gdb (and ddd) Guide

gdb and ddd for debugging C and C++ programs


About gdb and ddd
Getting Started with gdb
Common Comands

Sample gdb sessions

the info command for getting application and debugger state
Setting conditional breakpoints and setting breakpoints in C++ methods
Keyboard shortcuts in gdb
invoking make in gdb
Advanced features debugging threaded programs, attaching to an already running process, debugging a child process after a fork, signal handling

Assembly code debugging (using gdb to debug at the assembly code level, and how to examine memory and register values)

Links to more gdb information
Some settings (and bug fixes) for ddd

Introduction to gdb and ddd
The purpose of a debugger is to allow you to see what is going on inside your C program while it runs. In addition, you can use gdb to see what your program was doing at the moment it crashed.

Here are some of the usful actions that gdb can perform:

For C and C++ programs, gdb and ddd are debuggers that you can use. ddd is a easy-to-use GUI wrapper around an inferior debugger (gdb for GNU compiled C or C++ code). ddd allows you to interact with the debugger by using either GUI menu options or the under-lying debugger's command line interface. In addition, ddd automatically displays source code when breakpoints are reached.

There are some example programs and some documentation on using gdb to debug them that you can copy from here: /home/newhall/public/gdb_examples/

Getting Started with gdb
C and C++ programs compiled with the GNU compiler and the -g option can be debugged using GNU's debugger gdb (actually, you can use gdb on code that is not compiled with -g, but unless you like trying to figure out how assembly code sequences map to your source code I wouldn't recommend doing so). Also, do not compile with an optimization flag (i.e. don't use -O2), or gdb will have a hard time mapping optimized machine code to your source code. For example:
% gcc -g myprog.c

To start gdb, invoke gdb on the executable file. For example:

% gdb ./a.out
If your program terminates with an error, then the operating system will often dump a core file that contains information about the state of the program when it crashed. gdb can be used to examine the contents of a core file:
% gdb core ./a.out
One good way to get started when you are trying to track down a bug, is to set breakpoints at the start of every function. In this way, you will quickly be able to determine which function has the problem. Then you can restart the program and step through the offending function line-by-line until you locate the problem exactly.

ddd is invoked in a similar way:

% ddd ./a.out

Common gdb Commands
An abreviated printable version of these is here

gdb also understands abreviations of commands, so you can just type up to the unique part of a command name ("cont" for "continue", or "p" for "print", "n" for next, "s" for step, ...)

info commands for examining runtime and debugger state:
gdb has a large set of info X commands for displaying information about different types of runtime state and about debugger state. Here is how to list all the info commands in help, and a description of what a few of the info commands do:
(gdb) help info             # lists a bunch of info X commands
(gdb) help status           # lists more info and show commands

(gdb) info frame            # list information about the current stack frame
(gdb) info locals           # list local variable values of current stack frame
(gdb) info args             # list argument values of current stack frame
(gdb) info registers        # list register values
(gdb) info breakpoints      # list status of all breakpoints

using gdb to debug assembly code and examine memory and register values:
ddd is probably easier to use when steping through assembly code than gdb because you can have separate windows that show the disassembled code, the register values, and the gdb prompt.

Here are some gdb commands that are useful for debugging at the assembly code level (the Common Comands section has more general details about some of these commands, particularly formatting options for print and x):

disass   list the assembly code for a function or range of addresses
  disass <func_name>           lists assembly code for function
  disass <start>  <end>  lists assembly instructions between start and end address 

break   Set a breakpoint at an instruction 
  break *0x80dbef10   Sets breakpoint at the instruction at address 0x80dbef10

stepi, si   Execute next machine code instruction, 
             step into function call if it is a call instr
nexti,  ni  Execute next machine code instruction, 
             treats function call as single instr

   info registers        # list register values

   print $eax    # print the value stored in the eax register
   print *(int *)0x8ff4bc10  # print int value stored at memory addr 0x8ff4bc10

