On Sat, Mar 27, 2010 at 11:30 AM, Jeremy Leibs <leibs@willowgarage.com> wrote:
> Someone that knows more about this can probably help, however I've
> seen something like this before.
>
> I believe that image_proc does not actually do it's subscription /
> processing anything until something subscribes to one of the topics
> that it's producing.
>
> This is mentioned in the image_proc documentation:
>
> "All processing is on demand. Color processing is performed only if
> there is a subscriber to a color topic. Rectification is performed
> only if there is a subscriber to a rectified topic. While there are no
> subscribers to output topics, image_proc unsubscribes from the
> image_raw and camera_info topics."
>
> This is "nice" in that it doesn't consume unnecessary CPU processing
> images that nobody cares about (in turn since it's not subscribed to
> the camera, the camera doesn't actually have to be streaming data
> either if the driver were smart).  But, it's also very confusing in
> that it appears that it's not working until you actually subscribe.
>
> Do you have a downstream node subscribed to one of the processed image topics?

I did not, and that was the cause of the "failure" to subscribe.

Running image_view on the image_color output does indeed cause
image_proc to subscribe to image_raw. Unfortunately, there are still
no messages published on image_color, so that display window is blank.

But, that seems to get me closer to reproducing Eric's original problem.

Thanks, Jeremy!

I can still view the original image_raw topic, so the driver does seem to work.

Further suggestions are (always) welcome...
-- 
 joq