[ros-users] [PCL-users] best practice to install PCL in ROS

[Radu B. Rusu]

Dejan,

On 07/08/2011 04:57 PM, Dejan Pangercic wrote:
> Hi there,
>
> could anyone please tell us what is the best (recommended) way to
> install PCL under ROS?

There is only one way (and there has always been only one way): use whatever PCL package ROS provides, until we indicate 
otherwise.

Please RTFM in the future before posting. http://www.pointclouds.org/downloads/ clearly states that for PCL integration 
with ROS you should visit http://pcl.ros.org.

> We normally follow the instructions given on
> http://www.ros.org/wiki/perception_pcl which are made to check out the
> trunk version of PCL
> (http://svn.pointclouds.org/ros/trunk/perception_pcl/pcl/) but that
> trunk does not seem to have e.g. visualization library and some other
> components such as e.g. "pcl/console/*".

ROS is always going to be at least 6 months behind the development of PCL, due to a policy which does not allow the 
introduction of new features into a distribution. The files that you are searching for are either part of other packages 
(pcl_visualization in the first case), or are not present at all.

Bottom line is: if you use ROS, you need to be patient and wait until those features become available in ROS, until we 
change the above mentioned policy.

> In addition we've also noticed lately Radu suggesting installing PCL
> standalone when users reported having problems with PCL in ROS - but
> that can then not be used in ROS right?

You misunderstood. The pcl-users@ mailing lists deal with PCL problems in general. When someone sends an e-mail there 
saying that something doesn't work, the first thing that we do is try to isolate whether the problem is with PCL, ROS, 
or other 3rd party libraries that they are using.

The reason why people complain about having problems with PCL in ROS is because they try to use unstable/untested new 
features (pulling unstable stacks or using trunk), or they try to mingle PCL standalone with PCL in ROS. The standard 
answer is to try the same piece of code either with PCL standalone so we can see whether we have a bug in PCL, or with 
the perception_pcl stack released in a _STABLE_ ROS distribution (i.e. Diamondback at the moment), to see if there is a 
bug in users' code.


Cheers,
Radu.