x     Display the contents of the memory location given an address.  
         NOTE: the format is sticky (need to explictly change it)

   x $ebp-4          # examine memory at address: (value stored in register ebp)-4
                     # if this location stores an address x/a, an int x/wd, a ...
   x/s 0x40062d      # examine the memory location 0x40062d as a string
     0x40062d   "Hello There"
   x/4c 0x40062d     # examine the first 4 char memory locations 
                     # starting at address 0x40062d
     0x40062d   72 'H'	101 'e'	108 'l'	108 'l'
   x/d 0x40062d     # examine the memory location 0x40062d in decimal 
     0x40062d   72    # NOTE: units is 1 byte, set by previous x/4c command
   x/wd 0x400000     # examine the memory location 0x400000 as 4-byte in decimal 
     0x400000   100    # NOTE: units was 1 byte set, need to reset to w 

set set the contents of memory locations and registers
   set $eax = 10                 # set the value of register eax to 10
   set $esp = $esp + 4           # pop a 4-byte value off the stack 
   set *(int *)0x8ff4bc10 = 44   # at memory address 0x8ff4bc10 store int value 44 

display  at every breakpoint display the given expression
   display $eax

Quick summary of common commands for assembly debugging

  ddd a.out
  (gdb) break main
  (gdb) run  6              # run with the command line argument 6
  (gdb) disass main         # disassemble the main function
  (gdb) break sum           # set a break point at the beginning of a function
  (gdb) cont                # continue execution of the program
  (gdb) break *0x0804851a   # set a break point at memory address 0x0804851a
  (gdb) ni                  # execute the next instruction
  (gdb) si                  # step into a function call (step instruction)
  (gdb) info registers      # list the register contents
  (gdb) p $eax              # print the value stored in register %eax
  (gdb) p  *(int *)($ebp+8) # print out value of an int at addr (%ebp+8)
  (gdb) x/d $ebp+8          # examine the contents of memory at the given
                            # address (/d: prints the value as an int)
                            # display type in x is sticky: subsequent x commands
                            # will display values in decimal until another type
                            # is specified (e.g. x/x $ebp+8   # in hex)
  (gdb) x/s 0x0800004       # examine contents of memory at address as a string
  (gdb) x/wd 0xff5634       # after x/s, the unit size is 1 byte, so if want
                            # to examine as an int specify both the width w and d
Here is some more information about assembly-level debugging in gdb, including a simple example session: Debugging IA32 Assembly Code with gdb (and ddd)

Sample gdb sessions
Below is output from two runs of gdb on programs from ~newhall/public/gdb_examples/.

  1. Run 1 is a gdb run of badprog.c. It demonstrates some common gdb commands, and it finds one of the bugs in this program...there are others.

  2. Run 2 is a gdb run of segfaulter.c. It demonstrates how to find out where your program is segfaulting (and perhaps why...although valgrind will help more with this type of error).

Run 1: badprog.c

This is an example run of gdb on badprog.c. It demonstrates commonly used gdb commands in a debugging session.
% gcc -g badprog.c           #-- compile program with -g flag

% gdb a.out                  #-- invoke gdb with the executable

GNU gdb 6.4.90-debian
Copyright (C) 2006 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for details.
This GDB was configured as "i486-linux-gnu"...Using host libthread_db 
library "/lib/tls/".

(gdb) break main         #-- set a breakpoint at start of the program's execution

Breakpoint 1 at 0x8048436: file badprog.c, line 36.

