Re: c/4258: gcc-2.95.2 produces wrong code

Re: c/4258: gcc-2.95.2 produces wrong code

[Zack Weinberg]

The following reply was made to PR c/4258; it has been noted by GNATS.

From: Zack Weinberg <zack@codesourcery.com>
To: Ingo Rohloff <rohloff@in.tum.de>
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: c/4258: gcc-2.95.2 produces wrong code
Date: Mon, 10 Sep 2001 09:23:18 -0700

 On Mon, Sep 10, 2001 at 11:54:38AM +0200, Ingo Rohloff wrote:
 > > However, additional tests can never hurt.  If you
 > > rewrite your program in a form that can go into the test suite, we'll
 > > add it.  (In a nutshell: the program must produce no output, and
 > > should return 0 from main if it executed correctly, call abort if it
 > > didn't.)
 > 
 > Ok, i changed it that way. (The code itself is also simpler now.)
 
 Thanks.  That file is now c-torture/execute/20010910-1.c.  Since you
 reported this as a bug in 2.95, I put it onto all three active
 branches (2.95, 3.0, mainline).
 
 zw

Re: c/4258: gcc-2.95.2 produces wrong code

[rodrigc]

The following reply was made to PR c/4258; it has been noted by GNATS.

From: rodrigc@gcc.gnu.org
To: gcc-bugs@gcc.gnu.org, gcc-gnats@gcc.gnu.org, gcc-prs@gcc.gnu.org,
  nobody@gcc.gnu.org, rohloff@in.tum.de
Cc:  
Subject: Re: c/4258: gcc-2.95.2 produces wrong code
Date: 2 Dec 2001 04:05:32 -0000

 Synopsis: gcc-2.95.2 produces wrong code
 
 State-Changed-From-To: open->closed
 State-Changed-By: rodrigc
 State-Changed-When: Sat Dec  1 20:05:31 2001
 State-Changed-Why:
     Fixed in gcc 2.95.3 and added to testsuite as
     c-torture/execute/20010910-1.c
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&pr=4258&database=gcc

Re: c/4258: gcc-2.95.2 produces wrong code

[rodrigc]

Synopsis: gcc-2.95.2 produces wrong code

State-Changed-From-To: open->closed
State-Changed-By: rodrigc
State-Changed-When: Sat Dec  1 20:05:31 2001
State-Changed-Why:
    Fixed in gcc 2.95.3 and added to testsuite as
    c-torture/execute/20010910-1.c

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&pr=4258&database=gcc

Re: c/4258: gcc-2.95.2 produces wrong code

[Zack Weinberg]

The following reply was made to PR c/4258; it has been noted by GNATS.

From: Zack Weinberg <zack@codesourcery.com>
To: rohloff@in.tum.de
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: c/4258: gcc-2.95.2 produces wrong code
Date: Fri, 7 Sep 2001 09:27:47 -0700

 On Fri, Sep 07, 2001 at 03:42:53PM -0000, rohloff@in.tum.de wrote:
 > Compiling the attached code with -O2 produces a 
 > binary which skips a loop iteration.
 
 I have compiled this code with -O1 and -O2 with 
 "gcc-2.95.4 20010902 (Debian prerelease)" and get identical output.
 It seems, therefore, that the bug has been fixed in 2.95.3 or
 2.95.4-to-be.
 
 I cannot be certain, however, because you do not say what the program
 should output.  This is what I get:
 
 BFFFFA18
 BFFFFA1C
 BFFFFA20
 BFFFFA24
 BFFFFA28
 
 I don't know if that is correct or not.
 
 zw

c/4258: gcc-2.95.2 produces wrong code

[rohloff]

>Number:         4258
>Category:       c
>Synopsis:       gcc-2.95.2 produces wrong code
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          wrong-code
>Submitter-Id:   net
>Arrival-Date:   Fri Sep 07 08:46:00 PDT 2001
>Closed-Date:
>Last-Modified:
>Originator:     Ingo Rohloff
>Release:        gcc-2.95.2
>Organization:
>Environment:
Linux 2.4.9 i686
>Description:
Compiling the attached code with -O2 produces a 
binary which skips a loop iteration.
>How-To-Repeat:
Compile the following code with -O2 and execute
the resulting binary.
------------------------------------------------
struct epic_rx_desc {
  unsigned int next;
};

struct epic_private {
  struct epic_rx_desc *rx_ring;
  unsigned int rx_ring_dma;
  unsigned char* rx_skbuff[32];
};

static void epic_init_ring(struct epic_private *ep)
{
  int i;

  for (i = 0; i < 32; i++) {
    ep->rx_ring[i].next = ep->rx_ring_dma +
              (i+1)*sizeof(struct epic_rx_desc);
    ep->rx_skbuff[i] = 0;
  }
  ep->rx_ring[i-1].next = ep->rx_ring_dma;
}

int main()
{
  struct epic_private ep;
  struct epic_rx_desc rx_ring[32];
  int i;

  for (i=0;i<32;i++)
    rx_ring[i].next=0;

  ep.rx_ring=rx_ring;
  ep.rx_ring_dma=(unsigned int)rx_ring;
  epic_init_ring(&ep);

  for (i=0;i<5;i++)
  {
    printf("%8X\n",ep.rx_ring[i].next);
  }
}  
>Fix:
Compile with -O1.
>Release-Note:
>Audit-Trail:
>Unformatted: