Re: [ros-users] image_proc doesn't work with camera1394

On Sat, Mar 27, 2010 at 7:13 PM, Jeremy Leibs <leibs@willowgarage.com> wrote:
> On Sat, Mar 27, 2010 at 2:24 PM, Jack O'Quin <jack.oquin@gmail.com> wrote:
>> On Sat, Mar 27, 2010 at 2:57 PM, Eric Perko <wisesage5001@gmail.com> wrote:
>>
>>> Has anyone else been able to get any 1394 camera working with
>>> image_proc using either camera1394 or an older 1394 driver?
>>
>> Yes!
>>
>> I just noticed that camera1394 was publishing a zero time stamp.
>> (Didn't we used to get warnings about that?).
>>
>
> That explains the problem.  The time-synchronizer, which synchronizes
> the image and camera_info messages requires that the two messages have
> the same timestamp.

They actually were the same: both zero. I guess it needs the times to
actually progress, as well.

> We no longer get warnings because, I believe, the time in the header
> is no longer a magic/privileged field since auto-filling based on
> ros::Time::now() is seldom actually the right behavior.

Agreed.

> However, I am surprised that the time-synchronizer (inside of
> image_proc) is not producing warnings about not finding matching
> timestamps, this seems like something that would be very helpful in
> debugging this kind of problem.

It should not just silently throw everything away.

An appropriate warning would be good. This probably won't be the last
time someone porting a camera driver to ROS makes that mistake.

Thanks for your help, Jeremy.
--
joq