(gdb) run                     #-- run the program

Starting program: /home/newhall/public/gdb_examples/a.out 

Breakpoint 1, main () at badprog.c:36      #-- gdb stops at breakpoint	

36	  int arr[5] = { 17, 21, 44, 2, 60 };

(gdb) list                         #-- list the source code near the break point
31	  return 0;
32	}
34	int main(int argc, char *argv[]) {
36	  int arr[5] = { 17, 21, 44, 2, 60 };
38	  int max = arr[0];
40	  if ( findAndReturnMax(arr, 5, max) != 0 ) {

(gdb) list 11                            #-- list source code around line 11

11	// this function should find the largest element in the array and
12	// "return" it through max
13	// 	array: array of integer values
14	// 	len: size of the array
15	// 	max: set to the largest value in the array 
16	//  	reuturns: 0 on success and non-zero on an error
17	//
18	int findAndReturnMax(int *array1, int len, int max) {
20	  int i;

(gdb) list                              #-- list the next few lines of code
22	  if(!array1 || (len <=0) ) { 
23	    return -1; 
24	  }
25	  max = array1[0];
26	  for(i=1; i <= len; i++) {
27	    if(max < array1[i]) { 
28	      max = array1[i];
29	    }
30	  }

(gdb) next                             #-- execute the next instruction
38	  int max = arr[0];

(gdb)                          #-- hitting Enter executes the previous command
                               #--  (next in this case)
40	  if ( findAndReturnMax(arr, 5, max) != 0 ) {

 #-- also you can use the up and down arrows to scroll through previous commands

(gdb) print max                        #-- print out the value of max
$1 = 17

(gdb) p arr                            #-- p is short for the print command
$2 = {17, 21, 44, 2, 60}

(gdb) step        #-- step into the function call
                  #-- if had entered 'next', entire function call would be executed
                  #-- 'step' takes us to the entry point of findAndReturnMax

findAndReturnMax (array1=0xbfc5cb3c, len=5, max=17) at badprog.c:22
22	  if(!array1 || (len <=0) ) {             

(gdb) print array1[0]                   #-- lets see what the param values are
$3 = 17
(gdb) p max 
$4 = 17

(gdb) list
17	//
18	int findAndReturnMax(int *array1, int len, int max) {
20	  int i;
22	  if(!array1 || (len <=0) ) { 
23	    return -1; 
24	  }
25	  max = array1[0];
26	  for(i=1; i <= len; i++) {

(gdb) break 26          #-- set a breakpoint at line 26 (inside findAndReturnMax)

Breakpoint 2 at 0x80483e7: file badprog.c, line 26.

(gdb) cont                                #-- continue the execution

                                          #-- gdb hits the next breakpoint 
Breakpoint 2, findAndReturnMax (array1=0xbfc5cb3c, len=5, max=17)    
    at badprog.c:26
26	  for(i=1; i <= len; i++) {

(gdb) p i
$5 = 0

(gdb) n                                   #-- n is short for next
27	    if(max < array1[i]) { 

(gdb) display max          #-- display prints out value everytime a breakpoint hit
1: max = 17                                      

(gdb) display array1[i]
2: array1[i] = 21

(gdb) break 27                            #-- set a breakpoint inside the loop
Breakpoint 3 at 0x80483f0: file badprog.c, line 27.

(gdb) cont                                #-- continue execution

Breakpoint 3, findAndReturnMax (array1=0xbfc5cb3c, len=5, max=21)
    at badprog.c:27
27	    if(max < array1[i]) { 
#-- display prints these out:
2: array1[i] = 44                        
1: max = 21

(gdb) cont

Breakpoint 3, findAndReturnMax (array1=0xbfc5cb3c, len=5, max=44)
    at badprog.c:27
27	    if(max < array1[i]) { 
2: array1[i] = 2
1: max = 44

(gdb) cont

Breakpoint 3, findAndReturnMax (array1=0xbfc5cb3c, len=5, max=44)
    at badprog.c:27
27	    if(max < array1[i]) { 
2: array1[i] = 60
1: max = 44

(gdb) cont

Breakpoint 3, findAndReturnMax (array1=0xbfc5cb3c, len=5, max=60)
    at badprog.c:27
27	    if(max < array1[i]) { 
2: array1[i] = 17
1: max = 60                     #-- so max is 60 here

