Data breakpoints are now becoming a part of common breakpoint vocabulary. They help in detecting heap corruption, inadvertent data overwrites and writing past buffer boundaries.
Most programmer’s restrict the definition of data breakpoints to breakpoints that help halting the execution of code in the debugger when memory is written to. This is the kind of breakpoint that helps in catching most of the corruption bugs.
GDB provides data breakpoints (at least on Intel platforms) that do slightly more than that. As data breakpoints are implemented through hardware assistance, it is the hardware platform that provides the different kinds of data breakpoints and the debuggers provide the required interface.
watch
watch is gdb’s way of setting data breakpoints which will halt the execution of a program if memory changes at the specified location.
watch breakpoints can either be set on the variable name or any address location.
watch my_variable
watch *0x12345678
where 0x12345678 is a valid address.
Usually a crash because of heap corruption or invalid outcome due to buffer overruns shows up in the debugger when it is too late to figure out what went wrong. The watch or write data breakpoints can be used to find when a memory location has changed. The debugger at that very instant shows the reason for the inadvertent change.
The cause usually is double deletion of memory, writing to deleted memory, writing past the buffer boundary, etc. In order to fix such issues, it is more important to know when the corruption happens than to know what happens when the corruption has taken place.
Another interesting use of write data breakpoints is to find out the cause of memory leaks in reference counted objects by monitoring the increase and decrease in the reference count of the objects.
rwatch
rwatch (read-watch) breakpoints break the execution of code when the program tries to read from a variable or memory location.
rwatch iWasAccessed
rwatch *0x12345678
where 0x12345678 is a valid address.
For example, say you are new to your project and would like to figure out where exactly are your encryption routines in code. As a start, you can first search your codebase but you may hit a few false positives. If you know the memory location of your password and you set your rwatch breakpoints correctly, it would not be long before the debugger breaks execution right in your encryption algorithms which have to read the password in order to perform their function. Getting evil ideas already?
Another use of read data breakpoints is finding out the code that is reading from memory that has already been deleted and is using this corrupt information later.
awatch
awatch or access watches break execution of the program if a variable or memory location is written to or read from. In summary, awatches are watches and rwatches all in one. It is a handy way of creating one breakpoint than two separate ones.
awatch *0x12345678
where 0x12345678 is a valid address.
The problem with data breakpoints like any other breakpoint is that they can be triggered far too many times and the programmer may lose track of the problem being debugged. To ensure the breakpoints are hit the least number of times, the data being worked upon should be minimal and the breakpoint should be set as late as possible.
Line breakpoints can be implemented either in hardware or software. This article discusses the latter in detail.
It is very useful to be able to break execution of code at a line number of your choice. Breakpoints are provided in debuggers to do exactly that. It is fun getting to the root of the problem by setting breakpoints in a debug session. It is even more fun to know how do breakpoints work in the first place.
Software breakpoints work by inserting a special instruction in the program being debugged. This special instruction on the Intel platform is “int 3”. When executed it calls the debugger’s exception handler.
Example
Let us look at a very simple example that inserts a breakpoint in a program at compile time and not through a debugger. The code uses the Intel instruction “int 3” and you may need to figure out the equivalent instruction for a non-Intel platform.
// The code below works well with Visual Studio.
int main()
{
__asm int 3;
printf("Hello World\n");
return 0;
}
// The code below works well with gcc + gdb
int main()
{
asm("int $3");
printf("Hello World\n");
return 0;
}
If you run this program in Visual Studio, you get a dialog saying “helloworld.exe has triggered a breakpoint“.
In gdb you get the message “Program received signal SIGTRAP, Trace/breakpoint trap.”
In the example above, a call to “int 3” invokes the debugger’s exception handler.
It is also interesting to note the assembly instructions generated for the program above.
In Visual Studio, right click on the code and click on “Show Disassembly”. Also ensure that “Show Code Bytes” is on in the same context menu.
