[Bug c/12324] New: undefined symbols with -O2 with aliased pointers

[Bug c/12324] New: undefined symbols with -O2 with aliased pointers

[bernie at develer dot com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324

           Summary: undefined symbols with -O2 with aliased pointers
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: bernie at develer dot com
                CC: gcc-bugs at gcc dot gnu dot org

Compiling this code snippet with -O2 or -Os 
 
	static void undefined(void) { } 
	void *bar = (void * []) { undefined }; 
 
incorrectly optimizes away undefined(), producing this 
nm output: 
 
	00000000 D bar 
	         U undefined

[Bug optimization/12324] [3.4 Regression] [unti-at-a-time] not emitting function

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324


pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hubicka at gcc dot gnu dot
                   |                            |org
           Severity|normal                      |critical
             Status|UNCONFIRMED                 |NEW
          Component|c                           |optimization
     Ever Confirmed|                            |1
           Keywords|                            |wrong-code
   Last reconfirmed|0000-00-00 00:00:00         |2003-09-18 04:35:34
               date|                            |
            Summary|undefined symbols with -O2  |[3.4 Regression] [unti-at-a-
                   |with aliased pointers       |time] not emitting function
   Target Milestone|---                         |3.4


------- Additional Comments From pinskia at gcc dot gnu dot org  2003-09-18 04:35 -------
I can confirm this on the mainline (20030916), the problem is unit-at-a-time.  It also happens in 
20030830 with -funit-at-a-time.

[Bug optimization/12324] [3.4 Regression] [unti-at-a-time] not emitting function referenced from a global variable (void* type)

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324


pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[3.4 Regression] [unti-at-a-|[3.4 Regression] [unti-at-a-
                   |time] not emitting function |time] not emitting function
                   |referenced from a global    |referenced from a global
                   |variable                    |variable (void* type)


------- Additional Comments From pinskia at gcc dot gnu dot org  2003-09-18 04:39 -------
This also fails:
       static void undefined(void) { }
        void*bar = (void(*[])(void)) {undefined };

void main()
{
}

[Bug optimization/12324] [3.4 Regression] [unti-at-a-time] not emitting function referenced from a compound statement

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324


pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[3.4 Regression] [unti-at-a-|[3.4 Regression] [unti-at-a-
                   |time] not emitting function |time] not emitting function
                   |referenced from a global    |referenced from a compound
                   |variable (void* type)       |statement


------- Additional Comments From pinskia at gcc dot gnu dot org  2003-09-22 08:44 -------
Here is a case where it does not take -funit-at-a-time to fail:
        static inline void undefined(void) { }
        void *bar = (void * []) { undefined };

[Bug optimization/12324] [3.4 Regression] [unti-at-a-time] not emitting function referenced from a compound statement

[bernie at develer dot com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324



------- Additional Comments From bernie at develer dot com  2003-09-25 20:48 -------
Not sure if it's the same bug, but 3.4 outputs
nothing for this code with -O2:

  static int foo __attribute__((unused)) = 42;

This breaks the Linux kernel (the __setup() macro).
GCC 3.3.1 did the right thing:

        .data
        .align 4
        .type   foo, @object
        .size   foo, 4
foo:
        .long   42

[Bug optimization/12324] [3.4 Regression] [unti-at-a-time] not emitting function referenced from a compound statement

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324



------- Additional Comments From pinskia at gcc dot gnu dot org  2003-09-25 20:51 -------
GCC 3.4 for this code:
  static int foo __attribute__((unused)) = 42;
is doing the right thing, you should be using the attribute used instead.

[Bug optimization/12324] [3.4 Regression] [unti-at-a-time] not emitting function referenced from a compound statement

[bernie at develer dot com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324



------- Additional Comments From bernie at develer dot com  2003-09-25 21:43 -------
It's right, attribute((used)) makes it work, but then we have two problems: 
 - it used to work before and it's breaking Linux. Shall we fix it there or in GCC? 
 - gcc.texi says that used can only be used with functions, but it works on 
   variables too. Shall we correct that?

[Bug optimization/12324] [3.4 Regression] [unti-at-a-time] not emitting function referenced from a compound statement

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324



------- Additional Comments From pinskia at gcc dot gnu dot org  2003-09-25 21:55 -------
In Linux because unused really means that it is might be unused but I do not want warns about it.

