Lab 8: Debugging tools

Due: 5/24 11:59pm

Please read the instructions carefully

This lab may only be done on CSIL and it is an INDIVIDUAL ASSIGNMENT. Pair programming is not permitted.

Note also, that with each submission, Gradescope will use the umail address associated with your Gradescope account to randomize the answers it expects for this lab. Therefore, throughout the lab, you need to pass in your umail address as the command line argument for the C++ programs you are running (see instructions below).

So, if you are tempted to copy answers from another student, please, please, please don't (for obvious reasons). The expected answers for each student will be unique.

1. Background

Thus far you've learned about object oriented design and Test Driven Development. These techniques were designed to give structure to your code and your development process. This structure reduces the chance that a bug will sneak its way into your program. Even the most experienced programmer occasionally introduces bugs into their program. Given that a significant portion of your time will likely be spent debugging your code, we want to optimize the debugging process and make you more productive.

2. Goals

There are two sections of this lab.

In first section, Steps 1 (gdb) and 2 (valgrind), you are guided through learning about gdb and valgrind. There is nothing to turn in from these steps, so there may be a temptation to skip over them. We encourage you, instead, to really take your time and work through those, to really learn about gdb and valgrind, because you'll need those skills for step 3.

Step 3 requires you to work with gdb and valgrind, compute some answers, and put those answers in text files.

Those files will then be submitted to Gradescope for grading.

2.1. Getting Started

This lab must be done individually. Pair programming is not allowed. As mentioned above, each individual student's answers will be unique.

2.2. Step 1: gdb

The GNU debugger (a.k.a. gdb) is a very powerful tool for analyzing what your program is doing at run-time.

First read through the tutorial tools: gdb at the general CS32 course website ( The rest of these instructions assume that you already read through that. So if you continue without first looking over that, you are likely to not know how to proceed. If you then ask for help, we will first try to figure out: did you actually read through Tools: gdb? If you did not, we'll tell you: ok, go read it.

Since you are already familiar with the code in previous labs, you should use them to practice using gdb. You will need to recompile your lab 3 code with the -g option to use gdb on it. Add -g to the CXXFLAGS in your Makefile. Note that your code for that lab does not need to work completely for this step, you can debug a buggy program and that is the point of debugging.

The order of flags on the CXXFLAGS line does not matter. You can add it to the end or the beginning. For example, you can have

CXXFLAGS = -g -std=c++17 -Wall -Wextra -Wno-unused-parameter -Wno-unused-private-field

Every time you change your Makefile you should always do a make clean first. Rebuild your tests.

Now start gdb ./testStudent00 and set a breakpoint in Student::setName. (If you have no idea how to do that, it's probably because you didn't read though the tools: gdb tutorial first. Do that now.)

When you run the program (after setting the breakpoint, you type run), gdb will suspend the execution at the start of the setName method. You can then use various gdb commands to examine the state of the program, including what's on the stack and what's on the heap.

Before moving on, you should experiment with each of the following gdb commands:

  • help
  • break
  • display
  • print
  • next
  • step
  • backtrace
  • watch

There are examples for these commands on the tools: gdb page. Since you might not always know what you are looking for when debugging, experimentation is very important. It helps to know your code, but sometimes you may be debugging code that you didn't write.

Later steps in this lab will assume that you really did figure out how to use the commands above. So, if you are just skipping over this step, or skimming it, eventually you'll make your way back here… Just saying.

2.3. Step 2: valgrind

After you are comfortable using gdb and you have stepped through a few routines in your program now it's time to learn about valgrind. We talked a bit about valgrind, but there's more to learn, so please now read about valgrind here at the tools: valgrind article on the website.

Run memcheck on your lab 3 and ensure that you don't have any memory leaks or errors that were detected by valgrind. Ideally, you should fix any errors that valgrind reports in lab 3 as an exercise. (Note that if your program doesn't use new and/or delete, there probably won't be any.)

Run each test with –leak-check=full

valgrind --leak-check=full ./testStudent00
valgrind --leak-check=full ./testStudentRoll00

2.4. Step 3: Debugging someone else's code

It's time to put your new skills to the test. You will be given an executable file, but not all the source code.

It is possible that you may have a program that calls a library and you won't be given all the source code to the library or the library was built without the -g option. Here we are reversing this a little bit. You are given some code, but the code to the main function is not given to you. You will not be able to step through all of the code. In a sense, you can treat main function like a 'black box', which calls the code you are given. You don't need to know what the main function is doing, but you will have to answer a few questions about how it calls the code you are given.

