[Bug c/55749] New: gcc 4.7.1 removes labels mistakenly

[Bug c/55749] New: gcc 4.7.1 removes labels mistakenly

[blue_3too at hotmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55749

             Bug #: 55749
           Summary: gcc 4.7.1 removes labels mistakenly
    Classification: Unclassified
           Product: gcc
           Version: 4.7.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: blue_3too@hotmail.com


gcc version 4.7.1 (GCC) x86_64-unknown-linux-gnu

in the following small c program, label-as-value always incorrectly uses the
begining of the function (.e. pointer p is always set to the begining of foo).
Turning of scheduling will generate the correct code.

Is this a known bug? any pointer will be appreciated. thanks
>cat m.c

void *p;
extern void bar();
void foo()
{
    p = &&my_label;
    bar();

my_label:

    bar();
}

> gcc -O2 -S m.c

> less m.s

       .globl  foo
        .type   foo, @function
foo:
.LFB0:
.L2:
        subq    $8, %rsp
.LCFI0:
        xorl    %eax, %eax
        movq    $.L2, p(%rip)
        call    bar
        xorl    %eax, %eax
        addq    $8, %rsp
.LCFI1:
        jmp     bar
.LFE0:
        .size   foo, .-foo
        .comm   p,8,8
        .comm   i,4,4

[Bug c/55749] gcc 4.7.1 removes labels mistakenly

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55749

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

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

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-12-20 05:31:56 UTC ---
This is not a bug as you don't use labels-as-a-value with a computed goto.  You
need a computed goto to use label correctly so since you don't have one, GCC
assumes you are never jumping to that label and moves it around.

[Bug c/55749] gcc 4.7.1 removes labels mistakenly

[blue_3too at hotmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55749

--- Comment #2 from blue_3too at hotmail dot com 2012-12-20 16:17:20 UTC ---
Thanks for the comments. I checked the document @
gcc.gnu.org/onlinedocs/gcc/Labels-as-Values.html . But find no description hat
label-as-a-value should be used with computed goto.  Can you be more specific
about where is the decription from?

But back to my problem: my_labe is a resuming point for a thread to acquire a
lock after being blocked on the lock and slept. When it is waken, it should
resume execution from my_label. So I guess it conradicts with following
description in the doc:
"You may not use this mechanism to jump to code in a different function. If you
do that, totally unpredictable things happen. The best way to avoid this is to
store the label address only in automatic variables and never pass it as an
argument. "

Because the label is passed out of the function, it is used to jump back the
function when the sleeping thread is awaken. Now

[Bug c/55749] gcc 4.7.1 removes labels mistakenly

[blue_3too at hotmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55749

--- Comment #3 from blue_3too at hotmail dot com 2012-12-20 16:18:23 UTC ---
Thanks for the comments. I checked the document @
gcc.gnu.org/onlinedocs/gcc/Labels-as-Values.html . But find no description hat
label-as-a-value should be used with computed goto.  Can you be more specific
about where is the description from?

But back to my problem: my_labe is a resuming point for a thread to acquire a
lock after being blocked on the lock and slept. When it is waken, it should
resume execution from my_label. So I guess it contradicts with following
description in the doc:
"You may not use this mechanism to jump to code in a different function. If you
do that, totally unpredictable things happen. The best way to avoid this is to
store the label address only in automatic variables and never pass it as an
argument. "

Because the label is passed out of the function, it is used to jump back the
function when the sleeping thread is awaken. Not sure how to implement this
properly.

[Bug c/55749] gcc 4.7.1 removes labels mistakenly

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55749

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-12-20 16:54:53 UTC ---
"To use these values, you need to be able to jump to one. This is done with the
computed goto statement1, goto *exp;."

> Not sure how to implement this properly.

Use either setjmp/longjmp or getcontext/setcontext/makecontext instead.

[Bug c/55749] gcc 4.7.1 removes labels mistakenly

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55749

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|INVALID                     |DUPLICATE

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
.

*** This bug has been marked as a duplicate of bug 44298 ***