[Bug optimization/12324] [3.4 Regression] [unti-at-a-time] not emitting function referenced from a compound statement

[hubicka at ucw dot cz]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324



------- Additional Comments From hubicka at ucw dot cz  2003-09-26 08:43 -------
Subject: Re:  [3.4 Regression] [unti-at-a-time] not emitting function referenced from a compound statement

> PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324
> 
> 
> 
> ------- Additional Comments From bernie at develer dot com  2003-09-25 20:48 -------
> Not sure if it's the same bug, but 3.4 outputs
> nothing for this code with -O2:
> 
>   static int foo __attribute__((unused)) = 42;
> 
> This breaks the Linux kernel (the __setup() macro).
> GCC 3.3.1 did the right thing:
> 
>         .data
>         .align 4
>         .type   foo, @object
>         .size   foo, 4
> foo:
>         .long   42

You shall use attribute used here.  Attribute unused merely supress the
warning, while attribute used makes it to be output.  I will look into
the other problem.
Note that attribute used did not work for variables for older GCC, so
you need ifdef machinery.

Honza
> 
> 
> 
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.

[Bug optimization/12324] [3.4 Regression] [unti-at-a-time] not emitting function referenced from a compound statement

[bernie at develer dot com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324



------- Additional Comments From bernie at develer dot com  2003-09-26 16:29 -------
Subject: Re:  [3.4 Regression] [unti-at-a-time] not
 emitting function referenced from a compound statement

hubicka at ucw dot cz wrote:

>>------- Additional Comments From bernie at develer dot com  2003-09-25 20:48 -------
>>Not sure if it's the same bug, but 3.4 outputs
>>nothing for this code with -O2:
>>
>>  static int foo __attribute__((unused)) = 42;
>>
>>This breaks the Linux kernel (the __setup() macro).
>>GCC 3.3.1 did the right thing:
>>
>>        .data
>>        .align 4
>>        .type   foo, @object
>>        .size   foo, 4
>>foo:
>>        .long   42
> 
> You shall use attribute used here.  Attribute unused merely supress the
> warning, while attribute used makes it to be output.

Yes, I tried and it works. The documentation does not mention attribute
used for variables. I think it should be fixed.

> I will look into the other problem.

Thanks. That one also breaks the Linux kernel (drivers/net/fec.c uses
an idiom similar to the test case).

> Note that attribute used did not work for variables for older GCC, so
> you need ifdef machinery.

Do you know the exact GCC version that introduced it? I'd like to provide
a patch for the kernel. I can only confirm that 'used' works in 3.3.1 for
variables.

[Bug optimization/12324] [3.4 Regression] [unti-at-a-time] not emitting function referenced from a compound statement

[hubicka at ucw dot cz]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324



------- Additional Comments From hubicka at ucw dot cz  2003-09-26 20:13 -------
Subject: Re:  [3.4 Regression] [unti-at-a-time] not emitting function referenced from a compound statement

> PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324
> 
> 
> 
> ------- Additional Comments From bernie at develer dot com  2003-09-26 16:29 -------
> Subject: Re:  [3.4 Regression] [unti-at-a-time] not
>  emitting function referenced from a compound statement
> 
> hubicka at ucw dot cz wrote:
> 
> >>------- Additional Comments From bernie at develer dot com  2003-09-25 20:48 -------
> >>Not sure if it's the same bug, but 3.4 outputs
> >>nothing for this code with -O2:
> >>
> >>  static int foo __attribute__((unused)) = 42;
> >>
> >>This breaks the Linux kernel (the __setup() macro).
> >>GCC 3.3.1 did the right thing:
> >>
> >>        .data
> >>        .align 4
> >>        .type   foo, @object
> >>        .size   foo, 4
> >>foo:
> >>        .long   42
> > 
> > You shall use attribute used here.  Attribute unused merely supress the
> > warning, while attribute used makes it to be output.
> 
> Yes, I tried and it works. The documentation does not mention attribute
> used for variables. I think it should be fixed.
> 
> > I will look into the other problem.
> 
> Thanks. That one also breaks the Linux kernel (drivers/net/fec.c uses
> an idiom similar to the test case).
> 
> > Note that attribute used did not work for variables for older GCC, so
> > you need ifdef machinery.
> 
> Do you know the exact GCC version that introduced it? I'd like to provide
> a patch for the kernel. I can only confirm that 'used' works in 3.3.1 for
> variables.

