Re: linux-next: ftrace tree build failure

[Steven Rostedt <srostedt@redhat.com>]

On Tue, 2008-11-25 at 15:13 +1100, Stephen Rothwell wrote:
> Hi Steve,
> 
> On Mon, 24 Nov 2008 22:48:51 -0500 Steven Rostedt <srostedt@redhat.com> wrote:
> >
> > Could you post a git branch that had this failure, so I can compile it
> > to see what happened?
> 
> ssh://master.kernel.org/~sfr/ftrace master
> 
> to get the failure, I did:
> 
> make ARCH=x86_64 allmodconfig
> make ARCH=x86_64 arch/x86/boot/tty.o
> 
> My machine is Debian sid.
> 
> $ gcc --version
> gcc (Debian 4.3.2-1) 4.3.2
> $ ld --version
> GNU ld (GNU Binutils for Debian) 2.18.0.20080103
> 
> The original failure was with a gcc 4.3.2 cross compiled for a powerpc
> host and a binutils 2.19.
> 
> You can see which trees are merged by looking for commits by me.

I traced it down to this.

It compiles without -Os.

But I leave -Os in...

Doing a make arch/x86/boot/tty.i  I get this:

static int kbd_pending(void)
{
 u8 pending;
 asm volatile("int $0x16; setnz %0"
       : "=rm" (pending)
       : "a" (0x0100));
 return pending;
}

void kbd_flush(void)
{
 for (;;) {
  if (__builtin_constant_p((!kbd_pending())) ? !!(!kbd_pending()) :
({ int ______r; static struct ftrace_branch_data
__attribute__((__aligned__(4)))
__attribute__((section("_ftrace_branch"))) ______f = { .func =
__func__, .file = "arch/x86/boot/tty.c", .line = 85, }; ______r = !!(!
kbd_pending()); if (______r) ______f.hit++; else ______f.miss++;
______r; }))
   break;
  getchar();
 }
}


By removing the "static" and "__attribute__((section...))" which is
needed to remove static. It compiles.

It also compiles if I only remove the "if (______r) ______f.hit++; else
______f.miss++;

There's something buggy with gcc where it can not handle -Os and this
combination of having this static variable in the if statement.

-- Steve