amplxe-gui : segmentation fault on startup

[Alexis_R_]

First time user of the amplifier here. Apologies if this is a FAQ.

When running amplxe-gui, the splash screen comes up for a split second, and I then get Segmentation fault.

$ amplxe-gui
Segmentation fault

This is a fresh install without root privileges. Here are some details about the install and the system (edited hostname):

$ uname -a
Linux host.name.edu 2.6.18-308.16.1.el5 #1 SMP Tue Sep 18 07:21:07 EDT 2012 x86_64 x86_64 x86_64 GNU/Linux

$ cat /etc/redhat-release
Red Hat Enterprise Linux Client release 5.8 (Tikanga)

$ cat /proc/cpuinfo | grep Xeon | tail -1
model name      : Intel(R) Xeon(R) CPU           X5690  @ 3.47GHz

$ amplxe-cl --version
Intel(R) VTune(TM) Amplifier XE 2013 Update 7 (build 288061) Command Line Tool

Please advise how to proceed from there or if you need more info. Thanks.

[Bernard]

>>>0x00002aaaac536996 in cfgmgr2::ConfigDescriptorRegistry::ConfigDescriptorRegistry(cfgmgr2::config_type_t, char const*) () from ${INSTALL_DIR}/vtune_amplifier_xe_2013/bin64/../lib64/libamplxe_cfgmgr_2.15.so
(gdb) backtrace
#0  0x00002aaaac536996 in cfgmgr2::ConfigDescriptorRegistry::ConfigDescriptorRegistry(cfgmgr2::config_type_t, char const*) () from ${INSTALL_DIR}/vtune_amplifier_xe_2013/bin64/../lib64/libamplxe_cfgmgr_2.15.so>>>

Can you  post value of char const* at the time of segfault? I suppose that value can point to bad memory region(unaccessible).

[Bernard]

@A.Rohou,

do you have any updates on your problem?

[Alexis_R_]

Yes, I have tried to run the qfagent.cmd script manually after fixing its encoding. However, it never completes, so i cannot provide the full system information that Mr Anderson requested. Here is what I tried to do - perhaps I have the syntax wrong when calling the script?

$ /tmp/amplxe-log-alr99/2013-06-06-Thu-16-25-37-853042.amplxe-feedback/qfagent.cmd /tmp/amplxe-log-alr99/2013-06-06-Thu-16-25-37-853042.amplxe-feedback /gusr/alr99/linux/intel/vtune_amplifier_xe vtune_amplifier_xe
Quality Feedback Agent
Version: Quality Feedback Agent for Linux*
Time: Mon Jun 10 13:31:18 2013


This diagnostic script generates detailed system information about your system.
When  you forward  the output of this command to Intel(R) Support, the enclosed
server data will be used solely for the purposes of troubleshooting the problem
you have reported to Intel(R) Support. This information will not be shared with
any other companies or users, nor will it be used for any other purpose.

The script is collecting info. Please wait ...
Getting Hardware Information ...
        * Detecting architecture type
        * CPU info
        * CPU affinity
        * Amount of memory
        * PCI/AGP cards installed
Getting Software Information ...
        * Kernel version and type (UP/SMP)
        * glibc version
        * gcc version
        * icc version
        * NPTL version
        * Detecting Linux* distribution
Warning: The product has not been validated with this platform
It may work, but it is not supported
        * Checking IA32 libraries
        * File system disk space usage
        * Intel(R) Software Products installed
        * List loaded kernel modules
        * Check Linux* kernel sources availability
        * Check if sampling driver can be compiled
        * Environment variables
        * System configuration variables
        * System limits and kernel parameters
        * Permission to directories
        * Detecting GTK* and Motif* versions
        * Current locale
        * SELinux configuration
        * The list of running product processes
        * Query runlevel information for system services
        * Information on IPC facilities
        * Current user credentials
Product specific files ...
        * Product install dir
        * Product version
        * Product Driver Kit version
        * Product licenses
        * Printing bootup messages by dmesg
        * Printing messages from file /var/log/messages
        * Install log
        * Uninstall log
        * Product core files ...

At that point, the script hangs for 2+ days (I've tried this twice already).

Mr Anderson, I would like to provide you with the info you need to reproduce this - can you think of another way?

[Bernard]

Did you try to check with GDB what I have suggested in one of my previous posts [ 06/07/2013 - 01:44]?

