Switch branches/tags
Nothing to show
Find file
Fetching contributors…
Cannot retrieve contributors at this time
400 lines (248 sloc) 9.25 KB


docs/debugger.pod - The Parrot Debugger


This document describes parrot_debugger, the Parrot Debugger.


Starting from version 0.0.6 Parrot has its own debugger, which is modeled after Perl's one. Its name is parrot_debugger, and is an interactive environment that let you step through bytecode, set breakpoints, evaluate assembly instructions and peek at the interpreter status.

A good (well, at least some) knowledge of the Parrot internals is obviously required to read this document. Some familiarity with debugging principles is also mandatory beyond this point.

BUILDING parrot_debugger

The debugger is built along with Parrot when you run 'make', but if you want to build *only* the debugger, then you can run:

make parrot_debugger

Which will create a new parrot_debugger executable in the same directory as the parrot executable.


To start the debugger type:

parrot_debugger file.pbc

That is, parrot_debugger takes exactly one argument, which is the Parrot file that you're going to debug. This file may be Parrot bytecode (*.pbc), PASM source code (*.pasm) or PIR (*.pir). parrot_debugger will automatically load and disassemble the bytecode file for you.

Note that you can't pass command line arguments to your program when you invoke the debugger. See the run (r) command below for this.

After the version banner, you'll see the friendly debugger prompt:


parrot_debugger is ready to receive commands and give output. To list the available commands type 'h'. To quit the debugger type 'q'.

As with the Perl debugger, whenever it halts and shows you a line of code, it is always the line it's about to execute, not the one that it has just executed.


Always remember that you can enter 'h' to get a list of commands (this document may be outdated in respect to the actual debugger, so let it speak for itself).

Most commands can be shortened to their first letter. When available, this is signaled by the letter in parentheses after the command name Thus, help (h) means that the command can be given as 'help' or just 'h'. On the other hand, load can only be given as 'load', verbatim. Debugger commands are case sensitive (LOAD is not a valid command) but Parrot register names are not.

A blank line always repeats the last command entered.

Also note that at this point in its development, parrot_debugger has very poor error checking on commands and their arguments, so type carefully or something bad will happen. Feel free to report bugs, or better yet patch the source code (see "FILES" below).


Assign to a Parrot register. For example:

(pdb)    a I0 42

    I0 = 42

(pdb)    a N1 3.14

    N1 = 3.14

The first command sets I0 to 42 and the second sets N1 to 3.14. Register names are not case sensitive, so you can use either i0 or I0 .


Disassemble a loaded bytecode file. This will turn a file loaded with load into proper Parrot assembler.


Load a source code (assembler) file. The syntax is:

load FILE
list (l)

List the source code. The syntax is:

list [FROM] [NUM]

Both arguments are optional. By default FROM is from where the last list command ended (or the first line if this is the first invocation) and NUM is 10. That is, it lists the source code ten lines at a time.

Note that the disassembled source code is not the same as the original source code: labels take the names L1 .. Ln and opcodes are fully qualified (eg. set_i_ic instead of just set). See also eval (e).


# lists the first three source code lines
(pdb) l 1 3
1  set_i_ic I1,0
2  L3:  print_sc "fact of "
3  print_i I1
run (r)

Run (or restart) the program. The syntax is:


Any arguments you give are passed as command line arguments to the program (ie. they populate P0).

After the program has ended, you can run it again with this command. See also the continue (c) command.


(pdb) r
fact of 0 is: 1
fact of 1 is: 1
fact of 2 is: 2
fact of 3 is: 6
fact of 4 is: 24
fact of 5 is: 120
fact of 6 is: 720
Program exited.
break (b)

Add a breakpoint. The syntax is:


If you want a conditional breakpoint you should first specify the register that is involved in the condition (at least one must be), the comparison and then the third argument can be either another register or a constant, which must be of the same type as the first register specified.

The command returns a number which is the breakpoint identifier. You should note this number for the delete (d) command (see below).


