Fix map display for moving TF frame

[jacobperron]

Instead of the current time, use Time(0) to get the latest available transform as a fallback.
This is the same logic that is applied in RViz from ROS 1.

Resolves #332

Although we are using TF message_filters to deliver the map to the display, the display itself is doing a transformation:

rviz/rviz_default_plugins/src/rviz_default_plugins/displays/map/map_display.cpp

Lines 576 to 579 in 045227a

	if (!context_->getFrameManager()->transform(frame_, transform_time, current_map_.info.origin,
	position, orientation) &&
	!context_->getFrameManager()->transform(frame_, context_->getClock()->now(),
	current_map_.info.origin, position, orientation))

And even though it is using the current time (just a few lines before), it is off by less than a second causing the extrapolation exception. I noticed the ROS 1 implementation is slightly different, defaulting to the latest available transform by passing Time(0) (the related PR is ros-visualization/rviz#1066):

  if (!context_->getFrameManager()->transform(frame_, transform_time, current_map_.info.origin, position, orientation) &&
      !context_->getFrameManager()->transform(frame_, ros::Time(0), current_map_.info.origin, position, orientation))

https://github.com/ayrton04/rviz/blob/8f482f494ee4031ed04b3065a9755710b61176bf/src/rviz/default_plugin/map_display.cpp#L719-L720

[jacobperron]

• Linux
• Linux-aarch64
• macOS
• Windows

[jacobperron]

IT'S ALIVE!

[Martin-Idel]

I guess using Time(0) should be the same in all displays? If that is true, a quick search reveals that this is also wrong in the axes display (my bad) and all pointcloud displays (see pointcloud common). The background is that when we ported most of the displays, the time API wasn't finished and we mostly tested with static publishers.

Should this be done in a different ticket or do you prefer to do it here?

[jacobperron]

Should this be done in a different ticket or do you prefer to do it here?

I also noticed these other occurrences. I was going to investigate further to see if I can reproduce the same issue with the other displays before proceeding. I can create a ticket for (potentially) fixing the other displays.

[jacobperron]

I've opened up a follow-up #484

[wjwwood]

lgtm, thanks for looking into it!