Usually yes. Assuming your program
crashed outside GDB due to SIGSEGV
and left a core dump, you can:
A: find out which
instruction actually caused the
access violation:
(gdb) x/i $pc
This will usually be a memory access
instruction, e.g. "movl
$1,8(%eax)". What's important is
then what value does the register
which is supposed to point to valid
memory have.
B. find out value
of that register:
(gdb) p/x $eax
Often this would be 0 (you are
writing through a NULL pointer),
or some nonsense value, e.g.
0x32314043 (you've corrupted the
pointer, or overwrote it with an
ASCII string).
The GDB "info symbol" command will tell you which symbol
(if any) is near the given address.
Use the same "info symbol" command for addresses slightly smaller and slightly larger the address of your "target" variable.
Update:
The info symbol doesn't work on local (automatic) variables, because such variables don't have a (symbol table) symbol associated with them.
To find info about local variables, do "info locals". You can then print their addresses with
(gdb) print &a_local_variable
I don't know of any way to do the inverse (i.e. map an address of a local variable back to its symbolic name). However, if you only have a small number of locals, it is usually trivial to map an address into one of them "by hand". And if you have too many locals, that's a bad "code smell" -- you should probably refactor your code so that you don't.