[Sbcl-devel] disable-debugger

[James Y K. <fo...@fu...>]

The check for *invoke-debugger-hook* being nil in disable-debugger is  
causing a minor problem for me, as I'd like to start my saved core  
with --disable-debugger, sometimes. But I have overridden invoke- 
debugger-hook as follows, to work around another issue I ran into:

> ;;; Patch up the debugger so that it doesn't run serve-events  
> callbacks.
> ;;; Running app code asynchronously while you're trying to debug it
> ;;; is *NOT* helpful.
> (defvar *old-descriptor-handlers* nil) ; for inspecting in debugger
> (defun invoke-debugger-without-event-callbacks (condition &optional  
> val)
>   (declare (ignore val))
>   (let ((*old-descriptor-handlers* sb-impl::*descriptor-handlers*)
>         (sb-impl::*descriptor-handlers*  ()))
>     (invoke-debugger condition)))
> (setf sb-impl::*invoke-debugger-hook* #'invoke-debugger-without- 
> event-callbacks)

The --disable-debugger command-line option thus doesn't do anything,  
since invoke-debugger-hook has already been set at core-load-time.

Of course, I can work around the disable-debugger problem easily enough:

> (sb-ext:without-package-locks
> (defun sb-debug::disable-debugger ()
>    ;; override sbcl's disable-debugger so that it unconditionally  
> disables the debugger.
>    (setf sb-impl::*invoke-debugger-hook* 'sb-debug::debugger- 
> disabled-hook)
>    (sb-alien:alien-funcall (sb-alien:extern-alien  
> "disable_lossage_handler" (function sb-alien:void)))))

But perhaps one or both of these issues can be fixed in sbcl rather  
than layering hack upon hack in my code. :)

James