A Python core developer has requested some changes be made to your pull request before we can consider merging it. If you could please address their requests along with any other requests in other reviews from core developers that would be appreciated.

Once you have made the requested changes, please leave a comment on this pull request containing the phrase I have made the requested changes; please review again. I will then notify any core developers who have left a review that you're ready for them to take another look at this pull request.

@1st1

I actually think that Antoine solution is more preferable

I prefer Antoine's solution as well as far as elegance and API design goes, my only concern is that it would be fairly disruptive with regards to backwards compatibility (from raising an error upon passing reuse_address=True) to apply all the way back to 3.5+. I'll wait for @asvetlov's feedback though.

Edit: If the consensus is to use Antoine's proposal instead, I can modify this PR accordingly. As long as it's within the next couple of days, I should have the time to do so. If I don't have enough time, I'll close this PR and let someone else implement it.

Due to Windows compatibility concerns, I no longer think that changing SO_REUSEADDR -> SO_REUSEPORT is going to be a viable option. As a result, I will start working on implementing Antoine's proposal. For more details, see https://bugs.python.org/msg357143.

@1st1 I finished changing the PR based on @pitrou's proposal, due to the Windows compatibility issues with the previous implementation.

I have made the requested changes; please review again.

Thanks for making the requested changes!

@1st1: please review the changes made to this pull request.

I'll add a Misc/NEWS entry.

The latest round of test failures is being caused by explicitly passing a default value for the reuse_address parameter (_not_set/_unset from @1st1's suggestion) specifying a socket through the sock parameter:

ValueError: socket modifier keyword arguments can not be used when sock is specified. (reuse_address=<object object at 0x7f9335b97be0>)

This occurs from:

        if sock is not None:
            ...
            if (local_addr or remote_addr or
                    family or proto or flags or
                    reuse_address or reuse_port or allow_broadcast):
                        ...
                        raise ValueError(
                            f'socket modifier keyword arguments can not be used '
                            f'when sock is specified. ({problems})')

My suggested solution is to remove or reuse_address from the above conditional, which I will apply in the next commit. This should be fine, since reuse_address will no longer have any functionality after the changes made in this PR.

Edit: Fixed.

Based on @ned-deily's suggestion to manually set the versionchanged to separate values for the 3.7, 3.6, and 3.5 branches, I'll remove the needs backport to 3.7 and needs backport to 3.6 labels so that @miss-islington doesn't open the backport PRs automatically.

If this PR is accepted and merged, I'll use cherry-picker to manually open the backport PRs for 3.7, 3.6, and 3.5.

Regarding the "What's New" entries, I would prefer to do those separately to simplify the merge conflicts. I'll have definitely time to take care of this on Monday, but for today and tomorrow I'll likely be busy with finishing up my final exams (last week of the Fall semester).

I should have time to make very small changes or reply to feedback, but I will likely not have time to do anything substantial until Monday (Dec 9th), unless I'm able to finish my work early on Sunday.

Edit: If any larger changes are needed before Monday, feel free to open a PR to my branch at https://github.com/aeros/cpython/tree/bpo37228-create_datagram_endpoint-reuse_address and I should be able to merge it. I've also ticked allow edits from maintainers, which allows minor changes to the contents of the PR through the GitHub UI.

@ambv: Please replace # with GH- in the commit message next time. Thanks!

Thanks @aeros for the PR, and @ambv for merging it 🌮🎉.. I'm working now to backport this PR to: 3.8.
🐍🍒⛏🤖

GH-17529 is a backport of this pull request to the 3.8 branch.

Note: merged for 3.8 and 3.9. @aeros removed backport labels to 3.6 and 3.7 here. The issue on BPO still lists those versions as affected. Please fix either way.

@ambv

@aeros removed backport labels to 3.6 and 3.7 here. The issue on BPO still lists those versions as affected. Please fix either way.

This is because they need to be manually backported, due to a difference in the versionchanged note. I'll take care of this process and attach the backport PRs to the bpo issue.