[Bug other/11669] New: [3.4 regression] Calling of static functions does not work (build of glibc broken)

[Bug other/11669] New: [3.4 regression] Calling of static functions does not work (build of glibc broken)

[aj 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=11669

           Summary: [3.4 regression] Calling of static functions does not
                    work (build of glibc broken)
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: critical
          Priority: P3
         Component: other
        AssignedTo: geoffk at gcc dot gnu dot org
        ReportedBy: aj at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: i686-linux-gnu
  GCC host triplet: i686-linux-gnu
GCC target triplet: i686-linux-gnu

This is a testcase reduced from a failure to compile glibc on i686-linux-gnu
with current GCC:

gromit:~/tmp:[0]$ /opt/gcc/3.4-devel/bin/gcc -c test-failure.c -O2     
gromit:~/tmp:[0]$ nm test-failure.o 
00000010 T call_fixup
00000000 t fixup
gromit:~/tmp:[0]$ /opt/gcc/3.4-devel/bin/gcc -c test-failure.c -O2 -DUSE_DECLARATION
gromit:~/tmp:[0]$ nm test-failure.o 
00000010 T call_fixup
         U fixup
00000000 t fixup.0

The fixup.0 declaration is wrong here.  I suspect the IMI patches and assigned
it therefore to Geoff Keating.

To reproduce use the simple follwing simple code snippet:

#if USE_DECLARATION
/* Removing the prototype makes this work.  */
static int fixup (int i) __attribute__ ((used));
#endif

static int __attribute__ ((__used__))
fixup (int i)
{
  return i+1;
}

int
call_fixup (void)
{

  asm ("call fixup\n\t");

  return 0;
}

[Bug other/11669] [3.4 regression] Calling of static functions does not work (build of glibc broken)

[pinskia at physics dot uc dot edu]

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

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11669



------- Additional Comments From pinskia at physics dot uc dot edu  2003-07-25 11:24 -------
I always thought using control instructions in asm is invalid and wrong to do?

[Bug other/11669] [3.4 regression] Calling of static functions does not work (build of glibc broken)

[pinskia at physics dot uc dot edu]

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

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11669


pinskia at physics dot uc dot edu changed:

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


------- Additional Comments From pinskia at physics dot uc dot edu  2003-07-25 12:53 -------
>From bug 11563:
You cannot alter control flow in an inline asm.  This is clearly documented.
So glibc is wrong here.

[Bug other/11669] [3.4 regression] Calling of static functions does not work (build of glibc broken)

[schwab at suse dot de]

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

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11669


schwab at suse dot de changed:

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


------- Additional Comments From schwab at suse dot de  2003-07-25 14:44 -------
Please check more carefully.  This has nothing at all to do with control flow.  
GCC is generating the wrong label.  Just replace "call fixup" with "movl 
$fixup,%eax".

[Bug other/11669] [3.4 regression] Calling of static functions does not work (build of glibc broken)

[pinskia at physics dot uc dot edu]

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

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11669


pinskia at physics dot uc dot edu changed:

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


------- Additional Comments From pinskia at physics dot uc dot edu  2003-07-25 14:50 -------
But you should never assume the name of static variable then so the code is still wrong.
The way to fix it is to use asm to make sure the name will be the same as the asm name.
#if USE_DECLARATION
/* Removing the prototype makes this work.  */
static int fixup (int i)__asm__("fixup") __attribute__ ((used)); 
#endif

static int __attribute__ ((__used__))
fixup (int i)
{
  return i+1;
}

int
call_fixup (void)
{

  asm ("call fixup\n\t");

  return 0;
}

[Bug other/11669] [3.4 regression] Calling of static functions does not work (build of glibc broken)

[drow at mvista 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=11669



------- Additional Comments From drow at mvista dot com  2003-07-26 00:21 -------
Subject: Re:  [3.4 regression] Calling of static functions does not work (build of glibc broken)

This also "breaks" GDB in some cases.  We already decided to use the
old names for static functions in single-file mode.  Being inconsistent
about that is bad.

[Bug other/11669] [3.4 regression] Calling of static functions does not work (build of glibc broken)

[aj 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=11669


aj at gcc dot gnu dot org changed:

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


------- Additional Comments From aj at gcc dot gnu dot org  2003-07-26 05:27 -------
GCC changes behaviour and breaks in this way GCC and GDB.
For single-file mode the behaviour should not be changed.

I agree that for multiple-file mode, the usage of asm ("fixup") is the right way
but let's not force this for single-file - and it was already agreed on 
the mailing lists that this should be changed back.

Therefore I'm reopening the bug.

[Bug other/11669] [3.4 regression] Calling of static functions does not work (build of glibc broken)

[geoffk 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=11669


geoffk at gcc dot gnu dot org changed:

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


------- Additional Comments From geoffk at gcc dot gnu dot org  2003-07-26 10:16 -------
This should be fixed in current HEAD, with

        * c-decl.c (c_static_assembler_name): Remove TREE_STATIC test.

[Bug other/11669] [3.4 regression] Calling of static functions does not work (build of glibc broken)

[mec at shout dot net]

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

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11669



------- Additional Comments From mec at shout dot net  2003-07-26 16:17 -------
This happens to me on native i686-pc-linux-gnu with -gstabs+ with the following
source file:

  static void f1 ();
  static void f1 () { return; }
  static void f2 () { return; }

f1 is getting qualified as "f1.0".  f2 is not getting qualified, which is right.
 This causes gdb to not find the symbol "f1" with -gstabs+ debugging format.

I started a date search.  The bug did not happen with gcc HEAD from 2003-07-19
15:00:00 UTC, and it did happen with gcc HEAD from 2003-07-22 02:00:00 UTC.

Geoff K says it is fixed now and I am building a current compiler to try it out.

[Bug other/11669] [3.4 regression] Calling of static functions does not work (build of glibc broken)

[mec at shout dot net]

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

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11669



------- Additional Comments From mec at shout dot net  2003-07-26 17:07 -------
It works for me.  Tested with gcc HEAD 2003-07-26 16:00:00 UTC, native
i686-pc-linux-gnu, binutils 2.14, -gstabs+.  Thanks!

[Bug other/11669] [3.4 regression] Calling of static functions does not work (build of glibc broken)

[geoffk 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=11669


geoffk at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED


------- Additional Comments From geoffk at gcc dot gnu dot org  2003-07-26 18:37 -------
Confirmed fixed by mec@shout.net.