# sets a breakpoint on line 10 (will be breakpoint 0)
(pdb) b 10
Breakpoint 0 at line 10

# another breakpoint on line 11 (will be breakpoint 1)
(pdb) b 11
Breakpoint 1 at line 11

# break at line 4 if I16 is less than or equal to 123456
(pdb) b 4 if I16 <= 123456
Breakpoint 2 at line 4

# break at line 4 if N27 is greater than 5.23
(pdb) b 5 if N27 > 5.23
Breakpoint 3 at line 5

# break at line 4 if S2 is equal to S13
(pdb) b 6 if S2 == S13
Breakpoint 4 at line 6

# break at line 4 if S5 is equal to "stop"
(pdb) b 7 if S2 == "stop"
Breakpoint 5 at line 7
watch (w)

Add a watchpoint. The syntax is:


The condition has the same format as in break

delete (d)

Delete a breakpoint. The syntax is:


The NUM argument is the breakpoint number (from 0 to N) as emitted by the break (b) command. It is NOT the line that has the breakpoint.


# delete the first breakpoint (was on line 10, see example above)
(pdb) d 0

Disable a breakpoint. The syntax is the same as for the delete command. Disabled breakpoints can be re-enabled with enable.


Re-enable a disabled breakpoint. The syntax is:

enable [NUM]

where NUM is the number of the breakpoint.

continue (c)

Continue the program execution. The syntax of this command is:

continue [NUM]

Without arguments, the command just runs the source code until a breakpoint is found (or until the end of the program).

If you specify a number, it will skip the next NUM breakpoints it encounters.

When the program has ended, continue will do nothing. Use run (r) to execute it again.

next (n)

Run the next instruction. The syntax is:

next [NUM]

NUM defaults to 1, but you can give a number of instructions to execute before stopping again.

eval (e)

The eval command is currently unimplemeneted.

Run an instruction. The syntax is:



(pdb) e set I0, 42

(pdb) e print I0

(pdb) p i
I0 =           42
I1 =            0
trace (t)

Trace the next instruction. The syntax is:

trace [NUM]

It executes the next NUM instructions (default is 1) just as next (n) does, but printing additional trace information. This is the same as the information you get when running Parrot with the -t option.


# executes 2 instructions and trace them
(pdb) t 2
PC=0; OP=67 (set_i_ic); ARGS=(I1=0, 0)
PC=3; OP=24 (print_sc); ARGS=("fact of ")
fact of
3  print_i I1
print (p)

Print the interpreter registers. Register names are not case sensitive. The syntax is:

print VALUE

VALUE may be:

A register name: I3

Prints out the single register specified.

A register type: i, n, s, or p

Prints out all registers of the given type

An aggregate key: P0[1]

Looks up the given (integer- or string-valued) key in a PMC register.

For PMC registers, the command will print the number, the class of the PMC (in square brackets) and its string representation (when available). It prints <PMCNULL> for uninitialized PMC registers.


# prints the content of I2
(pdb) p i2
I2 =              0

# prints the content of P0
(pdb) p P0
P0 = [ResizablePMCArray]

# prints the content of all string registers
(pdb) p s
 S0 = Just
 S1 = Another
 S2 = Parrot
 S3 = Hacker

Print interpreter information.


(pdb) info
Total memory allocated = 81936
GC mark runs = 6
GC collect runs = 0
Active PMCs = 8197
Active buffers = 7
Total PMCs = 21840
Total buffers = 48
Header allocations since last collect = 0
Memory allocations since last collect = 2
quit (q)

Exit the debugger.

help (h)

Prints information about debugger commands. The syntax is:

help [COMMAND]

If COMMAND is omitted, prints a list of the available commands.



This is the file that will produce the executable. Nothing fancy here, it mostly consists of the main function.


Most of the debugger is implemented here. You may want to start from the PDB_run_command function and go down from there for the real meat.


Parrot_debug, the function which launches the debugger, is implemented here.


This defines all the PDB structures, which hold data used by the debugger.


Version 1.0

First version (SVN debug.c revision 1.24), authored by Aldo Calpini