topic Re: hpux B.11.31 core crash but not in B.11.23 in Operating System - HP-UX

hpux B.11.31 core crash but not in B.11.23

Meir Yanovich — Mon, 24 Nov 2008 10:16:09 GMT

Hello all
we compiling the application on :
HP-UX B.11.23 U ia64 with aCC HP C/aC++ B3910B A.06.15 linked with STLport 5.10

and when we run the application on B.11.23 machine , every thing is working great
but when we move the application to B.11.31
we get core that say :

1 0x4000000000aabe30:0 in stlp_std::basic_string,stlp_std::allocator >::_M_assign(char const*,stlp_std::allocator)+0x560 ()

how can we test what is the problem ?

Re: hpux B.11.31 core crash but not in B.11.23

Steven E. Protter — Mon, 24 Nov 2008 11:03:35 GMT

Shalom,

Install and use the gdb debugger for HP-UX 11.31 and analyze the core file.

use tusc to analyze the process as running if possible.

tusc is a system utility for hp-ux that mimics the functionality of strace on Linux.

SEP

Re: hpux B.11.31 core crash but not in B.11.23

Meir Yanovich — Mon, 24 Nov 2008 15:10:00 GMT

well we did try to analyze the core with gdb
we didn't found any thing except the problem with the :

#1 0x4000000000aabe30:0 in stlp_std::basic_string,stlp_std::allocator >::_M_assign(char const*,stlp_std::allocator)+0x560 ()

now we compiled the application on hpux 31 and
and run the application on hpux31 in our local server and its running great but when we move to different hpux 31
we still we getting this error .
what else we can try and find in the machine configuration that can impact ?

Re: hpux B.11.31 core crash but not in B.11.23

Meir Yanovich — Mon, 24 Nov 2008 15:42:34 GMT

also im getting this error in gdb :
#0 0x4000000003057620:0 in __milli_memmove+0x80 ()

Re: hpux B.11.31 core crash but not in B.11.23

Dennis Handly — Mon, 24 Nov 2008 19:45:54 GMT

>when we run the application on B.11.23 machine, every thing is working great
but when we move the application to B.11.31

11.31 should be forward compatible. If it fails it means your application is illegal and may have some uninitialized variables, or some problems with different inputs.

1 0x4000000000aabe30:0 stlp_std::basic_string,stlp_std::allocator >::_M_assign(char const*, stlp_std::allocator)+0x560
>how can we test what is the problem?

By printing out the variables that are used there. (Have you compiled with -g0?)
Or by looking at the machine code with:
(gdb) frame 1
(gdb) print *this
(gdb) disas 0x4000000000aabe30-16*20 0x4000000000aabe30+16*4
(gdb) info reg

Are you in a thread? You could have a thread stack overflow.

>we didn't found anything except the problem with the:

But did you look at the statement in question and look at all the variables?

>what else we can try and find in the machine configuration that can impact?

What does your application do? Does it look at your machine configuration?

>I'm getting this error in gdb:
#0 0x4000000003057620:0 __milli_memmove+0x80

This is a user error. You have bad addresses or a length in your call to memcpy or memmove.

What does this show?
(gdb) p /x $r32
(gdb) p /x $r33
(gdb) p /x $r34

Does it match what you see in frame 1?

Re: hpux B.11.31 core crash but not in B.11.23

Meir Yanovich — Tue, 25 Nov 2008 06:56:50 GMT

hello and thanks for the reply i will try to answer the questions
>Are you in a thread? You could have a thread stack overflow.
well i don't know .. how can i check this ?
>But did you look at the statement in question and look at all the variables?
yes we set all variables.
>What does your application do? Does it look at your machine configuration?
the application reads and process text files
in multi threaded mode. it does not read machine configuration just it local application configuration
just reminder we have the same machine
with the same configuration ( as we know .. )
and the application works fine .
im attaching the output of the gdb command in text file i do not know how to analyze it.
can you please take a quick look at it .
Thanks

Re: hpux B.11.31 core crash but not in B.11.23

Duncan Edmonstone — Tue, 25 Nov 2008 08:49:42 GMT

>> just reminder we have the same machine
with the same configuration ( as we know .. )
and the application works fine .

So have you compared the pathc level and product levels of the 2 systems?

If they're on the same network then:

sysdiff

will do a quick comparison for you.

Otherwise an "swlist -l product" output from both systems run through diff should show any differences.

HTH

Duncan

Re: hpux B.11.31 core crash but not in B.11.23

Dennis Handly — Tue, 25 Nov 2008 08:56:12 GMT

>well i don't know. how can I check this?

Provide the full stack trace.

>yes we set all variables.

Do they all have valid values? You need to print the contents of the string.

>reads and process text files in multi-threaded mode.

You might want to explain that? No more than one thread per file?

>I'm attaching the output of the gdb command

You need to redo this:
(gdb) bt
(gdb) frame 0
(gdb) disas $pc-16*20 $pc+16*4
(gdb) info reg

It appears you have a null pointer as a target. Relink your application with -z.

Re: hpux B.11.31 core crash but not in B.11.23

Meir Yanovich — Tue, 25 Nov 2008 11:35:40 GMT

Hello and thanks for the fast reply
well i cant do sysdiff because the second hp
in different place not related to our network×¥
also swlist -l product is problem because i have no root privileges but i emailed to the sysadmin and i hope i will have the diff shortly
about the questions:
>Do they all have valid values? You need to print the contents of the string.
which variable to print ? there is attached stack trace can you please point me where ?

>You might want to explain that? No more than one thread per file?
no its balk of files per thread

>Provide the full stack trace.
attached
>You need to redo this:
>(gdb) bt
>(gdb) frame 0
>(gdb) disas $pc-16*20 $pc+16*4
>(gdb) info reg
also in the attached file
Thanks

Re: hpux B.11.31 core crash but not in B.11.23

Meir Yanovich — Tue, 25 Nov 2008 11:54:22 GMT

im attaching the the swlist diff print
the test1 is the problematic machine
on the left.
thanks

Re: hpux B.11.31 core crash but not in B.11.23

Meir Yanovich — Tue, 25 Nov 2008 14:53:06 GMT

just small update i recompile the application with -Z and still the same error.

Re: hpux B.11.31 core crash but not in B.11.23

Dennis Handly — Tue, 25 Nov 2008 22:31:06 GMT

>which variable to print? there is attached stack trace can you please point me where?

The stack trace doesn't appear to have debug info. You should recompile with -g0.
Your stack trace shows you are in a thread: __pthread_bound_body
Are your strings in question, if global, protected by mutexes?

The function _M_assign is passing a NULL pointer to memcpy. It is trying to copy 1 byte.

What is the source line and variables used in frame 2, ActivationRecord::Clear?

>I'm attaching the the swlist diff

Nothing was obvious.

>I recompile the application with -Z and still the same error.

I said relink with -z. You can also use "chatr -z".