If you visit the following link, you can browse the 4 files:

You should see a listing of several C++ programs. We are going to copy those into your ~/cs32/lab8 github repo all at once with the following command:

cp ~emre/public_html/cs32/code/lab8/* ~/cs32/lab8

2.4.1. Guidelines for Answering

To be sure that your answers pass the tests on Gradescope, be sure that you record them carefully. The tests on Gradescope are very picky about filenames, formatting, etc.

  • Do not add extra text to answers (e.g. reduce answers like The answer is: x. to only x )
  • If a question asks for you to write a string "my string", do not include the quotation marks (e.g. "my string" => my string )
  • All questions should be placed in separate files matching the question name (e.g. your answer to q1a should be placed in a file called q1a; DO NOT create q1a.txt, etc )
  • If a question asks you to enter a number 5,000 only include the digits without the comma (e.g. 5,000 => 5000)

2.4.2. HINT: Suggestion for Creating the Answer Files

You can use the echo command and redirect the output to a file. There is an example below. Note that the quotations in the echo command are not included in the file named 'testfile'. The cat command is just printing the file to the terminal so you can see that the contents were in fact written.

$ echo "100" > testFile
$ cat testFile

The source code we've given you is deliberately poorly written. Take your time and look through it, but note that these are examples of code that should be rewritten. However, that is not your job today. Instead, you will use debugging tools to understand why these programs are failing.

First run the segProgram, which is designed to segfault.

Segmentation fault (core dumped)

gdb was designed to make handling segfaults trivial. Use gdb to pry into segProgram and discover the reason why it is crashing. When using gdb, make sure you run gdb from the same directory as the program you are analyzing (gdb relies on the current working directory as a path to find the source code).

2.5. Question 1 (q1a, q1b)

Start up gdb and load in segProgram.

gdb ./segProgram

Now run the program with the run command followed by your umail address:

(gdb) run

It will stop as soon as it encounters the segfault. Note that when you use the command run [blah], it takes [blah] as the command line argument (i.e. the value of argv[1]). This is a necessary step in order to produce the answer Gradescope is expecting.

  • q1a: What index of the array is the program trying to access? Write your answer in q1a. That's q-one-a, not q-L-a.
  • q1b: What value is the program trying to store in the array when the fault occurred? Write your answer in file q1b

2.6. Question 2 (q2)

For the remainder of the questions use the debugProgram executable.

gdb ./debugProgram 

Remember to pass in your umail address when running through gdb in order to obtain the correct values.

(gdb) run

q2: What is the value of the string 'a' when debugFunction1 is called? Remember to omit the quotation marks in your answer. Write your answer to a file named q2. Hint: use the break feature to stop the program

2.7. Question 3 (q3)

q3: What is the value of the string 'a' when debugFunction1 is called a second time?

Remember to omit the quotation marks in your answer. Write your answer to a file named q3.

2.8. Question 4 (q4)

q4: How many recursive calls to recursiveFunction are in the backtrace (also called a stack trace) when variable a = 100 in =recursiveFunction? Put your answer in q4.

Hints: * Use backtrace (abbreviated bt), along with conditional breakpoints: break someplace if condition * What is the number of stack frames when the last instance of recursiveFunction is created? You may need to page through them. * Then think about how many of those stack frames are actually for the recursive function, as opposed to other functions that have been put on stack before or after.

2.9. Question 5 (q5a, q5b)

Hint: For this question, consider using the watch command.

  • q5a: The main function assigns values to watchdata[ ]. What is watchdata[25]'s new value after being assigned a new value by main. Put the new value of watchdata[25] (after it is modified by main) into the file q5a
  • q5b: In addition, what is the value of the variable watchsomevalue at the instant that watchdata[25] is updated? Please put the value of watchsomevalue at that moment into the file q5b.

2.10. Question 6 (q6)

q6: The function dataLost is leaking memory. How much data is lost in the dataLost function? Write your answer in a file called q6.


  • Run your program using valgrind with --leak-check=full
valgrind --leak-check=full ./debugProgram
  • Don't include data lost elsewhere.
  • Valgrind formats its numerical output with commas. Don't include commas in the answer.
  • Only submit the number. Don't use any units like Bytes.

3. Submitting via Gradescope

You will turn in your files for each question above (q1a, q1b, q2, q3, q4, q5a, q5b, q6).

The lab assignment "Lab8" should appear in your Gradescope dashboard in CS 32. If you haven't submitted anything for this assignment yet, Gradescope will prompt you to upload your files. For this lab, you are required to submit your files with your github repo.

If you already submitted something on Gradescope, it will take you to their "Autograder Results" page. There is a "Resubmit" button on the bottom right that will allow you to update the files for your submission.

If everything is correct, you'll see a successful submission passing all of the autograder tests.

Author: Mehmet Emre


The material for this class is based on Prof. Richert Wang's material for CS 32