VoIP Mailing List Archives

[brian at freeswitch.org]

This looks like gcc is segfaulting can you provide me a complete backtrace of the core file that dumps from FreeSWITCH?

http://wiki.freeswitch.org/wiki/Reporting_Bugs


It sounds like you might have bad ram or bad hardware... gcc crashing is usually a sign something is really wrong with your machine.


/b

On Sep 21, 2009, at 5:46 PM, Tihomir Culjaga wrote:

[tculjaga at gmail.com]

hi Brian,

well, there is no coredump at all... and when i start FS with gdb it doesn't crash
I need to do some more testing and will come back to you.

T.

On Tue, Sep 22, 2009 at 1:22 AM, Brian West <brian@freeswitch.org (brian@freeswitch.org)> wrote:

[brian at freeswitch.org]

FreeSWITCH compiles with debug symbols by default but you showed an
output where GCC was segfaulting so you have bad ram or bad hardware
and I suspect that is your problem.

/b

On Sep 22, 2009, at 7:36 AM, Tihomir Culjaga wrote:

[anthony.minessale at g...]

see this from your own log?

make[2]: Entering directory `/opt/freeswitch-trunk/libs/pcre'
g++ -DHAVE_CONFIG_H -I.     -O2 -MT pcrecpp_unittest.o -MD -MP -MF .deps/pcrecpp_unittest.Tpo -c -o pcrecpp_unittest.o pcrecpp_unittest.cc
g++: Internal error: Segmentation fault (program cc1plus)
Please submit a full bug report.
See <file:///usr/share/doc/gcc-4.3/README.Bugs> for instructions.
make[2]: *** [pcrecpp_unittest.o] Error 1
make[2]: Leaving directory `/opt/freeswitch-trunk/libs/pcre'
make[1]: *** [all] Error 2
make[1]: Leaving directory `/opt/freeswitch-trunk/libs/pcre'
make: *** [libs/pcre/libpcre.la] Error 2


This is a FATAL error to have on your machine.
It's failing during the build.  This is your compiler crashing while trying to build the software.
This is very bad.
You most likely have a hardware failure and need to replace the machine or at the very least all of the memory chips.




On Tue, Sep 22, 2009 at 7:36 AM, Tihomir Culjaga <tculjaga@gmail.com (tculjaga@gmail.com)> wrote:

[tculjaga at gmail.com]

Hi Anthony,

it is not the machine ... and yep there was some memory related issue ... but this was caused by my module ....

So, to summarize.. i had two issues:

1. FS crashing without any notice (at 5 CPS)
2. Unable to recompile FS with gdb support

The first issue was actually related to "-hp" switch i was using in my startup script. With it, FS was crashing without any notice (even on low traffic) and regardless if i load my custom modules or not.
The second issue was related to many FS crashes having my module loaded... I found it later and fixed that.


So, after the machine cleanup I rebuild FS with gdb support without any issues.
Of course i sow this log .. but i didn't realize for a while... and after that i was fighting with crashes caused by "-hp" ... also, it was quite late as well ended up at 3 AM



Anyhow, the poit is; FS works well with my custom module. It just finished 2 mil. calls (with media) at 100 CPS having ~1600 simultaneous calls... well, thats something .



T.


On Tue, Sep 22, 2009 at 4:35 PM, Anthony Minessale <anthony.minessale@gmail.com (anthony.minessale@gmail.com)> wrote:

[brian at freeswitch.org]

The issue is you clearly show GCC crashing trying to compile
freeswitch which is BAD that indicates a larger problem with the
hardware or memory. Its physical issues not logical ones.

/b

On Sep 22, 2009, at 11:29 AM, Tihomir Culjaga wrote:

[diego.viola at gmail.com]

Why don't you try to do the same on another machine to see if you get the same results?

I think it's hardware related as Anthony and Brian pointed out.

Diego

On Tue, Sep 22, 2009 at 4:29 PM, Tihomir Culjaga <tculjaga@gmail.com (tculjaga@gmail.com)> wrote:

[tculjaga at gmail.com]

hmmm, how to track that down?
this is gonna be tricky...


i have another machine but quite different i can try on that as well and we will see....

T.

On Tue, Sep 22, 2009 at 6:41 PM, Brian West <brian@freeswitch.org (brian@freeswitch.org)> wrote:

[diego.viola at gmail.com]

Why don't you try changing RAM? Or run memtest86 or try another machine? Or...

On Tue, Sep 22, 2009 at 5:46 PM, Tihomir Culjaga <tculjaga@gmail.com (tculjaga@gmail.com)> wrote:

[anthony.minessale at g...]

One of the things that -hp does is call "memlockall" which disables swapping which uses more memory which makes hitting a land mine in your ram chip much more likely.

On the other hand:

Since you are talking about "with" and "without" gcc support I am going to guess you are on Solaris which you probably should have mentioned before.
it's possible that some of the more aggressive things activated by -hp is not possible on that platform.  If so we either have to identify that and disable it or disable hp completely for Solaris.

Either way, gcc randomly crashing is never ok and is a symptom of a pretty serious issue.

Are you using 2 separate fresh checkouts for both suncc and gcc builds because it's not possible to switch the same source tree once it's already configured for one of them.


On Tue, Sep 22, 2009 at 11:29 AM, Tihomir Culjaga <tculjaga@gmail.com (tculjaga@gmail.com)> wrote:

[tculjaga at gmail.com]

Hi,

Nope, I'm still on Debian 5.0... in transit to CentOS 5.3 but it needs to wait a bit.
i was talking about gdb, not gcc and was trying to recompile FS with debug symbols on: CFLAGS="-g -ggdb" MOD_CFLAGS="-g -ggdb".

yes, I understand that gcc segfault most probably means only one thing... HW isues. This is sometihng that I'm going to check tomorrow.... running memtest to see what i get. Also, I will repeat the same test with a new block of RAM.


Maybe i didn't explain myself well... apologize.

T.



On Tue, Sep 22, 2009 at 8:42 PM, Anthony Minessale <anthony.minessale@gmail.com (anthony.minessale@gmail.com)> wrote:

[diego.viola at gmail.com]

Doesn't FS already compiles with debug symbols by default?

On Tue, Sep 22, 2009 at 7:04 PM, Tihomir Culjaga <tculjaga@gmail.com (tculjaga@gmail.com)> wrote:

[brian at freeswitch.org]

yes

On Sep 22, 2009, at 2:32 PM, Diego Viola wrote:

[diego.viola at gmail.com]

He's doing an extra effort... just compile it as you would normally and you will have the debug symbols.

On Tue, Sep 22, 2009 at 8:11 PM, Diego Viola <diego.viola@gmail.com (diego.viola@gmail.com)> wrote:

[diego.viola at gmail.com]

Then why is Tihomir trying to compile with debug symbols?

On Tue, Sep 22, 2009 at 8:00 PM, Brian West <brian@freeswitch.org (brian@freeswitch.org)> wrote: