[klibc] sample sleep command killed

[Phil Howard]

On Thu, Jul 15, 2010 at 6:02 AM, maximilian attems <max at stro.at> wrote:
> On Wed, Jul 14, 2010 at 07:47:43PM -0400, Phil Howard wrote:
>>
>> > axe your wrong built and do it the recommended way.
>>
>> Can you put the recommended way into some kind of complete document?
>
> sure, here the quick way. It assumes you have a ~src directory in
> your home as common practise. nproc is found in recent coreutils (8.1).
>
> in short words the recommend way is to built klibc against
> configured linux-2.6 tree:
>
> cd ~/src
> git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
> cd linux-2.6
> make defconfig
> make -j $(nproc)
> git clone git://git.kernel.org/pub/scm/libs/klibc/klibc.git
> cd klibc
> make  KLIBCKERNELSRC=`pwd`/../linux-2.6 -j $(nproc)

The differences I see are:

1.  You are running more processes in the build.  I fail to see how
using one processor instead of four would make the sleep executable
image (but not others) be built wrong.

2.  You use the latest git versions, rather than the latest released
tarballs.  Are you saying that Klibc is broken in 1.5.18 and git has
the unreleased fix?

3.  You used defconfig whereas I used oldconfig originally.  I tried
both.  No difference.

4.  You are using the original kernel tree itself, rather than
installed headers.  Earlier recommendations suggest the installed
headers.  Turns out the kernel source tree header layout changes.
Installed header layout should be more stable.  Can you tell us why
you think not using installed headers is better?

I'll try other things.  This is, afterall, failing BEFORE the sleep
executable even gets executed.  And it is the kernel that is
apparently deciding something is wrong with the executable while in
sys_execve.  I think I'll focus on what is wrong with the kernel or
the toolchain.