Solaris C++ program crash. Help on mdb debugging - c++

Question

I'm trying to debug a segmentation fault using mdb in solaris (as this is the only available option for me). The following is the core dump
`threading model: multi-threaded
status: process terminated by SIGSEGV (Segmentation Fault)
C++ symbol demangling enabled
librax.so void RajHistory_Backend::addMessage+0x58()
librax.sovoid RajErrorSink::output+0xe0()
librax.so void RajErrorSink_bg::DoWrite+9()
librax.so int RajErrorSink_Backend::run+0x128()
libmtcpp.so void*startThread+0x1e()
libc.so.1 _thr_setup+0x5b()
libc.so.1 _lwp_start()`
Code snippet
//
std::deque<char*> _queue;
//
void RajHistory_Backend::addMessage(char *msg)
{
_mutex.lock();
_queue.push_front(msg);
while ( _queue.size()-1 > _size )
{
char *b = _queue.back();
_queue.pop_back();
delete [] b;
}
_mutex.unlock();
}
I'm actually struggling to find the crash reason as I'm new to mdb . I did some debugging using this link
When I tried to print the variables in this method
> history_size/d
libenv.so`history_size:
libenv.so`history_size: 20
>msg/C
libc.so.1`msg:
libc.so.1`msg: H
>msg/s
libc.so.1`msg:
libc.so.1`msg: HGHcöHc°HÐÃUHåLeàLmèIüLuðH]ØAõL}øHì0
öIÖ|u5M
ötA>`
Does this mean the received message, char* msg is invalid?
How do I get the exact line in addMessage() method which caused the problem? Any hints on this how I can debug in mdb?

Answer 1

When mdb prints msg as libc.so.1`msg it’s indicating it found a value in libc, not in your code. Similarly, it’s showing libenv.so`history_size so found a similarly named entry in libenv. mdb won’t know the names of arguments to your functions.
You’ll have a much easier time if you build your binaries with debugging info (-g flag to the compiler) and use the source level debugger for your compiler (gdb if you're using gcc, dbx if you're using the Sun Studio or Solaris Studio compilers).