[ros-users] odd bug? with rospack

Brian Gerkey gerkey at willowgarage.com
Mon May 17 16:09:47 UTC 2010


On Sun, May 16, 2010 at 12:32 AM, David Feil-Seifer
<david.feilseifer at gmail.com> wrote:
> I inadvertently created a package with the same name as an existing
> package. That caused me some strange errors with rosbuild when that
> happened. The package dependencies were not being read from the new
> package's manifest.xml, but the old one. That caused the package to
> build with a different set of package dependencies. Could this be
> remedied by adding a warning when make is run in a directory but
> rospack finds a manifest.xml file (for assessing dependencies) in
> another directory? This would at least alert the user to this odd
> behavior.

hi Dave,

Good idea.  I added a check for that situation, r9745 on ros/trunk.
It is now a fatal error if the directory in which you're trying to
build a package is not the same (according to Python os.path.realpath)
as where rospack says the package lives.

> Also, it might be a good warning for roswtf if there are
> multiple packages with the same name.

That's probably not a good idea, because having multiple packages by
the same name in a tree is a common use case for many developers, who
use overlays on top of standard installations.  ROS was designed
specifically to support this pattern.

	brian.



More information about the ros-users mailing list