Re: Insight build problem on cygwin (tentative patch enclosed)

[ac131313@redhat.com (Andrew Cagney)]

> [reply-to set]
> On Mon, Feb 17, 2003 at 01:38:02PM -0500, Andrew Cagney wrote:
> >>-  static const struct interp_procs tk_procs =
> >>-  {
> >>-    tk_init,
> >>-    gdbtk_resume,
> >>-    gdbtk_suspend,
> >>-    gdbtk_exec,
> >>-    gdbtk_prompt_p,
> >>-    gdbtk_command_loop,
> >>-  };
> >>-
> >>-  interp_add (interp_new ("gdbtk", NULL, NULL, &tk_procs));
> >>-
> >
> >FYI,
> >
> >Removing from _initialize*() is wrong.  Interpreters should only be 
> >registered in _initialize*() function.
> >
> >I think the underlying problem is general confusion over the separation 
> >of powers between interp_add() and interp_init().
> >
> >BTW, even with that NULL, it worked for me.  Is there something cygwin 
> >centric going on?
> 
> Don't see how it could be.  interp_set does this:
> 
>   uiout = interp->interpreter_out;
> 
> So, at some point uiout becomes NULL.  I wiped out my old non-working version
> so I can't give the specific instance right now but it seems obvious that this
> could happen.  If you set interpreter_out to NULL and call interp_set at some
> point then uiout will be NULL.

I don't know.  Probably didn't use it in anger.

I think the underlying problem is with interp_add().  It shouldn't
even take that ui-out parameter.  Instead the ui-out should be set
using some other mechanism (interp_init()?).

Andrew

PS: Note that the current gdbtk_init() relies on the `going to be
deleted real soon now' init_ui_hook.