Back to library index.

Package std-debug (in std.i) - debug commands

Index of documented functions or symbols:

catch

DOCUMENT catch(category)
  Catch errors of the specified category.  Category may be -1 to
  catch all errors, or a bitwise or of the following bits:

     0x01 math errors (SIGFPE, math library)
     0x02 I/O errors
     0x04 keyboard interrupts (e.g.- control C interrupt)
     0x08 other compiled errors (YError)
     0x10 interpreted errors (error)

  Use catch by placing it in a function before the section of code
  in which you are trying to catch errors.  When catch is called,
  it always returns 0, but it records the virtual machine program
  counter where it was called, and longjumps there if an error is
  detected.  The most recent matching call to catch will catch the
  error.  Returning from the function in which catch was called
  pops that call off the list of catches the interpreter checks.

  To use catch, place the call near the top of a function:

     if (catch(category)) {
       ......
     }
     ......

  If an error with the specified category occurs in the "protected"
  code, the program jumps back to the point of the catch and acts
  as if the catch function had returned 1 (remember that when catch
  is actually called it always returns 0).

  In order to lessen the chances of infinite loops, the catch is
  popped off the active list if it is actually used, so that a
  second error will *not* be caught.  Often, this is only desirable
  for the error handling code itself -- if you want to re-execute
  the "protected" code, do this, and take care of the possibility
  of infinite loops in your interpreted code:

     while (catch(category)) {
       ......
     }
     ......

  After an error has been caught, the associated error message
  (what would have been printed had it not been caught) is left
  in the variable catch_message.

  ***WARNING***
  If the code protected by the catch contains include or require
  calls, or function references which force autoloads, and the
  fault occurs while yorick is interpreting an included file,
  catch will itself fault, and the error code will not execute.
  If a fault occurs after an include has pushed a file onto
  the include stack for delayed parsing and you catch that fault,
  the include stack will not unwind to its condition at the time
  catch was called.  That is, catch is incapable of protecting
  you completely during operations involving nested levels of
  include files.

  In some cases, after_error is a more appropriate way to recover
  from errors.

SEE ALSO: error, after_error

dbauto

SEE: dbexit

dbcont

SEE: dbexit

dbdis

SEE: dbexit

dbexit

DOCUMENT Debug mode.

Yorick errors fall into two general categories: Syntax errors discovered
during parsing, and runtime errors discovered when a Yorick program is
actually running.  When a runtime error occurs, Yorick offers the
choice of entering "debug mode", which you can do by typing the 
key immediately after the error occurs.  Typing a non-blank line exits
debug mode automatically by default.  In debug mode, the Yorick prompt
becomes "dbug>" instead of the usual ">".  When you see this prompt,
Yorick has halted "in the middle of" the function in which the error
occurred, and you can print, plot, modify, or save the local variables
in that function by means of ordinary Yorick commands.  Debug mode is
recursive; that is, you can debug an error which occurred during
debugging to any number of levels.

You can exit from debug mode in several ways:

   dbexit            -- exit current debug level, discarding all
                        active functions and their local variables
   dbexit, 0         -- exit all debug levels
   dbexit, n         -- exit (at most) N debug levels

   dbcont            -- continue execution of the current function
      Continuing is useful if you have managed to repair the
      problem which caused the error.  The expression in which the
      error occurred will be evaluated a second time, so beware of
      side effects.

   dbret, value      -- continue execution by returning VALUE (which
                        may be nil or omitted) to the caller of the
                        function in which the error occurred.
      This is useful if the function in which the error occurred is
      hopelessly confounded, but you know the value it should return.

Yorick does not allow "single stepping" directly, although you can
execute the statements in a function by copying them, then tell
Yorick to skip those statements you have executed "by hand".  There
are two functions for skipping execution:

   dbskip            -- skip the next logical line (This will be only
                        a portion of a source line if several statements
                        are stacked on the source line.)
   dbskip, n         -- skip next N (positive or negative) logical lines

   dbup              -- discard the current function, so that you are
                        debugging its caller -- there is no way to go
                        back "down", so be careful

There are two functions which print information (like other print
functions, if called as functions instead of subroutines, their
result is returned as a string array with one line per string):

   dbinfo            -- returns current function and source line

   dbdis             -- returns disassembled virtual machine code
                        for the next line (use the disassemble function
                        to get the entire function)
      This allows you to see exactly where in a line the error occurred.

Finally,

   dbauto            -- toggles whether debug mode will be entered
                        automatically when a runtime error occurs
   dbauto, 1         -- enter debug mode automatically after an error
   dbauto, 0         -- type  after error to enter debug mode

dbinfo

SEE: dbexit

dbret

SEE: dbexit

dbskip

SEE: dbexit

dbup

SEE: dbexit

disassemble

DOCUMENT disassemble(function)
      or disassemble, function
  Disassembles the specified function.  If called as a function, the
  result is returned as a vector of strings; if called as a subroutine,
  the disassembly is printed at the terminal.  If the function is nil,
  the current *main* program is disassembled -- you must include the
  call to disassemble in the main program, of course, NOT on its own
  line as a separate main program.

error

DOCUMENT exit, msg
         error, msg
  Exits the current interpreted *main* program, printing the MSG.
  (MSG can be omitted to print a default.)
  In the case of exit, the result is equivalent to an immediate
  return from every function in the current calling chain.
  In the case of error, the result is the same as if an error had
  occurred in a compiled routine.

SEE ALSO: print, write, batch, catch

exit

SEE: error