The Perl Debugger

Many people use the Perl debugger with no more than these commands. Once you are comfortable with those, however, an additional four commands can make your debugging more efficient, especially for programs that use object-oriented code:

You can search and display code that matches a string or regular expression with / for forward searches and ? for backward searches. There should be no space before the string you're looking for:

  DB<6> /name
6:      my $name = "Pengu";

The S and m commands are useful for exploring what subroutines or methods are available: S lists every subroutine and method available to the program. These are in reverse order of when they were loaded by use or require, and they include routines loaded from the debugger, such as Term::ReadLine. The m command lists every method available to an object or by way of a package. Here is a sample:

  DB<7> use CGI

  DB<8> $q = new CGI

  DB<9> m $q
AUTOLOAD
DESTROY
XHTML_DTD
_compile
_make_tag_func
_reset_globals
_setup_symbols
add_parameter
all_parameters
[...]

Actions, breakpoints and watchpoints provide even more control over the debugger and the running program. You may prefer using them from a graphical Perl debugging front end, such as ddd, ptkdb or Activestate Komodo. The most common complaint about the Perl debugger is remembering the proper command-line shortcut for each command, and these commands add still more shortcuts to remember.

Additionally, in Perl 5.8 some of the keyboard commands have changed to make them more internally consistent. Often, though, people need to use both 5.8 and an earlier version, so it may be easier to use a GUI. I describe the commands from the command line below; the principles remain the same.

An action is used to wedge code into your program without modifying the source file. It can be useful when the code is in production and you want to test a change. It's also useful if you're in the middle of a debugging run and want to change code without restarting the debugging session from scratch.

You set an action like so, a <line-number> <code>. An example could be:

DB<10> a 9 $index = $_;

This adds a new command inside the foreach loop that stores the index count, which is incremented each time through. If you list the program, you see an a next to the line number that has the action. The action is executed before the line to which it is attached. You can list actions you've set with L and delete an action by specifying a with the line number without a command. The previous is for Perl 5.6 and earlier; in Perl 5.8, delete an action with A and the line number of the action.

Breakpoints and watchpoints return control to the debugger from continuous execution, such as from r and c described above. They are useful for jumping ahead to the particular iteration of a loop that is having problems, without repeatedly stepping through the loop by hand.

A breakpoint is set on a line number or subroutine, with an optional condition that must be met. A breakpoint is set with b as shown here:

b shout

If you list the program, you can see a b next to the line number at the first line of the subroutine shout. Press C to continue execution, and it stops inside the subroutine.

If you followed the previous example and set the action on line 9, you could set a breakpoint to stop on a particular iteration of this loop:

b shout ($index eq 8)

This should give you an idea of the power of actions and breakpoints, if you imagine debugging a longer program with complex conditional statements and external data sources.

You can list breakpoints with L and delete one with d in Perl 5.6 and earlier. In Perl 5.8, you delete a breakpoint with B.

A watchpoint probably is better known as a watch expression. It halts the program as soon as a specified expression changes. In Perl 5.6, it is set with W as shown here:

W $name

You can list watchpoints with L and delete all of them by specifying no parameter to W. In Perl 5.8, add a watchpoint with w and delete it with W.