ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
0

catkin build ONLY the message target of a package

asked 2019-07-30 07:43:02 -0500

banerjs gravatar image

updated 2019-07-30 08:00:05 -0500

I have packages as follows:

package_A/
|-- msg/
    |-- CustomMsg.msg
    \-- ... (other messages)
|-- include/ ... (some code)
|-- src/ ... (some Python and C++ code)

package_B/ ... (also depends on package_A/CustomMsg.msg)

my_msgs/
|-- msg/ ... (some messages; one depends on package_A/CustomMsg.msg)
\-- ... (srv and action definitions that depend on msg)

my_package_1/ ... (depends on my_msgs)

my_package_2/ ... (depends on my_msgs)

I'm working in Melodic and I have some bag data, pkl files, etc., but unfortunately, I need to work with that data in another distro where it's infeasible for me to setup my workspace again (package_A and package_B won't build because of various dependencies, for example). The way I see it, I have 2 options:

Option 1:

I only need the messages built to work with my data, but I cannot figure out a way to get catkin to only build the message targets and ignore the code that comes in package_A. Is there a way to do this?

CMake doesn't come naturally to me, so when trying to build catkin build package_A_generate_messages failed, I decided to come here for help. If there is a way to build just CustomMsg.msg, even better!

Option 2:

I know that you can work across distros if the message hashes remain unchanged. But I don't know how those hashes are generated. Is the package name taken into account? I think the answer is yes, but wanted to check what else might be. CustomMsg.msg actually has a pretty simple message definition, so if I redefine the message in my own (faux) package called package_A on the other machine and build, will I end up with the same hash, which will then allow me to reuse my data files?

edit retag flag offensive close merge delete

Comments

I don't know if there is a way to force catkin to only generate messages, but you could comment out all the add_executable/library/dependencies and target_link_libraries in your packages CMakeLists.txt.

pcoenen gravatar image pcoenen  ( 2019-07-30 09:09:37 -0500 )edit

Yeah, that's essentially what I'm doing with option 2. I think it seems to be working; I'll post back here when I'm sure. Thanks!

banerjs gravatar image banerjs  ( 2019-07-30 12:00:53 -0500 )edit

3 Answers

Sort by ยป oldest newest most voted
1

answered 2019-08-01 13:23:55 -0500

gvdhoorn gravatar image

What you ask for is impossible, as the message generation targets are only created upon a successful CMake configuration phase run.

Success of that phase depends on the find_package(..) and other statements in the CMakeLists.txt of all pkgs in a workspace to succeed. That can only happen if all dependencies are present (which would appear to be exactly your problem: "package_A and package_B won't build because of various dependencies, for example").

The targets are defined in the Makefile that CMake generates after all this succeeds. At that point it would be trivial to invoke only the message generation target:

catkin_make package_A_generate_messages

or:

catkin build --make-args package_A_generate_messages

I'm not aware of any option to CMake (which Catkin will invoke) to make it ignore failing commands, so editing the CMakeLists.txt of package_A and package_B would seem to be your only option.

edit flag offensive delete link more

Comments

Got it. That makes a lot of sense. Thanks!

Just so I have this clear, assume for instance that instead of missing dependencies, which is indeed my problem, the problem was that dependency APIs had changed (for instance the arguments to a function had changed; something only caught when an actual make command is run)---then the option of running some cmake command followed by a make <package>_generate_messages would have been doable, right?

banerjs gravatar image banerjs  ( 2019-08-02 07:53:26 -0500 )edit

If CMake is happy with whatever it found for packages, and it's just compilation that would fail, then yes, I believe you should be able to run Catkin first (which is essentially CMake) and then the message generation targets should be there.

gvdhoorn gravatar image gvdhoorn  ( 2019-08-02 08:08:16 -0500 )edit
1

answered 2019-07-30 12:55:42 -0500

jayess gravatar image

As far as I'm aware, there is no way to build only parts of a package such as messages. There is, however, a third option: Create a new package that contains only the messages. This actually a very common practice and would help avoid this issue.

edit flag offensive delete link more

Comments

Yes, I know. But package_A and package_B are not maintained by me and while I do have access to the code, I do not want to incur the cost of a) collecting data again or of b) refactoring all the packages that depend on CustomMsg.msg since in reality there are many more than 3 or 4 packages in my workspace that depend on the message. So option 3, while valid, is a no-go.

banerjs gravatar image banerjs  ( 2019-07-30 13:21:57 -0500 )edit
0

answered 2019-07-31 07:16:00 -0500

banerjs gravatar image

updated 2019-08-02 08:17:14 -0500

I solved my problem with Option 2. It's essentially the principle behind this wonderful package: https://github.com/uts-magic-lab/mess.... But if there is a set of CMake commands that one can easily use to build just the messages, I'd love to know. (I know it's possible because my CLion IDE gives me the option of building just the generate_messages target; I just don't know how to get the actual command that CLion is using under-the-hood).

Edit: As mentioned in @gvdhoorn's answer, it appears that CLion gives me the option to only build the generate_messages target because it has already had a successful CMake build in the past. On the new machine, the extra dependencies within the CMakeLists.txt would cause the CMake build to fail, and therefore the build target of generate_messages would not exist. Conclusion - option 1 is impossible.

edit flag offensive delete link more

Question Tools

2 followers

Stats

Asked: 2019-07-30 07:43:02 -0500

Seen: 2,756 times

Last updated: Aug 02 '19