lkml.org 
[lkml]   [2009]   [Dec]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: 2.6.33-rc1 Reboot right after bootloader
Johannes Stezenbach wrote:
> On Sun, Dec 20, 2009 at 09:46:56AM +0100, Sam Ravnborg wrote:
>> On Sun, Dec 20, 2009 at 12:34:57AM +0100, Michael Guntsche wrote:
>>> I took me some time but I tracked down my reboot problem. The culprit is
>>> commit
>>>
>>> 4a2ff67c88211026afcbdbc190c13f705dae1b59: kbuild: fix bzImage build for
>>> x86
>> Just to check what is going wrong here could you try to execute the following two commands:
>> printf \\xa8\\x51\\x37\\x00 > x ; hexdump x
>> /usr/bin/printf \\xa8\\x51\\x37\\x00 > x ; hexdump x
>> echo -ne \\xa8\\x51\\x37\\x00 > x ; hexdump x
>>
>> Please try as above and also with full path to printf (/usr/bin/printf)
>
> Debian dash has a built-in printf which doesn't support \x.
> /usr/bin/printf works fine.

So it looks like I was wrong with the last patch here.
I changed `/bin/echo -ne' to printf. And while it worked
for me at that time, apparently it does not work for others.

Now I wonder why it worked for me. I can confirm that
in current debian testing (dash-0.5.5.1-3) and in debian
stable (dash-0.5.4-12) dash's built-in printf does not
interpret \x escape sequence. I sure verified the fix
I proposed, rebuilding kernels in a freshly-installed
debian testing with /bin/sh pointing to dash. I'll try
to investigate this. (Side note: since the time this
issue first hit me, I carry 4a2ff67c88211026afcbdbc190c13f705dae1b59
locally.)

It is even more interesting. I re-read POSIX description
of printf, see
http://www.opengroup.org/onlinepubs/000095399/utilities/printf.html
And this one, too, does NOT mention \x at all. Printf is
still "better" than echo because for echo _no_ interpretation
of escape sequences is mandated -- see
http://www.opengroup.org/onlinepubs/000095399/utilities/echo.html
but that does not help much since for printf, while \-sequences
are mandated, particular \x is not, so \x is a {GNU|common sense|...}
extension.
What the...

So now I don't know what to do. According to the standard, there's
no utility that will work here. ;) Maybe the attached (together
with 4a2ff67c88211026afcbdbc190c13f705dae1b59) will make everyone happy?

/mjt
--- a/scripts/Makefile.lib.orig 2009-12-19 18:36:01.944153109 +0300
+++ a/scripts/Makefile.lib 2009-12-20 13:36:04.014530573 +0300
@@ -215,6 +215,6 @@
dec_size=$$(expr $$dec_size + $$fsize); \
done; \
-printf "%08x" $$dec_size | \
- sed 's/\(..\)\(..\)\(..\)\(..\)/\\\\x\4\\\\x\3\\\\x\2\\\\x\1/g' \
+printf "%012o" $$dec_size | \
+ sed 's/\(...\)\(...\)\(...\)\(...\)/\\\\4\\\\3\\\\2\\\\1/g' \
)
\
 
 \ /
  Last update: 2009-12-20 11:41    [from the cache]
©2003-2011 Jasper Spaans