New compiler warning

[dirk-thomas]

As mentioned in ros-visualization/qt_gui_core#114.

E.g. http://build.ros.org/job/Mpr__qt_gui_core__ubuntu_bionic_amd64/1/warnings22Result/

[mikaelarguedas]

It's unclear to me why this shows up on this package. This function should be called transitively by createInstance.

Call stack being:

catkinFindLib
called by getCatkinLibraryPaths (l323)
called by getAllLibraryPathsToTry ('330)
called by getClassLibraryPath (l412)
called by loadLibraryForClass (608)
called by:
  -> createClassInstance (l129)
  -> createInstance (l158)
  -> createUniqueInstance (l190)
  -> createUnmanagedInstance (l229)

I built the melodic version of class_loader, pluginlib, qt_gui_core and rqt_gui with this additional print

Debug output of rosrun rqt_gui rqt_gui:

[DEBUG] [1522718065.187724347]: XML file specifies lookup name (i.e. magic name) = rqt_gui_cpp/RosCppPluginProvider.
[DEBUG] [1522718065.187735353]: Exiting determineAvailableClasses()...
[DEBUG] [1522718065.187742513]: Finished constructring ClassLoader, base = qt_gui_cpp::PluginProvider, address = 0x118eb80
[DEBUG] [1522718065.187829905]: Attempting to create managed instance for class rqt_gui_cpp/RosCppPluginProvider.
[DEBUG] [1522718065.187843431]: Class rqt_gui_cpp/RosCppPluginProvider maps to library lib/librqt_gui_cpp in classes_available_.
[DEBUG] [1522718065.187850493]: this function is actually called

@dirk-thomas can you please do the same experiment on a bionic system and see if that print appears?

[dirk-thomas]

Same here - I don't see any differences.

I would suggest to just move the code from the anonymous catkinFindLib function into the only function calling it, namely getCatkinLibraryPaths. That gets rid of the warning.

(I don't know why the anonymous function was created in the first place in #49 though.)

[dirk-thomas]

@mikaelarguedas How about my proposal from the previous comment?

[mikaelarguedas]

Yes I had the same idea, originally 👍 , as the function is private I think that's fine to not have it in its' own private namespace. I wanted to have a simple reproducible example to submit to GCC before removing it because that should be reported upstream.