Re: [PATCH 1/2] namespace.pl: fix source tree name mangling

[Stephen Hemminger]

On Wed, 29 Sep 2010 12:31:12 +0800
Américo Wang <xiyou.wangcong@xxxxxxxxx> wrote:

> On Tue, Sep 28, 2010 at 05:44:02PM +0900, Stephen Hemminger wrote:
> >The current namespace.pl script does not find source files correctly.
> >The problem is that the current directory is not the base of the kernel
> >tree at the point where it calls objdump.
> >
> >Signed-off-by: Stephen Hemminger <shemminger@xxxxxxxxxx>
> >
> >--- a/scripts/namespace.pl	2010-09-27 17:53:29.919433444 +0900
> >+++ b/scripts/namespace.pl	2010-09-28 09:24:36.891552993 +0900
> >@@ -167,8 +167,10 @@ sub do_nm
> > 		printf STDERR "$fullname is not an object file\n";
> > 		return;
> > 	}
> >-	($source = $fullname) =~ s/\.o$//;
> >-	if (-e "$objtree$source.c" || -e "$objtree$source.S") {
> >+	$fullname =~ s/\.o$//;
> >+	$source = $basename;
> >+	$source =~ s/\.o$//;
> 
> With your patch applied, $source will be the basename of
> an obj path with .o stripped.
> 
> >+	if (-e "$objtree$fullname.c" || -e "$objtree$fullname.S") {
> > 		$source = "$objtree$source";
> > 	} else {
> > 		$source = "$srctree$source";
> >
> 
> So here we will get a non-exist path stored in $source.
> 
> Are you serious? What problem did you meet? I see no problem
> with the original code here.

The original script is broken, if you run it on a compiled
kernel tree.

No source file found for arch/x86/boot/a20.o
No source file found for arch/x86/boot/bioscall.o
No source file found for arch/x86/boot/cmdline.o
No source file found for arch/x86/boot/copy.o
No source file found for arch/x86/boot/cpucheck.o
No source file found for arch/x86/boot/early_serial_console.o
No source file found for arch/x86/boot/edd.o
No source file found for arch/x86/boot/main.o
No source file found for arch/x86/boot/mca.o
No source file found for arch/x86/boot/memory.o
No source file found for arch/x86/boot/pm.o
No source file found for arch/x86/boot/pmjump.o
No source file found for arch/x86/boot/printf.o

...

Running under perl debugger shows that the script has
done chdir prior to the failing test:

	if (! -e "$source.c" && ! -e "$source.S") {
		# No obvious source, exclude the object if it is conglomerate
>>	   open(my $objdumpdata, "$objdump $basename|")
		    or die "$objdump $fullname failed $!\n";

For the first error:
DB<3> x $source
0  'arch/x86/boot/a20'
  DB<4> !!pwd
/home/shemminger/kernel/net-next-2.6/arch/x86/boot
  DB<5> x $fullname
0  'arch/x86/boot/a20.o'
  DB<6> x $basename
0  'a20.o'

Therefore the problem was that $source was full path not the base
of the file name 'a20'
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/