In gdb type disassemble at the gdb command.
disassemble
0x0040107a <main+42>: int3
Now obtain the opcode of the int3 instruction using the x (examine memory) command
(gdb) x/x main+42
0x40107a <main+42>: 0xcc
As seen above, the breakpoint opcode we inserted during compilation is 0xCC .
How Do Debuggers Insert Breakpoints?
For practical reasons, it is unwise to ask for a recompilation whenever a breakpoint is added or deleted. Debuggers change the loaded image of the executable in memory and insert the “int 3” instruction at runtime. The common steps a debugger performs to provide the functionality of a line breakpoint to a user are as follows –
When a user inserts a breakpoint in a line of code, the debugger saves the opcode at that given location and replaces it with 0xCC (int 3).
When the program is run and it executes the “int 3” instruction, control is passed to the debugger’s exception handler.
The debugger notifies the user that a breakpoint has been hit. Say that the user instructs the debugger to resume execution of the program.
The debugger replaces the opcode 0xCC with the one it had saved earlier. This is done to restore the instructions to their original state.
The debugger then single steps the program.
It then resaves the original instruction and re-inserts the opcode 0xCC. If this step were not done, the breakpoint would have been lost. Temporary breakpoints on the other hand skip this step.
The debugger then resumes execution of the program.
Hardware breakpoints are limited in number but debuggers are able to provide unlimited breakpoints by implementing them through software.
Knowing what goes behind the scenes makes debugging a bit easier. A debugger may defer setting a breakpoint if the module is not loaded in memory yet. It needs to replace some opcode with 0xCC and that can happen only when the module is in memory. Likewise, a mismatch between a binary, its sources and its debug symbols (or the lack of it) may cause breakpoints to be hit at unexpected locations because the debugger is not able to correctly map the source line to the opcode that it needs to replace with 0xCC. At times debuggers complain about the mismatch and refuse to set the breakpoints.
Many of the setup issues with breakpoints become obvious once we know how they work internally. And when all else fails and release build breakpoints adamantly refuse to work, you always have the option of compiling an “int 3” breakpoint right into your code.
Have you faced the problem of breakpoint clutter where breakpoints keep piling up only to hinder the debugging session? It is then that one realizes that there are some breakpoints that can be deleted and others disabled.
A useful feature in a debugger is a temporary breakpoint that automagically gets deleted when hit thereby reducing the clutter of unnecessary breakpoints. These breakpoints are useful when you wish to stop at a code location only once and do not require the execution to stop at that location ever again.
For example, say you are trying to determine whether a particular test scenario invokes a specific line of code or not, in that case a temporary breakpoint can be used as the breakpoint is not useful once it has been hit alteast once.
Below are steps on how to set temporary breakpoints in various debuggers.
gdb
Use thetb command to set a temporary breakpoint in gdb. It is similar to the break command but the breakpoint will automatically be deleted when hit.
(gdb)help tb
Set a temporary breakpoint.
Like “break” except the breakpoint is only temporary,
so it will be deleted when hit. Equivalent to “break” followed
by using “enable delete” on the breakpoint number.
Windbg
In Windbg, breakpoints set in the Command window using the bl /1 command can be used to create temporary breakpoints. The /1 tells Windbg that the breakpoint should be deleted when hit.
In Windbg temporary breakpoints are also known as “one shot breakpoints”.
Visual Studio
I found it a bit painful to create temporary breakpoints in Visual Studio. The only way I could create one was by setting a breakpoint and then setting the hit count for the breakpoint to be equal to 1. The article here explains how to set a hit count in Visual Studio.
The amount of work involved to do this sometimes doesn’t make temporary breakpoints worthwhile to set. Moreover the breakpoint lingers on and doesn’t actually get deleted when hit.