[Michael_B_13]

Today I installed the Update4 and have the same error. Ubuntu 11, Cluster Studio 2013, all recent updates.

[Bernard]

Can you post stack trace

[Michael_B_13]

Sure:

NAME@NODE:/opt/intel/vtune_amplifier_xe/bin64$ gdb amplxe-gui
GNU gdb (Ubuntu/Linaro 7.3-0ubuntu2) 7.3-2011.08
Copyright (C) 2011 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-linux-gnu".
For bug reporting instructions, please see:
<http://bugs.launchpad.net/gdb-linaro/>...
Reading symbols from /opt/intel/vtune_amplifier_xe_2013/bin64/amplxe-gui...(no debugging symbols found)...done.
(gdb) run
Starting program: /opt/intel/vtune_amplifier_xe_2013/bin64/amplxe-gui
[Thread debugging using libthread_db enabled]

Program received signal SIGSEGV, Segmentation fault.
0x00007ffff5f3c996 in cfgmgr2::ConfigDescriptorRegistry::ConfigDescriptorRegistry(cfgmgr2::config_type_t, char const*) ()
from /opt/intel/vtune_amplifier_xe_2013/bin64/../lib64/libamplxe_cfgmgr_2.15.so
(gdb) backtrace
#0 0x00007ffff5f3c996 in cfgmgr2::ConfigDescriptorRegistry::ConfigDescriptorRegistry(cfgmgr2::config_type_t, char const*) ()
from /opt/intel/vtune_amplifier_xe_2013/bin64/../lib64/libamplxe_cfgmgr_2.15.so
#1 0x00007ffff5f3c698 in cfgmgr2::IConfigDescriptorRegistry::create(cfgmgr2::config_type_t, gen_helpers2::sptr_t<cfgmgr2::IContextValueMap const>) ()
from /opt/intel/vtune_amplifier_xe_2013/bin64/../lib64/libamplxe_cfgmgr_2.15.so
#2 0x00007fffe9504cea in perfclient_1_3::AnalysisTypeHolder::fill() () from /opt/intel/vtune_amplifier_xe_2013/lib64/libamplxe_client_stub_1.3.so
#3 0x00007fffe94fe5e9 in perfclient_1_3::IdeCallback::initialize(eil1::IDE_Env*, char const*) ()
from /opt/intel/vtune_amplifier_xe_2013/lib64/libamplxe_client_stub_1.3.so
#4 0x00007fffe9502c11 in StartUp () from /opt/intel/vtune_amplifier_xe_2013/lib64/libamplxe_client_stub_1.3.so
#5 0x0000000000467480 in ?? ()
#6 0x0000000000433abc in ?? ()
#7 0x0000000000450c56 in ?? ()
#8 0x000000000045146a in wxAppConsole::CallOnInit() ()
#9 0x00007ffff70b61b4 in wxEntry(int&, char**) () from /opt/intel/vtune_amplifier_xe_2013/bin64/../lib64/libwx_gtk2-2.8.so.0
#10 0x000000000044dc1b in ?? ()
#11 0x00007ffff3e4430d in __libc_start_main (main=0x44dbe0, argc=1, ubp_av=0x7fffffffdb18, init=<optimized out>, fini=<optimized out>,
rtld_fini=<optimized out>, stack_end=0x7fffffffdb08) at libc-start.c:226
#12 0x000000000041b3ea in ?? ()
#13 0x00007fffffffdb08 in ?? ()
#14 0x000000000000001c in ?? ()
#15 0x0000000000000001 in ?? ()
#16 0x00007fffffffdda0 in ?? ()
#17 0x0000000000000000 in ?? ()
(gdb)

 Because when I said: The same error, this implied, that the call stack was the same :)

[Michael_B_13]

I removed vtune amplifier completely and reinstalled Update8 -> no change

[Bernard]

Thanks:)

This exactly the same call stack like in previous user post.

>>>0x00007ffff5f3c996 in cfgmgr2::ConfigDescriptorRegistry::ConfigDescriptorRegistry(cfgmgr2::config_type_t, char const*) ()
from /opt/intel/vtune_amplifier_xe_2013/bin64/../lib64/libamplxe_cfgmgr_2.15.so
(gdb) backtrace
#0 0x00007ffff5f3c996 in cfgmgr2::ConfigDescriptorRegistry::ConfigDescriptorRegistry(cfgmgr2::config_type_t, char const*) ()
from /opt/intel/vtune_amplifier_xe_2013/bin64/../lib64/libamplxe_cfgmgr_2.15.so>>>

I suppose that this parameter  char const*  can be root cause of the problem.Can you resolve with gdb the address char pointer ?

