Re: Segfaults with eglibc-2.16 on ARM v6 (hardfp)

[Carlos O'Donell <carlos_odonell at mentor dot com>]

On 7/19/2012 12:34 PM, Stephan Raue wrote:
> Am 19.07.2012 17:48, schrieb Carlos O'Donell:
>> On 7/19/2012 10:05 AM, Stephan Raue wrote:
>>> Hi,
>>>
>>> after upgrading my embedded distribution from eglibc-2.15 to 2.16 i get libc related segfaults with every program. Because this distro uses shell based init scripts and busybox, which also segfaults provides the shell i cant boot into the system with eglibc-2.16. With eglibc-2.15 all is working.
>>>
>>> I have tried to boot in to a system with eglibc-2.15 and copy the eglibc-2.16 filessystem to a writabable location in /storage:
>>>      root ~ # LD_LIBRARY_PATH=/storage/system/lib/ /storage/system/bin/busybox
>>>      Segmentation fault
>>>
>>> if i replace /storage/system/lib/libc.so.6 with the one from eglibc-2.15 system:
>> WARNING: You should replace all core libraries with the 2.15 versions. The dynamic linker and libc share state and should be at matching versions.
>>
>>>      root ~ # LD_LIBRARY_PATH=/storage/system/lib/ /storage/system/bin/busybox
>>>      BusyBox v1.20.2 (2012-07-19 08:32:53 CEST) multi-call binary.
>>>      Copyright (C) 1998-2011 Erik Andersen, Rob Landley, Denys Vlasenko
>>>      ...
>>>
>>> how i can debug this system to get more infos to provide to find the issue? i have GDB and STRACE installed on the system, if i run gdb:
>>>      root ~ # gdb /storage/system/bin/busybox
>>>      GNU gdb (GDB) 7.4.50.20120716
>>>      ...
>>>      (gdb) set environment LD_LIBRARY_PATH=/storage/system/lib/
>>>      (gdb) run
>>>      Starting program: /storage/system/bin/busybox
>>>      During startup program terminated with signal SIGSEGV, Segmentation fault.
>>>      (gdb) bt
>>>      No stack.
>> This is likely a crash in the dynamic loader.
>>
>> You need to debug the dynamic loader.
>>
>> Start by running with LD_DEBUG=all and see how far you get.
>>
>> Cheers,
>> Carlos.
> thanks much!
> 
> i get this output: http://pastebin.com/udL0kgn3
> 
> is this a known and already reported issue?

Never seen this before.

You'll need to debug the dynamic loader.

Cheers,
Carlos.
-- 
Carlos O'Donell
Mentor Graphics / CodeSourcery
carlos_odonell@mentor.com
carlos@codesourcery.com
+1 (613) 963 1026