comment #6 from pinskia at gcc dot gnu dot org  2007-07-06 04:12 -------
This patch is most likely better:
Index: fold-const.c
===================================================================
--- fold-const.c        (revision 126396)
+++ fold-const.c        (working copy)
@@ -2063,9 +2063,7 @@
      appropriately sign-extended or truncated.  */
   t = force_fit_type_double (type, TREE_INT_CST_LOW (arg1),
                             TREE_INT_CST_HIGH (arg1),
-                            /* Don't set the overflow when
-                               converting a pointer  */
-                            !POINTER_TYPE_P (TREE_TYPE (arg1)),
+                            0,
                             (TREE_INT_CST_HIGH (arg1) < 0
                              && (TYPE_UNSIGNED (type)
                                  < TYPE_UNSIGNED (TREE_TYPE (arg1))))



So don't set overflow for conversion.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32628

[Bug middle-end/32628] [4.3 Regression] bogus integer overflow warning

[pinskia at gcc dot gnu dot org]

------- Comment #7 from pinskia at gcc dot gnu dot org  2007-07-18 15:22 -------
>So don't set overflow for conversion.

When I tested this patch, it caused Ada not to bootstrap so I have to look more
into it.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32628

[Bug middle-end/32628] [4.3 Regression] bogus integer overflow warning

[mmitchel at gcc dot gnu dot org]

-- 

mmitchel at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32628

[Bug middle-end/32628] [4.3 Regression] bogus integer overflow warning

[rguenth at gcc dot gnu dot org]

------- Comment #8 from rguenth at gcc dot gnu dot org  2008-01-15 14:44 -------
Reduced testcase:

int f(char *device)
{
  return device == ((char *)0 + ~0UL);
}

for building the POINTER_PLUS_EXPR we convert the unsigned offset to
sizetype which causes the double-int (-1, 0) to sign-extend to (-1, -1)
signalling overflow (the positive value doesn't fit in sizetype).

In the context of POINTER_PLUS_EXPRs the sizetype offset depending on
evaluation context has an effective range of [-INT_MIN, UINT_MAX].

So in principle we want to not treat the sign/zero extension
that happens when converting from/to sizetype as overflow.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|pinskia at gcc dot gnu dot  |rguenth at gcc dot gnu dot
                   |org                         |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32628

[Bug middle-end/32628] [4.3 Regression] bogus integer overflow warning

[rguenth at gcc dot gnu dot org]

------- Comment #9 from rguenth at gcc dot gnu dot org  2008-01-16 21:52 -------
Subject: Bug 32628

Author: rguenth
Date: Wed Jan 16 21:51:57 2008
New Revision: 131579

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=131579
Log:
2008-01-16  Richard Guenther  <rguenther@suse.de>

        PR middle-end/32628
        * fold-const.c (fold_convert_const_int_from_int): Do not
        set overflow if that occured only because of a sign extension
        change when converting from/to a sizetype with the same
        precision and signedness.

        * gcc.dg/overflow-warn-7.c: New testcase.

Added:
    trunk/gcc/testsuite/gcc.dg/overflow-warn-7.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/fold-const.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32628

[Bug middle-end/32628] [4.3 Regression] bogus integer overflow warning

[rguenth at gcc dot gnu dot org]

------- Comment #10 from rguenth at gcc dot gnu dot org  2008-01-16 21:54 -------
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=32628

[Bug middle-end/32628] [4.3 Regression] bogus integer overflow warning

[ebotcazou at gcc dot gnu dot org]

------- Comment #11 from ebotcazou at gcc dot gnu dot org  2010-04-13 15:48 -------
Subject: Bug 32628

Author: ebotcazou
Date: Tue Apr 13 15:47:38 2010
New Revision: 158274

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=158274
Log:
        PR middle-end/32628
        * c-common.c (pointer_int_sum): Disregard overflow that occured only
        because of sign-extension change when converting to sizetype here...
        * fold-const.c (fold_convert_const_int_from_int): ...and not here.

        * fold-const.c (fold_binary_op_with_conditional_arg): Do not restrict
        the folding to constants.  Remove redundant final conversion.
        (fold_binary) <associate>: Do not associate if the re-association of
        constants alone overflows.
        (fold_binary) <FLOOR_MOD_EXPR>: Move transformation into BIT_AND_EXPR
        to the end of the list.
        (multiple_of_p) <COND_EXPR>: New case.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/c-common.c
    trunk/gcc/fold-const.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32628