tf chaining broken on pr2 about geometry2

[ahendrix]

Done for Gazebo launch. Though I would hope that was unnecessary, hoping it is easy for you to reproduce.

Anyway, it seems the cause is not too mysterious: in tf.cpp:Transformer::chainAsVector(const std::string & target_frame, ros::Time target_time, const std::string & source_frame, ros::Time source_time, const std::string& fixed_frame, std::vector<std::string>& output) const

None of the input parameters other than output are even being used, this is makred as //@todo. The code that worked in diamondback is still in tf.cpp, but commented out.

For my other problems, it seems I cannot reproduce, maybe there was something wrong with my waitForTransform at the time.

[ahendrix]

@ahendrix: Obsolete with the use of tf2 as the backend for tf. Closing.

@tkruse: I beg to differ. Not only is tf still there and the tf_monitor script part of it, but also tf2/src/buffer_core.cpp:BufferCore::_chainAsVector is a copy&paste implementation (sigh) of tf/src/tf.cpp:Transformer::chainAsVector.

I would prefer you to reopen the issue, and then close it once that either the tf_monitor script is not part of a ROS distro anymore, or until the tf_monitor script delivered with a distro does not have that problem anymore (e.g. by relying on a fixed tf2 implementation of chainAsVector), please.

[tfoote]

This method was stubbed out when the datastructure was converted and currently lists the parent frames not the spanning set. It needs to be fixed in both implementations. Fortunately this problem is only in the chainAsVector methods which are used for console output. But it needs to be fixed in both tf and tf2.

[jacquelinekay]

Proposed solution in pull request 95. @tkruse and @ahendrix : feel free to review.