As a developer, the first step for solving any bug is to reproduce it. This is an important step before the investigation actually begins. Once the issue is reproduced, the developer starts investigating the bug with the knowledge, tools and debugging skills at hand. The source of the problem is then determined, a fix is proposed and later checked in after completing the unit testing and other required processes followed by the team. This may sound very simple but there are obstacles at every step unless the bug is very trivial. This is routine work for any seasoned software developer.
One mistake developers often make is to jumpstart the debugging process sooner than needed. They fire up their favorite debugger, set the breakpoints and then start debugging when actually they should have first reduced the data set required for reproducing the bug.
The rule of debugging is very simple –
“Every developer should strive to reproduce the bug by hitting the required set of breakpoints the least number of times. This allows faster and efficient debugging sessions.”
So if you are hitting a breakpoint hundred times more than you should have, then you are either not using the debugger efficiently or you have not worked enough to reduce the data required to reproduce the bug.
The latter problem is more common as sometimes developers put in little effort to reduce the input data and start debugging prematurely. This leads to longer debugging cycles where a lot of time is wasted on investigating code that is not even relevant to the problem being solved.
For example, if a there is 1MB text file that is processed by your program and crashes it, then one should first try minimizing the text file such that if it is reduced any further, the crash goes away. The text file obtained in such a manner is the smallest input data set required to reproduce the crash. Once this goal is achieved, the program will now be processing lesser data and effectively less code will be executed which results in important breakpoints being hit fewer times.
Well how does the developer go about reducing the input data set? There is no single method but the one that commonly works is to run a binary search on the input data. For example in the above case, the text file should be split into two. Two text files of half MB each would be obtained and one should test to see if the crash reproduces with either one of the files. If you still see the crash, you have halved the data set. Then the smaller file should be again split into two and the process should be repeated till a very small text file is obtained that still causes the crash to occur. Depending on what your program does, you can even reduce the 1 MB text file to a single character file. Debugging your program with a single character file is much simpler than using the initial 1MB file.
Reducing The Data Set
The Caveat
Sometimes a large data set may have multiple issues. By reducing it as described above, one may solve a partial problem but other problems may go unnoticed in the reduced set. Therefore once a bug has been solved on a reduced input data set, it should be tested against the one provided with the original bug. This ensures that no other issue that should have been fixed got ignored in a bid to made debugging more efficient.
Final Note
Reducing the input data is essential before starting the debugging process and a great productivity aid too. If possible, this should be a part of the bug reporting process for quality engineers or customers who often log the issue. Reducing the data may not always be possible but it is certainly worth an attempt.
If you are familiar with hit count breakpoints already, you may want to click here to jump to the advanced tricks shared in this article.
What is the hit count of a breakpoint?
A debugger allows users to set a breakpoint at a specific line in code. When the execution reaches that line, the breakpoint is said to have been *hit* and the execution of program being debugged is suspended.
Internally the debugger also keeps a count of the number of times the breakpoint has been hit. This is called the hit count of a breakpoint. Debuggers allow users to set conditions based on the hit count of the breakpoint. For example, you can specify that the execution of the program should only be suspended when the hit count is greater than or equal to 250. To put it in other words, the breakpoint will be skipped for the first 249 times it is hit.
The advantage of being able to set a condition with the hit count of a breakpoint is to make the process of debugging faster.
How can hit count based breakpoints be set?
Debuggers today have either a command line or a graphical user interface. Mostly all debuggers provide a means to set hit count based breakpoints. Below are steps on setting such breakpoints in some of the debuggers I have used.
Microsoft Visual Studio 2005
Set a breakpoint at a line in your code.
Right click the breakpoint and then click on “Hit Count”. You can also go to Debug -> Windows -> Breakpoints and right click on the breakpoint that was just created and select “Hit Count”.
In the dialog that pops up, you can choose from four ways of controlling the breakpoint based on its hit count. The default is to ignore the hit count and suspend the program always when the breakpoint is hit. It is good to take note of the other three options.
