[noetic] Rosrun uses heuristic to pick python devel-space relay script

[sloretz]

This is to support ros/catkin#1044 .That PR adds a devel-space relay script for installed python executables so their shebang can be rewritten to match ROS_PYTHON_VERSION. It has a drawback that rosrun will warn about non-unique executables because it sees one in the devel space and one in src space. This PR adds an exception to that warning.

The first commit is a slight refactoring that should not change rosrun's behavior.

1. 6773140 makes the script use pkg_name and file_name instead of $1 and $2 because it is confusing to read when adding a function that also uses arguments.
2. 61d017c makes rosrun pick the first executable if there are exactly two and one came from the --libexec list and another came from the --share list.

[dirk-thomas]

06a2805 splits the catkin_find invokaction into two: once with --libexec and once with --share.

This will slow down the command since Python needs to be invoked another time. Since this will affect every invocation of rosrun I am not sure that is ok. The alternative would be to leave the single invocation as is and only if it returned 2 matches query catkin_find once again (with only one of the two arguments). That would only increase the runtime when there are actually two matches which should be much less common.

[sloretz]

This will slow down the command since Python needs to be invoked another time. Since this will affect every invocation of rosrun I am not sure that is ok.

It does seem to slow it down; I'll implement an alternative. The average run time of 6 runs on my machine (bailing before exec) is 0.498s before 06a2805, and 0.781s after.

[sloretz]

The alternative would be to leave the single invocation as is and only if it returned 2 matches query catkin_find once again (with only one of the two arguments). That would only increase the runtime when there are actually two matches which should be much less common.

Implemented in 35c874f 61d017c

[sloretz]

@dirk-thomas mind having another look at this PR? I assume this will need a noetic-devel branch since it's only purpose is to support ros/catkin#1044 which has been merged already on catkin's noetic-devel branch.

[dirk-thomas]

LGTM

Will do the branching and merging when we are about to release this repo.

[sloretz]

@ros-pull-request-builder retest this please

[dirk-thomas]

@ros-pull-request-builder retest this please

[dirk-thomas]

Moved the first commit into a separate PR: #239.

[sloretz]

Looks like Noetic PR job passed 🎉

[sloretz]

Since this is a companion to ros/catkin#1044 which was merged onto the noetic-devel branch, I think this PR might need to be re-targeted. @dirk-thomas mind creating a noetic-devel branch?

[dirk-thomas]

I think this PR might need to be re-targeted.

Yes, as the [noetic] prefix in the title indicates.

mind creating a noetic-devel branch?

Not yet, before forking I need to land code cleanup changes.

[dirk-thomas]

before forking I need to land code cleanup changes.

Done in #240 and #241.