Hi Michael

Am Freitag, 26. September 2014, 08:46:21 schrieb Michael Olbrich:
> On Wed, Sep 24, 2014 at 03:43:10PM +0200, Tim Sander wrote:
> > > I'm happy to announce that I've just released
> > > OSELAS.Toolchain-2013.12.2.
> > > 
> > > This is a bugfix-only release. The relevant changes since 2013.12.1 are:
> > >  - The toolchains can now be built with make 4.0
> > >  - Some toolchains did not optimize for the correct CPU. The code works
> > >  but
> > >  
> > >    might be slower. This is fixed now.
> > >  
> > >  - All security relevant fixes mentioned in the glibc 2.19 changelog
> > >  have
> > >  
> > >    been added for glibc 2.18 based toolchains.
> > >  
> > >  - All generated toolchain archives are now XZ comressed.
> > 
> > While i have used this toolchain arm-v7a-linux-gnueabihf with an Xilinx
> > Zynx compiling nearly the same Project with a different Kernelconfig /DTB
> > for Altera SOC fails with this toolchain !?! The kernel boots but on
> > mounting the rootfs it fails to execute any binary?
> 
> Is the dynamic linker (whatever /lib/ld-linux-armhf.so.3 points to)
> executable? Do you have a working system? Boot it and try to chroot into
> your new system. You get better error messages that way.
Ok i have now setup booted into the environment compiled with 
OSELAS.Toolchain-2012.12.1/arm-cortexa9-linux-gnueabihf, mounted the other 
system via nfs and tried to start any executable:

/mnt/bin/busybox: line 1: syntax error: unexpected word (expecting ")")

(thats with OSELAS.Toolchain-2013.12.2/arm-v7a-linux-gnueabihf on Altera SOC)

I have taken a look at both busyboxes and have run them through "readelf -e" 
(attached). The only difference which sprung to my eye was the hardfloat 
support, which i don't understand as both toolchains should be hardfloat 
toolchains.

Starting the binary with strace gives:
execve("mnt/bin/busybox", ["mnt/bin/busybox"], [/* 11 vars */]) = -1 ENOEXEC 
(Exec format error)
write(2, "strace: exec: Exec format error\n", 32strace: exec: Exec format 
error
) = 32
exit_group(1)                           = ?
+++ exited with 1 +++

So i have an execution error and the shell error seems to be the shell 
fallback from the kernel. 

Another thing which caught my attention was the fact that:
objdump -d bin/busybox |grep UNDEFINED |wc -l
is 364 for the old and 13478 for the new toolchain. 

Best regards
Tim