Skipping dependency strange behavior

[eliapellegrino]

After cleaning a forest workspace

tree@tree-comexpress1:~/circulo9_ws2$ rm -rf build/*
tree@tree-comexpress1:~/circulo9_ws2$ rm -rf install/*

and try to grow a recipe with some dependency, a strange behavior happened:

tree@tree-comexpress1:~/circulo9_ws2$ forest grow all -j8 -m xeno
building all with 8 parallel jobs
[all] depends on xbot2 -> not found, installing..
..[xbot2] depends on XBotInterface -> found
..[xbot2] depends on matlogger2 -> found
..[xbot2] depends on gazebo -> found
..[xbot2] depends on matio -> not found, installing..
....[matio] libmatio-dev already installed
....[matio] no build action required
....[matio] ok
..[xbot2] depends on xbot_msgs -> found
..[xbot2] depends on orocos_kdl -> found
..[xbot2] depends on eigen_conversions -> found
..[xbot2] depends on urdf -> found
..[xbot2] depends on tf2_kdl -> found
..[xbot2] depends on tf -> found
..[xbot2] depends on tf2_eigen -> found
..[xbot2] source code  already exists, skipping clone
..[xbot2] running cmake...
..[xbot2] building...
..[xbot2] ok
[all] depends on circulo9_device -> build found, building..
..[circulo9_device] skipping dependency ec_xbot2_client   ** THIS DEPENDECY SHOULD NOT BE SKIPPED **
..[circulo9_device] depends on xbot2 -> build found, building..
....[xbot2] skipping dependency XBotInterface
....[xbot2] skipping dependency matlogger2
....[xbot2] skipping dependency gazebo
....[xbot2] skipping dependency matio
....[xbot2] skipping dependency xbot_msgs
....[xbot2] skipping dependency orocos_kdl
....[xbot2] skipping dependency eigen_conversions
....[xbot2] skipping dependency urdf
....[xbot2] skipping dependency tf2_kdl
....[xbot2] skipping dependency tf
....[xbot2] skipping dependency tf2_eigen
....[xbot2] source code  already exists, skipping clone
....[xbot2] already built, skipping
....[xbot2] ok

As you can see, the dependency ec_xbot2_client should not be skipped.
Trying again to do the exact same thing, bring to a different result, where the dependency is not skipped. It seems a random behavior

tree@tree-comexpress1:~/circulo9_ws2$ rm -rf build/*
tree@tree-comexpress1:~/circulo9_ws2$ rm -rf install/*
tree@tree-comexpress1:~/circulo9_ws2$ ls build/
tree@tree-comexpress1:~/circulo9_ws2$ ls install/
tree@tree-comexpress1:~/circulo9_ws2$ forest grow all -j8 -m xeno
building all with 8 parallel jobs
[all] depends on xbot2 -> not found, installing..
..[xbot2] depends on XBotInterface -> found
..[xbot2] depends on matlogger2 -> found
..[xbot2] depends on gazebo -> found
..[xbot2] depends on matio -> not found, installing..
....[matio] libmatio-dev already installed
....[matio] no build action required
....[matio] ok
..[xbot2] depends on xbot_msgs -> found
..[xbot2] depends on orocos_kdl -> found
..[xbot2] depends on eigen_conversions -> found
..[xbot2] depends on urdf -> found
..[xbot2] depends on tf2_kdl -> found
..[xbot2] depends on tf -> found
..[xbot2] depends on tf2_eigen -> found
..[xbot2] source code  already exists, skipping clone
..[xbot2] running cmake...
..[xbot2] building...


..[xbot2] ok
[all] depends on circulo9_device -> not found, installing..
..[circulo9_device] depends on ec_xbot2_client -> not found, installing..
....[ec_xbot2_client] depends on xbot2 -> build found, building..
......[xbot2] skipping dependency XBotInterface
......[xbot2] skipping dependency matlogger2
......[xbot2] skipping dependency gazebo
......[xbot2] skipping dependency matio
......[xbot2] skipping dependency xbot_msgs
......[xbot2] skipping dependency orocos_kdl
......[xbot2] skipping dependency eigen_conversions
......[xbot2] skipping dependency urdf
......[xbot2] skipping dependency tf2_kdl
......[xbot2] skipping dependency tf
......[xbot2] skipping dependency tf2_eigen
......[xbot2] source code  already exists, skipping clone
......[xbot2] already built, skipping
......[xbot2] ok

[MarcoRuzzon]

I find somehow confusing that the circulo9_device build folder is found, even though the folder should be empty.
Since Forest does not build dependencies of projects already built by it, ec_xbot2_client build is skipped.

tree@tree-comexpress1:~/circulo9_ws2$ rm -rf build/*   <--- SHOULD EMPTY THE BUILD FOLDER
tree@tree-comexpress1:~/circulo9_ws2$ rm -rf install/*

tree@tree-comexpress1:~/circulo9_ws2$ forest grow all -j8 -m xeno
building all with 8 parallel jobs
[all] depends on xbot2 -> not found, installing..
..[xbot2] depends on XBotInterface -> found
..[xbot2] depends on matlogger2 -> found
..[xbot2] depends on gazebo -> found
..[xbot2] depends on matio -> not found, installing..
....[matio] libmatio-dev already installed
....[matio] no build action required
....[matio] ok
..[xbot2] depends on xbot_msgs -> found
..[xbot2] depends on orocos_kdl -> found
..[xbot2] depends on eigen_conversions -> found
..[xbot2] depends on urdf -> found
..[xbot2] depends on tf2_kdl -> found
..[xbot2] depends on tf -> found
..[xbot2] depends on tf2_eigen -> found
..[xbot2] source code  already exists, skipping clone
..[xbot2] running cmake...
..[xbot2] building...
..[xbot2] ok
[all] depends on circulo9_device -> build found, building..   <-- FOUND BUILD FOLDER OF circulo9_device
..[circulo9_device] skipping dependency ec_xbot2_client   <-- THIS DEPENDENCY SHOULD BE SKIPPED 
..[circulo9_device] depends on xbot2 -> build found, building..
....[xbot2] skipping dependency XBotInterface
....[xbot2] skipping dependency matlogger2
....[xbot2] skipping dependency gazebo
....[xbot2] skipping dependency matio
....[xbot2] skipping dependency xbot_msgs
....[xbot2] skipping dependency orocos_kdl
....[xbot2] skipping dependency eigen_conversions
....[xbot2] skipping dependency urdf
....[xbot2] skipping dependency tf2_kdl
....[xbot2] skipping dependency tf
....[xbot2] skipping dependency tf2_eigen
....[xbot2] source code  already exists, skipping clone
....[xbot2] already built, skipping
....[xbot2] ok

The expected behaviour is the second one.

[all] depends on circulo9_device -> not found, installing..
..[circulo9_device] depends on ec_xbot2_client -> not found, installing..

The build folder existence check is done only in src/forest/common/install.py and it's pretty straightforward. I don't see how this could give false positives.

In case this should happen again report it with re result of la command in the build folder.