Re: Proposed patch for threadsafe access to octave internal data

[John Swensen]

Michael Goffioul wrote:
> On 7/15/07, John Swensen <address@hidden> wrote:
> > I have attached my first shot at a patch to allow the IDE I am writing
> > threadsafe access to octave internal data.  The basic idea is that I
> > have created a class called octave_server that has a mutex protected set
> > of data and acts as an intermediary between my IDE and octave.  I have
> > registered a function with the octave_rl_event_hook so that when
> > readline is idle, a function is called that processes data sent from the
> > IDE to octave (e.g. requests for variables values, breakpoint
> > operations, etc) and then fills in buffers for data that is updated
> > regularly (e.g. the names and properties of variables that are currently
> > in scope, new lines of history information, etc.) for subsequent
> > asynchronous access by the IDE.  According to the readline
> > documentation, the event hook is called at most 10 times per second.
> > This is more than adequate for the needs of the IDE.
>
> What if some other code wants to install its own input event hook?
> (octplot or jhandles does it). Are they correctly chained?
>
> Michael.
No, it does not.  That will require a change to the gnu_readline class, 
so that multiple packages can register for the rl_event_hook and then it 
dispatches them (in the order they are registered).  At the current 
time, though I suppose it doesn't make much sense, octplot and jhandles 
could not coexist either.  That is, unless you are handling setting up 
more than one event hook in a different manner.

This isn't a very big change, so I will try to have another patch for 
this done by the end of today.

John Swensen