Miscellaneous scrapping fixes

[gergondet-woven]

Hello drake-ros maintainers,

Thank you for making this project available. We have been using it to build our software for the past months and we have built up a number of patches that we hope can be useful to everyone.

This PR is the first I'm opening and it contains 4 small patches related to various edge cases in the workspace scrapping that we have encountered.

Below is a short summary of the individual patches, I am happy to submit separate PRs if you prefer.

P1: Remove elements that have a condition attribute on ROS1

This is used by a number of packages that are trying to support both versions at the same time.

This patch is actually part of #336 we actually had our own patch that does pretty much the same thing (including packages that have the $ROS_VERSION == 2 condition instead of excluding the $ROS_VERSION == 1). However, that PR incorporates more changes this particular fix is not available on main yet so I hope it's ok to bring it in with this PR.

P2: Fix an issue with some Python packages that have multiple entries in the egg top_level.txt (seen in `launch_testing for example)

P3: Fix an issue with CMake's JSON where libraries' paths are escaped under some conditions (for example if the workspace path has a ~ character)

P4: Include *.inc files in C++ headers of library, afaik, only the domain_bridge package uses this extension

Thanks again for making bazel_ros2_rules and I'm looking forward to your feedback.

---

This change is 

[zachfang]

@IanTheEngineer I assigned you as the reviewer for now just to ensure it's on our radar. Feel free to delegate otherwise.

[IanTheEngineer]

Thank you for the PR! I kicked off the CI for completeness, but the failures do not necessarily indicate that something is amiss with your PR. I will run some tests locally overnight and report back.

Reviewable status: 0 of 5 files reviewed, 1 unresolved discussion

---

bazel_ros2_rules/ros2/resources/cmake_tools/file_api.py line 72 at r1 (raw file):

                    self.link_flags.append(fragment['fragment'])
                elif fragment['role'] == 'libraries':
                    # In the generated JSON a library can be in quotes (to escape some characters?)

Nit: We try to wrap lines at 80 characters whenever possible. This applies to the other files in the PR as well.

[gergondet-woven]

Hello @IanTheEngineer

Thanks for the feedback.

We try to wrap lines at 80 characters whenever possible. This applies to the other files in the PR as well.

Sorry about that, I have pushed an extra commit to wrap all lines here under 80 columns.

[IanTheEngineer]

I took another pass looking through the changes this evening. This looks quite good. I have a couple small questions and nits below. Thank you for iterating on this.

Reviewed 5 of 5 files at r2, all commit messages.
Reviewable status: all files reviewed, 4 unresolved discussions (waiting on @gergondet-woven)

---

bazel_ros2_rules/ros2/resources/ros2bzl/scraping/ament_python.py line 17 at r2 (raw file):

    top_level = dist.read_text('top_level.txt')
    top_level = top_level.rstrip('\n')
    return str(dist._path), "\n".join(

Nit: Consider storing this string in a variable and returning that for readability. Also, consider adding a comment as to what this is fixing.

---

bazel_ros2_rules/ros2/resources/ros2bzl/templates.py line 156 at r2 (raw file):

        sandbox(top_level_i)
        for _, top_level in properties.python_packages
        for top_level_i in top_level.split("\n")

Nit: This nested list comprehension is quite clever, but somewhat hard to understand at a glance. Consider breaking it up for readability.

---

bazel_ros2_rules/ros2/resources/ros2bzl/scraping/metadata.py line 15 at r2 (raw file):

                    elements_to_remove.append((parent, child))
                else:
                    child.attrib.pop("condition")

Nit: Consider conforming to the single quote string convention used by the rest of the file.

---

bazel_ros2_rules/ros2/resources/ros2bzl/scraping/metadata.py line 15 at r2 (raw file):

                    elements_to_remove.append((parent, child))
                else:
                    child.attrib.pop("condition")

Looking at the ElementTree iter documentation, it states that the behavior is undefined when iterating over a tree if it is then modified.

Would child.attrib.pop('condition') be considered a modification to the tree structure?

---

bazel_ros2_rules/ros2/resources/cmake_tools/file_api.py line 76 at r2 (raw file):

                    # This escaping creates problem down the line,
                    # "/usr/lib/libfoo.so" is *not* a path.
                    # We remove the escaping here so they can be treated normally

This comment block is fantastic for conveying the issue and fix. Thanks for this!