synth_device watchdog {

    ## How should the watchdog device decide that the watchdog has
    ## triggered, assuming the eCos application has failed to
    ## reset the device in time. There are two options: wallclock
    ## time, where the watchdog will trigger after a second
    ## irrespective of what the eCos application has been doing;
    ## or consumed cpu time, where the watchdog will only trigger
    ## if the eCos application has actually run for a whole second.
    ##
    ## Using wallclock time matches most real hardware,
    ## but can give spurious results if there are other processes
    ## running and consuming cpu cycles. Also, if the user has
    ## halted the application in gdb then the watchdog will trigger
    ## even though the application has no chance to reset the
    ## watchdog.
    ##
    ## Using consumed cpu cycles avoids these problems.
    ## However the default implementation of the idle thread action
    ## is to block in a select() system call, so if the eCos
    ## application spends most of its time idling but not resetting
    ## the watchdog then the device will not trigger. The
    ## configuration option CYGIMP_HAL_IDLE_THREAD_SPIN can be used
    ## to change the synthetic target HAL's implementation of the
    ## idle thread action.
    ##
    ## The default is to use consumed_cpu_time. Wallclock time can
    ## be selected by uncommenting the next line.
    # use wallclock_time
    # use consumed_cpu_time
    
    ## When running in GUI mode there will be a small window showing
    ## the current state of the watchdog: asleep, watching, or giving
    ## the alarm. This window can be packed as desired, using one
    ## of the container frames .main.nw .main.n .main.ne .main.w
    ## .main.e .main.sw .main.s or .main.se, and with a position of
    ## left, right, top or bottom. The default is to pack inside the
    ## .main.n frame, on the right, thus causing the watchdog to
    ## appear above the central text window.
    # watchdog_pack -in .main.n -side right

    ## If running in GUI mode, when the watchdog triggers and resets
    ## the eCos application this will be shown in the display.
    ## Optionally the watchdog device can also give an audible alert
    ## when this happens. This is disabled by default to avoid
    ## annoying co-workers, but can be enabled by specifying a suitable
    ## sound file. Two such files, sound1.au and sound2.au, are supplied
    ## with the watchdog device itself. Alternatively the user can
    ## specify a different sound file using a suitable path
    # sound sound1.au
    # sound sound2.au
    # sound ~/sounds/my_watchdog_alert.au

    ## If sound is enabled then the watchdog device needs some way to
    ## access the host's sound hardware. Rather than accessing
    ## /dev/audio or /dev/dsp directly the watchdog.tcl script will
    ## run an external program. This can avoid complications with
    ## different sound file formats etc. The default program is
    ## "play", a front-end to the sox utility, but a different
    ## program or script can be specified if desired. It will be run
    ## simply as "<program> <sound file> &"
    # sound_player play
}