foxy and galactic branch contains breaking changes since last release

[ijnek]

When I branched off foxyand galactic branches, I branched off from the ros2 branch, because that was the dev branch listed in rosdistro for the distributions. However, the ros2 branch contained a significant number of breaking changes, and the foxyand galactic branches shouldn't have been branched off devel.

Should try and revert all the breaking changes in the two branches to get the foxyand galactic branches working and make a release.

[ijnek]

I've gone through all PRs that were merged since the last release in Foxy and Galactic.

Changes that need to be reverted:

• Install includes to include/${PROJECT_NAME} and more modern CMake #419
• Use Boost::pythonXY target #421
• Use Boost::python3 if Boost < 1.67 #422

Changes that don't have to be reverted:

• delete __init__ for launch directory #352
• [ros2][cv_bridge] Updated cv_bridge.dll install location. #356
• Fix multiple undefined references #370
• Export sensor_msgs dependency #392
• fix endianness comparison #397
• Yuv422 yuy2 #396
• remove boost endian #399
• Rename opencv_tests install folder to deconflict #357
• Handle padded img msg python #400
• fix windows build for image_geometry #388
• Add equidistant distortion model #358
• Implement unrectifyImage() #359
• Fix rectifyRoi when used with binning and/or ROI #378
• substituted missing sphinx extension to fix Python docs #417
• prevent conversion of single channel 16bit integer images to/from colour #412
• Add header arg to cv2_to_imgmsg #326
• Forward-port PRs from noetic to ros2 #420
• Find Python3 if not on Android #423
• Minor cleanups to the ROS 2 branch #418
• [cv_bridge][ROS 2] Add missing dependency on rcpputils #427
• python3-opencv should be a more general dependency #439
• Update README.md #252
• test: modify python test #444
• update maintainer #451

[Ace314159]

Is there a reason to revert #421 and #422 as well? After the revert, builds do not work when using a Boost 1.67 or newer, as explained in #390 (comment)

from this document:

Note that Boost Python components require a Python version suffix (Boost 1.67 and later), e.g. python36 or python27 for the versions built against Python 3.6 and 2.7, respectively. This also applies to additional components using Python including mpi_python and numpy. Earlier Boost releases may use distribution-specific suffixes such as 2, 3 or 2.7. These may also be used as suffixes, but note that they are not portable.

[ijnek]

Hi @Ace314159, sorry for the inconvenience the reversion has caused and thanks for reporting it. Let's get it resolved.

Here's the original reasons:
#419 was reverted because that was the PR that was breaking API in foxy and galactic.
#421 was reverted because it depended a lot on changes from #419
#422 was reverted because it was fixing mistakes in #421.

However, now looking at #390, this is suggesting to me that the foxy/galactic build on windows might fail after the reversions I've made.

Could you describe the platform / distribution that's failing? And also, which branch you're building from source, (or whether you are using a binary).

[Ace314159]

Thanks for the prompt response! I am building from source using the foxy branch on Windows.

[ijnek]

Great, thanks for the information. Could you try out the branch: foxy-with-boost-python-fix I just put up?
It has changes from #390 applied, which was easier to port across than the changes from #421.

[Ace314159]

It works! Thanks for the quick turnaround!

[ijnek]

I'm going to re-open this issue until a release has been made. I'm currently trying to figure out how to get a pre-release test going before I make another release.

[ijnek]

Releases were made.

However, I've been reported that the backport to galactic broke Galactic builds on RHEL: https://build.ros2.org/job/Gbin_rhel_el864__cv_bridge__rhel_8_x86_64__binary/

I've reverted the release to an older version until we resolve this. Let's track this issue here.

[ijnek]

@Ace314159 could you review #495 please?

[Ace314159]

It looks good to me! Just wasn't sure about the COMPONENTS thing I mentioned.

[ijnek]

Closing, as should be fixed with #495 🤞