[Bug c/18666] New: Conversion of floating point into bit-fields

[Bug c/18666] New: Conversion of floating point into bit-fields

[jakub at gcc dot gnu dot org]

Is
/* { dg-do run } */
/* { dg-options "-std=c99 -pedantic-errors } */

struct A { unsigned int i : 1; } a;
extern void abort (void);

int
main (void)
{
  a.i = 16.0;
  if (a.i != 0)
    abort ();
  return 0;
}

a valid test or not?  This worked with 3.4.x and earlier, but doesn't any
longer.  The question is mainly if the type of a.i for the 6.3.1.4/1 purposes
is unsigned int (in this case it would be well-defined, 16 is representable
in unsigned int and storing 16 into unsigned int i : 1 bitfield is defined),
or if the type is integer type with precision 1.

-- 
           Summary: Conversion of floating point into bit-fields
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: jakub at gcc dot gnu dot org
                CC: aoliva at redhat dot com,gcc-bugs at gcc dot gnu dot
                    org,joseph at codesourcery dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18666

[Bug c/18666] Conversion of floating point into bit-fields

[joseph at codesourcery dot com]

------- Additional Comments From joseph at codesourcery dot com  2004-11-25 00:52 -------
Subject: Re:  New: Conversion of floating point into bit-fields

On Thu, 25 Nov 2004, jakub at gcc dot gnu dot org wrote:

> a valid test or not?  This worked with 3.4.x and earlier, but doesn't any
> longer.  The question is mainly if the type of a.i for the 6.3.1.4/1 purposes
> is unsigned int (in this case it would be well-defined, 16 is representable
> in unsigned int and storing 16 into unsigned int i : 1 bitfield is defined),
> or if the type is integer type with precision 1.

There are at least three DRs affirming that the type is unsigned:1, i.e. a 
type with precision 1.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18666

[Bug c/18666] [4.0 Regression] Conversion of floating point into bit-fields

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-12-02 13:41 -------
Reading comment #1, seems to me that the test is correct but still don't know for sure since it worked 
differently on 3.4 and before, I am marking as a regression.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
            Summary|Conversion of floating point|[4.0 Regression] Conversion
                   |into bit-fields             |of floating point into bit-
                   |                            |fields
   Target Milestone|---                         |4.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18666

[Bug c/18666] [4.0 Regression] Conversion of floating point into bit-fields

[aoliva at gcc dot gnu dot org]

------- Additional Comments From aoliva at gcc dot gnu dot org  2004-12-23 20:25 -------
http://www.open-std.org/jtc1/sc22/wg14/www/docs/dr_120.html seems to be pretty
clear on this matter: the type of the bit-field is informally unsigned int :1,
so the conversion invokes undefined behavior per 6.3.1.4/#1.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18666