(gdb) where                 #-- show the stack frames 

#-- findAndReturnMax is active function at line 27, was called by main at line 40:
#0  findAndReturnMax (array1=0xbfd043ec, len=5, max=60) at badprog.c:27
#1  0x08048479 in main () at badprog.c:40

 frame 1  #-- move into main's calling context (stack frame 1) to examine main's state
#1  0x08048479 in main () at badprog.c:40
40        if ( findAndReturnMax(arr, 5, max) != 0 ) {

(gdb) print max                  #-- in main's stack frame max is 17
$1 = 17

(gdb) cont                       #-- continue execution
max value in the array is 17     #-- main prints out  max after function call 

#-- This looks like a bug:" 
#-- findAndReturnMax set max to 60, but 60 isn't getting "passed back"  
#-- to main after the call.  To fix this we need to either have 
#-- findAndReturnMax return the value of max or pass max by reference

(gdb) quit                      #-- quit gdb

The program is running.  Exit anyway? (y or n) y

Run 2: segfaulter.c

This is an example run of gdb on the segfaulter.c program that segfaults when run.


  % gdb segfaulter

  (gdb) run              #-- just run segfaulter and let it seg fault

 Program received signal SIGSEGV, Segmentation fault.
 0x080483e1 in initfunc (array=0x0, len=100) at segfaulter.c:15
 15          array[i] = i;

  (gdb) where           #---  let's see where it segfaulted

 #0  0x080483e1 in initfunc (array=0x0, len=100) at segfaulter.c:15
 #1  0x0804846e in main () at segfaulter.c:38

 (gdb) list             #--- let's see code around segfaulting instruction 

 10      int initfunc(int *array, int len) {
 12        int i;
 14        for(i=1; i <= len; i++) {
 15          array[i] = i;
 16        }
 17        return 0;
 18      }

  (gdb) p array[0]     #--- let's print out some values and see what's going on
 Cannot access memory at address 0x0

 #-- it looks like array is a bad address (0x0 is NULL)

  (gdb) p array
 $1 = (int *) 0x0

  (gdb) frame 1             #--- let's see what main is passing to this funtion

 #1  0x0804846e in main () at segfaulter.c:38
 38        if(initfunc(arr, 100) != 0 ) {

  (gdb) print arr           #--- print out arr's value (what we pass to initfunc) 
 $2 = (int *) 0x0
 #--- oops, we are passing NULL to initfunc...
 #--- we forgot to initialize arr to point to valid memory

Keyboard Shortcuts in gdb
gdb supports command line completion; by typing in a prefix you can hit TAB and gdb will try to complete the command line for you.

Also, you can give just the unique prefix of a command as the command and gdb will execute it. For example, rather than entering the command print x, you can just enter p x to print out the value of x.

The up and down arrow keys can be used to scroll through previous command lines, so you do not need to re-type them each time.

If you just hit RETURN at the gdb prompt, gdb will execute the most recent previous command again. This is particularly useful if you are steping through the execution, then you don't have to type next each time you want to execute the next instruction, you can just type it one time and then hit RETURN.

Setting conditional breakpoints and some issues with setting breakpoints in C++ code

conditional breakpoints

A conditional breakpoint is one that only transfers control to gdb when a certain condition is true. This can be very useful when you only want gdb control after iteration 1000 of a loop, for example.

To set a condition on a breakpoint, use the condition command with the number of the breakpoint followed by the condition on which to trigger the breakpoint. Here is an example where I'm setting a conditional breakpoint in the loops.c program that will only be triggered when the condition (i >= 150) is true:

(gdb) break 28                                  # set breakpoint at line 28
(gdb) info break                                # list breakpoint information
  Num Type           Disp Enb Address    What
   1   breakpoint     keep y   0x080483a3 in foo at loops.c:28

(gdb) condition 1 (i >= 150)                # set condition on breakpoint 1
(gdb) run   (or continue if already running) 

Breakpoint 1, foo (val=200) at loops.c:28     # breakpoint 1 is reached
28	    if(i > val)
(gdb) print i                                 # only when (i >= 150) 
$1 = 150

breakpoints in C++ programs

One complication with gdb and C++ programs, is that you need to specify methods and data members using the "classname::" prefix. In addition, you often need to use a leading ' before a name for gdb to find the symbol, and if methods are overloaded, you need to specify which method it is by listing its full prototype (actually, if you hit TAB gdb will list all possible matches for you and you can pick one of those).