I believe it actually is 3.3.1.  It is safe to test for that version as
no earlier version elliminated unused variables.

Honza
> 
> 
> 
> 
> 
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.

[Bug optimization/12324] [3.4 Regression] [unti-at-a-time] not emitting function referenced from a compound statement

[bernie at develer dot com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324



------- Additional Comments From bernie at develer dot com  2003-09-27 02:40 -------
For what you may care, the attribute((used)) fix has now
been applied to Linus' kernel:

http://linux.bkbits.net:8080/linux-2.5/cset@1.1345?nav=index.html|ChangeSet@-1d

[Bug optimization/12324] [3.4 Regression] [unit-at-a-time] not emitting function referenced from a compound statement

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324


pinskia at gcc dot gnu dot org changed:

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

[Bug optimization/12324] [3.4 Regression] [unit-at-a-time] not emitting function referenced from a compound statement

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2003-11-29 02:14 -------
There are patches floating around but someone needs to approve one of them.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2003-09-18 04:35:34         |2003-11-29 02:14:51
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324

[Bug optimization/12324] [3.4 Regression] [unit-at-a-time] not emitting function referenced from a compound statement

[bernie at develer dot com]

------- Additional Comments From bernie at develer dot com  2003-11-29 16:04 -------
This patch by Jan Hubicka fixed this PR for me:

 http://gcc.gnu.org/ml/gcc-patches/2003-10/msg01402.html

There was longish thread in October on gcc-patches, and
this post by Richard Henderson seems to be approving it:

  http://gcc.gnu.org/ml/gcc-patches/2003-10/msg01451.html

May I go on and commit the patch?


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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324

[Bug optimization/12324] [3.4 Regression] [unit-at-a-time] not emitting function referenced from a compound statement

[hubicka at ucw dot cz]

------- Additional Comments From hubicka at ucw dot cz  2003-11-29 16:10 -------
Subject: Re:  [3.4 Regression] [unit-at-a-time] not emitting function referenced from a compound statement

> 
> ------- Additional Comments From bernie at develer dot com  2003-11-29 16:04 -------
> This patch by Jan Hubicka fixed this PR for me:
> 
>  http://gcc.gnu.org/ml/gcc-patches/2003-10/msg01402.html
> 
> There was longish thread in October on gcc-patches, and
> this post by Richard Henderson seems to be approving it:
> 
>   http://gcc.gnu.org/ml/gcc-patches/2003-10/msg01451.html
> 
> May I go on and commit the patch?
Actually the last incarnation
(http://gcc.gnu.org/ml/gcc-patches/2003-10/msg01402.html) has never been
approved.  Richard, would be possible to look into that one?  It is
really trivial.

Honza

> 
> 
> -- 
>            What    |Removed                     |Added
> ----------------------------------------------------------------------------
>                  CC|                            |rth at redhat dot com
> 
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324
> 
> ------- You are receiving this mail because: -------
> You are the assignee for the bug, or are watching the assignee.
> You are on the CC list for the bug, or are watching someone who is.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324

[Bug optimization/12324] [3.4 Regression] [unit-at-a-time] not emitting function referenced from a compound statement

[rth at redhat dot com]

------- Additional Comments From rth at redhat dot com  2003-12-02 00:08 -------
Subject: Re:  [3.4 Regression] [unit-at-a-time] not emitting function referenced from a compound statement

On Sat, Nov 29, 2003 at 05:10:37PM +0100, Jan Hubicka wrote:
> (http://gcc.gnu.org/ml/gcc-patches/2003-10/msg01402.html)

You've got some unneeded parenthesis in there (x || (y || z)),
but otherwise ok.


r~


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324

[Bug optimization/12324] [3.4 Regression] [unit-at-a-time] not emitting function referenced from a compound statement

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2003-12-03 23:04 -------
Subject: Bug 12324

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	hubicka@gcc.gnu.org	2003-12-03 23:04:23

Modified files:
	gcc            : ChangeLog toplev.c 

Log message:
	PR optimization/12324
	* toplev.c  (rest_of_decl_compilation): Do not deffer when compiling
	in unit-at-a-time mode.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.1903&r2=2.1904
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/toplev.c.diff?cvsroot=gcc&r1=1.848&r2=1.849



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324

[Bug optimization/12324] [3.4 Regression] [unit-at-a-time] not emitting function referenced from a compound statement

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2003-12-03 23:23 -------
Fixed for 3.4.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12324