[Michael_B_13]

Thanks. If you tell me how, i will try. Because I don't have really experience in Linux debugging. I normally work under Windows with Visual Studio.

[Bernard]

Sorry Michael I also do not have much experience in Linux debugging:( I am more experienced in Windows debugging.

[Bernard]

Maybe someone more experineced in Linux debugging can provide some help?

[robert-reed]

Michael B, are you also trying to install VTune Amplifier XE as a non-privileged user?  One person's failure is an anomaly.  Two persons' failures are a trend.  We need to understand the circumstances of these failures to try to identify the fault.  At this point it's not really clear.

[Bernard]

Hi robert,

do you know how to resolve with GDB memory addresses?

[Michael_B_13]

Hi Robert. I did two attemps. First with a sudo ./install.sh and the second one with a "sudo su" and then "./install.h". So I would say I tried it two times as a -privileged user.

[Bernard]

Michael B. wrote:

Hi Robert. I did two attemps. First with a sudo ./install.sh and the second one with a "sudo su" and then "./install.h". So I would say I tried it two times as a -privileged user.

What was the result?

[robert-reed]

Well, it seems that there is some common defect that is causing the configuration manager in VTune Amplifier to SEGV rather than handle the anomaly in an orderly fashion.  Unfortunately, my attempts to reproduce the symptoms on one of the machines available to me have so far ended in failure (i.e., failure to reproduce a fault--the code runs without a SEGV for me).  Worse, though both A. Rohou and Michael B. report failures with similar looking stack traces, nothing shared so far offers any clue about common factors between the two cases that might explain why these failures are occurring.  Moreover, all A. Rohou's attempts were following installs as a nonpriviledged user whereas Michael B. did the installations under sudo (which presumes that he had the rights to run sudo on this machine? ;-). So the unpriviledged install I think can be removed from the symptoms list.

 And yet we're still at this impasse, where I don't have enough information to reproduce the problem locally.  And so far A Rohou's attempts to gather more details have been stymied by a seemingly similar failure.  Michael B., do you also have that problem?  Have you been able to run amplxe-feedback as described previously in this thread?  Or does it also crash for you?

The stack traces above all seem to point to the configuration manager within VTune Amplifier but don't offer much in the way of diagnostics beyond that.  What we need is more insight into the environments where these failures have occurred.  We already know something of A. Rohou's environment (Intel processors, RHEL 5.8,...).  That all seems pretty normal.  Unfortunately.  I'll contact someone on the development team and show them what we know so far--maybe it will spark some insight--but without having sufficient information to reproduce the problem, it's very hard to investigate these failures.

[Alexis_R_]

@robert-reed: It might be helpful to know what information your feedback tool collects from the system. I could then get this information myself and forward it to you. I can imagine that the versions of some key libraries (glib etc) and whether some kernel modules are present would be needed, but if you could give me a list of things to look for, that'd help.

@ iliyapolak : I also do not know how to use gdb to get these values and I've been too busy to teach myself how to do so.

[Bernard]

@A.Rohou

I have found which command can be used to inspect the memory address beign pointed by pointer.I think that this command can be helpful

x/w address 
x/4b &variable 

[Michael_B_13]

Hi Robert,

sry I was away for a week. I will follow the steps to get informations.

We also use Intel machines (Sandybridge E26xx) on two socket boards with 128 GB RAM, running Ubuntu 11.

Update 1:

If I try the feedback tool, I get: 

Intel(R) VTune(TM) Amplifier XE 2013 (build 290588) feedback tool
Copyright (C) 2009-2013 Intel Corporation. All rights reserved.
Collecting pids tree
sh: /tmp/amplxe-log-root/2013-06-24-Mon-14-25-52-839477.amplxe-feedback/qfagent.cmd: not found

it creates a zip-file, but the *.log therein is empty.

[robert-reed]

I tried duplicating A Rohou's results when I first started looking at this thread but wasn't able to see a problem.  However, I tried to duplicate your symptoms, Michael B, with a new install of the latest update (9) and discovered that qfagent is found but fails to run for me.  I've reported it to the development team.  We should soon see a fix.