For example, to set a break point in funciton pinPage of the BufMgr class, I'd do the following:

	(gdb) break 'BufMgr::pinPage(int, Page *&, int)'
This looks pretty icky, but really I just type break 'BufMgr::p then hit TAB for automatic completion.
	(gdb) break 'BufMgr:: <tab> 
will list all methods of the BufMgr class, then you can just pick from the list the method you want to put the breakpoint in.

gdb and make
Within gdb you can invoke "make" to rebuid your executable (assuming that you have a makefile to build your program). This is a nice feature in the case when you have many breakpoints set and do not want to exit gdb, recompile, re-start gdb with the new a.out, and reset all the breakpoints. However, keep in mind that modifying and recompiling your source code from within gdb may result in your breakpoints not being where you think they should be (adding/removing lines of source code can result in your in your breakpoints no longer being where you want them to be in terms of the new version of your source code). You can use the disable or delete commands to disable or delete old breakpoints.

Some Advanced Features

debugging multi-threaed processes in gdb

Here are some basics about using gdb to debug pthread programs, and some general advice for debugging multi-threaded processes, and some links to other gdb and pthreads resources: debugging pthreads programs

attaching gdb to a running process

  1. get the process's pid
       # ps to get process's pid (lists all processes started in current shell):
       $ ps                     
       # list all processes pipe through grep for just those named a.out:
       $ ps -A | grep a.out     
         PID TTY          TIME CMD
       12345 pts/3     00:00:00 a.out
  2. attach gdb to the running process
       # gdb <executable> <pid>
       $ gdb a.out 12345
    # OR alternative syntax:
       # gdb attach <pid>  <executable>
       $ gdb attach 12345 a.out
    At this point the process is stopped by gdb; you have the gdb prompt that you can use issue gdb commands like setting breakpoints, or printing out program state before continuing execution.
Depending on if the process was explicitly stopped before attaching gdb or not (e.g. did the process call kill(getpid(), SIGSTOP) to stop itself like in the attach_example.c) you can continue its execution from the gdb prompt in one of two ways:
  (gdb) cont                 # try this first
  (gdb) singal SIGCONT       # try this if the cont command doesn't work

following a process on a fork

You can set gdb to follow either the parent or the child process on a fork (the default is to follow the parent): By setting breakpoints in the child path of code, and then doing:
(gdb) set follow-fork-mode child
gdb will follow the child process after the fork. See debugging forks for more information.

signal control

In gdb you can send the process a signal:
(gdb) signal SIGCONT
(gdb) signal SIGALARM 
Sometimes your process receives signals and you would like to have gdb perform some action when certain signals are delived to the debugged process. For example, if your program issues a bad adress, it will receive a SIGBUS signal and usually exit. The default behavior of gdb on a SIGBUS it to let the process exit. If, however, you want to examine program state when it recieves a SIGBUS, you can specify that gdb handle this singal differently:
  (gdb) handle SIGBUS stop    # if program gets a SIGBUS, gdb gets control 
You can list how gdb is handling signals using info:
  (gdb) info signal    # list info on all signals

  (gdb) info SIGALRM   # list info just for the SIGALRM signal

ddd settings and bug fixes
Running ddd creates a .ddd directory in your home directory and will save settings to files here, so that you don't need to reset all your preferences from scratch. You can click and drag to change the sizes of subwindows and choose Menu options to display (or not) certain menus, register values, machine code, etc.
common gdb commands (from above)
example gdb sessions (from above)
GDB quick reference card
A very complete GDB reference
Using GDB within Emacs by Ali Erkan