Sedb

The SmartEiffel debugger is an Embedded debugger. This means that, in order to use it, you need to compile your application with a special flag -sedb. The executable will then have the debugger inside, you can launch your application the usual way.

Note: sedb is only available with compile_to_c (or compile).

The debugger offers powerful commands which allows you to examine the bowels of your program:

• commands to control execution,
• commands to set or unset breakpoints,
• commands to print your data,
• other various commands.

Execution control

These commands allow your program to execute. You can either run step by step or make bigger strides, or even force execution until the end.

Whatever your choice is, the program will run until the next breakpoint (see below), or until the next ^C.

The available commands are (by increasing order of step size):

s	Move only one step forward. This means entering a routine if it is a feature call.
n	Move only one step forward (next), but without entering a routine (this means that a feature call is considered being one step).
f	Move to the next routine return (finish).
c	Move as far as possible (continue), so maybe until the end of the program.
C	Continue until the end of the program, ignoring all breakpoints (static and dynamic ones).

Breakpoints

There are two kinds of breakpoints:

• those explicitly placed in the code, with the instruction sedb_breakpoint,
• those dynamically placed in the debugger. To modify dynamic breakpoints, you can use the following commands:

b	Add a dynamic breakpoint. There are many possible criteria, described below. All chosen criteria must be valid to activate the breakpoint.
B	Print all dynamical breakpoints.
-<num>	Remove a dynamic breakpoint designed by its number. The numbers are printed by the B command.

The sedb_breakpoint instruction

The sedb_breakpoint instruction is defined in ANY. It places a static breakpoint directly in the Eiffel source. This command does nothing if the program is not compiled with the -sedb flag.

These breakpoints cannot be removed by the -<num> command. The only way to ignore them is to use the C command (continue until the end of the program).

Specifications of a dynamic breakpoint

To set up a dynamic breakpoint, you are able to specify one or more criteria.

• Name: the name of the method and the class name in which the method is defined, separated by a space, for example "item STRING". You can specify a substring of this name. Thus, if you specify item then the program will stop at the beginning of the item method of STRING, but also at the beginning of the one in ARRAY, etc. Similarly, if you specify STRING then the program will stop when encountering any method of STRING, but also of HASHED_DICTIONARY[STRING, INTEGER]. You can even be original: I let you guess the behaviour of the is_ specification :-)
• File: for example, string.e. As the file name is applied to the complete path, you can specify lib/kernel which will stop at all the methods in the classes of this cluster.
• Line numbers: you can specify an interval of lines, for example [12,13].
• Execution stack: this condition allows to track the size of the execution stack (useful for debugging a recursive function for example). For example, you can specify the limit of the stack size to 10, which will stop the execution when the stack size reaches 10. An optional automatic increment permits the limit to be incremented each time the breakpoint is reached. If you don't use this option, this breakpoint can be good way of track the memory consumption of the stack.

Of course, all these specifications are cumulative: in this case, they must all be true at the same time to activate the breakpoint.

Data printing

The following commands allow to print the data of your program:

e <exp>	Evaluates and prints the result of an expression. Note that the general Eiffel expressions are not supported.
	Only the following can be printed: Current local variables a parameter of the current routine the content of the case of a NATIVE_ARRAY and suffixing by a dot one of its index beginning at zero (for example storage.2 prints the third element of storage) the attributes of these objects, by using the classical notation with dots (for example my_string.count or my_string.storage.4), recursively the result of once functions at the condition that they have already been evaluated (i.e. the debugger does not procede to the function call; it only checks the result)
p p <exp>	Re-evalutes the same expression by suffixing a dot and the expression (if present). This is very useful to chain printing. You can also use the particular expression .. which goes one level up, .... which goes two levels up, etc.
	Example of the use of e and p : sedb> e Current STD_OUTPUT#0x807ab00 [ filter = Void buffer_position = 0 buffer = NATIVE_ARRAY[CHARACTER]#0x806c030 capacity = 4096 ] sedb> e buffer NATIVE_ARRAY[CHARACTER]#0x806c030 sedb> p.0 (sedb) e buffer.0 'H' sedb> p..1 (sedb) e buffer.1 'e' sedb> p....buffer_position (sedb) e buffer_position 0 sedb>
.	Print the current frame; that is the content of local variables of the current routine.
u	Goes up in the stack (i.e. goes to the caller). This means that the current routine becomes the calling routine. Note that the e, p and . commands follow the current routine.
d	Goes down in the stack (opposite of u).
S	Prints the execution stack. There are two printing modes: compact mode which only prints the name of the file and executed routine; an asterisk indicates the current routine complete mode which prints the entire stack, with all frames (as the printed stack when a program crashes outside sedb).

Various commands

q	Quits the debugger; the programm will be stopped. You can also use Q which doesn't ask for confirmation.
h ?	Prints help.
H	Prints detailed help.
G	Run the garbage collector, if it's present.
T	Changes the trace mode, if it's present (flag -trace when compiling). If the mode is active, the file trace.se fills up (very quickly!)
Return	Re-execute the last command.