Hit Count Window In Visual Studio 2005
When the program is in suspended mode, one can see the current hit count of the breakpoint in the breakpoint window. In the image below, the “Hit Count” column shows the current hit count of the breakpoints.
gdb
In gdb, the command continue is used to resume execution of the suspended program. When followed by a number N, the breakpoint is hit the Nth time.
(gdb)help continue Continue program being debugged, after signal or breakpoint.
If proceeding from breakpoint, a number N may be used as an argument,
which means to set the ignore count of that breakpoint to N – 1 (so that
the breakpoint won’t break until the Nth time it is reached). (gdb)continue 20
gdb is available on Mac OSX, Linux, AIX, Solaris, HPUX and Cygwin on Windows, etc so this is one command one should learn by heart.
On Mac OSX, the XCode IDE uses gdb internally and allows access to it through the menu (Debug -> Console Log). Through the command line interface continue can be used as described above.
In gdb the info breakpoints command can be used to view the current hit count of all breakpoints.
(gdb)info breakpoints
Num Type Disp Enb Address What
1 breakpoint keep y 0x0040118a in main at try.cpp:6
breakpoint already hit 246 times
3 breakpoint keep y 0x004011a5 in main at try.cpp:8
Visual Studio and gdb differ slightly in terminology. One allows setting breakpoints with a hit count and the other lets skipping of breakpoints for a certain count. However they are essentially the same features that allows the programmer the option of not having to stop always at a breakpoint. In the subsequent sections, the term “set a hit count breakpoint” is used instead of “skip the breakpoint n times”. It should be trivial to interpret the tricks in terms of skipping a breakpoint.
WinDbg
I did not find a way to set hit count breakpoints in windbg yet. Here is how you set a hit count breakpoint in WinDbg.
Go to the source view and set a breakpoint in the source code. The shortcut F9 can be used to toggle a breakpoint.
In the command window (alt + 1), list all breakpoints using the bl command.
Take note of the breakpoint that you just set and copy the location of the breakpoint which is listed in the format of <module_name>!<function_name>+<offset>. See example below.
Now redefine the breakpoint with the bp command. After the bp command paste the location that you copied in the previous step followed by the hit count.
0:000> bl
1 e x86 00000000`004113b2 0001 (0001) 0:**** test_project!wmain+0x42
0:000> bp test_project!wmain+0x42 2300
breakpoint 1 redefined
0:000> bl
0 e x86 00000000`004113b2 2300 (2300) 0:**** test_project!wmain+0x42
The hit count in the above example is set to 2300. This current hit count as shown above is decremented each time the breakpoint is hit but the execution stops only when this number is equal to 1. The number within the parentheses denotes the hit count that was originally set by the user.
Using Hit Count For Fun And Profit
Many developers set breakpoints without the hit count conditions. There are lot of nifty ways in which a hit count breakpoint can be used.
Below are some scenarios which developers will find useful while using hit count breakpoints:
Break In A Loop More Conveniently.
Setting an unconditional breakpoint in a loop (e.g. for, while, do-while) may break execution more often than needed. If you know the iteration of the loop when you want to suspend execution of the program, you can set a hit count breakpoint.
For example, in the while loop below if the intention is to break in the 21stiteration, a hit count based breakpoint will be more useful and simpler than a conditional one. Do note that in the loop below, the variable i does not increment by one.
<code>int i = 0;
while( !flag && i < N )
{
/* some code */
i *= 2;
}</code>
Likewise, the for-loop below traverses through the int vector using an iterator. If the intent is to break when the 10thelement in the vector is being processed in the loop, then a hit count breakpoint will be more useful and easy to set.
<code>std::vector<int>::iterator iter;
for( iter = vec.begin() ; iter != vec.end() ; ++iter )
{
/* some code */
}</code>
Create A Quick And Dirty Profiler And Much More.
Part I
Profilers that instrument code log the time taken by a function and the number of times it is called. It is the latter where hit count breakpoints are very useful. The greatest advantage of being able to track the number of times a function is called is that you don’t have a to run the code through a profiler but you get the results with the same accuracy. Moreover profilers may crash at times but debuggers are pretty stable when it comes to debugging code.
The trick here is to set a hit count breakpoint that will never be reached. For example, set a hit count to an unpractically large value (say 1000000) and set one breakpoint at the program termination (for example at the end of the main() function).
When the program is run, due to the large hit count, the breakpoint will never be hit and only the breakpoint at the end of the program will be hit. The debugger however has no knowledge that the breakpoint hit count is too large for it to be hit and therefore tracks the count whenever execution reaches the breakpoint.
At program termination, when the program gets suspended due to your second breakpoint, you have the debugger waiting to tell you what the hit count of the first breakpoint currently is. In other words it just told you how many times did the line of code get hit before the program terminated. That exactly is the kind of information that the profiler would have told you. Voila – you have that quick and dirty profiler ready for use :-).
Maybe someday I will write about how how a breakpoint works internally and then you can relate the similarity between what do the debugger and code instrumenting profiler have in common.
The above trick is explained in the C code snippet below.
<code>
void profile_me()
{
/* set hit count breakpoint here with a very large hit count */
/* function code */
}</code>
<code>
int main()
{
profile_me();
/* Set the second breakpoint here and when this is hit,*/
/* observe the hit count of the breakpoint set above */
return 0;
}</code>
Part II – Smart Breakpoints
Another use of hit breakpoints is very similar to the quick and dirty profiler trick. At times when one encounters a crash in a loop or in a repeated function call, it may make more sense to debug a few iterations prior to when the crash actually happens. For example, say a loop is processing tokens and a crash happens while processing the 2520th token. The crash itself may not make much sense once it has occured but it may help to know what happened 5 iterations prior to the crash. That way, the programmer can collect data for prior iterations and reach the crash condition. This will equip the programmer with relevant data needed to solve the crash at hand.
<code>while( token = get_token() )
{
/* some code */
switch( token )
{
case token_1: /*do code */
case token_2: /*do code */
/* more case statements */
}
}</code>
The trick here again is to set a very large hit count so that the breakpoint is never hit. Once the crash occurs, the hit count of the breakpoint is noted. Then the hit count of the breakpoint is reset to 5 minus the hit count obtained when the crash had occured. From now on whenver the hit count condition is met and the breakpoint is hit, the programmer will know that in 5 iterations a crash is expected. The data collected for the 5 iterations may be essential for resolving the crash.
Part III – Matching calls.
Hit count breakpoints have yet another use in debugging – matching the call count for a pair of functions. For example, for every malloc call a free call should have been made in order to have zero memory leaks. Similarly, a constructor (for now assume there is only one) and a destructor of a class should be called equal number of times. These calls have an opposite effect but their pair should match to ensure that resources don’t leak.
The trick is to set two hit count breakpoints with very large values that will never be reached in both the constructor and destructor above. Also a breakpoint should be set at the point of program termination (for example at the end of function main() ). The two breakpoints in the constructor and destructor will not be reached due to the very large values. When the program’s execution is suspended at program termination due to the final breakpoint, the hit count of the two breakpoints set in the constructor and destructor should be checked and hopefully their hit counts should match. Here I am assuming the class C was not involved in creating global or static objects. A mismatch of hit counts may suggest that not all objects of class C were destroyed and a possible resource leak should be looked into.
In summary, if there are two calls that should be called equally during the life span of a program, then this trick can be used to check that the call hit counts do indeed match.
Final Note
Hit count is a slightly under used feature of a debugger but it can be used in many innovative ways to gain better control over debugging. It is not a replacement for profilers but a great tool when you do not have one with you at hand. The infinite-hit-count-breakpoints are useful to keep track of code workflow as these breakpoints are set with the intention of never wanting them to be hit. However the information that such breakpoints can provide can be pretty useful and accurate.
