fix(NAPPCORE-2747): Broadcast only primary interface events.

[oleksii-skidan]

Problem

We have noticed that the socket protector on Apple appears to rebind sockets too frequently. For example, when a cellular connection is enabled and the system transitions from Wi-Fi, it will rebind sockets first due to the change in the Cellular interface, then due to the change in the Wi-Fi interface. What I noticed was that when this was happening, the Cellular interface was not the primary one yet.

We have a suspicion that this may be causing hard-to-track no-nets. I made a build for Rolandas Pakėnas with this fix (as he was the one who reported the most frequent occurrences of no-nets), which he daily drove for a week. His feedback was that the no-nets stopped occurring for him.

Solution

I added a bit of code (not included in this MR) that was dumping the changed DynamicStore keys. That's how I know that sometimes we try rebinding sockets due to changes in non-primary interfaces. From the numerous dumps, the keys that we are interested in the most seem to be these:

• State:/Network/Interface/<interface>/IPConfigurationBusy
• State:/Network/Global/IPv4
• State:/Network/Interface/<interface>/IPv4
• State:/Network/Interface/<interface-UUID>/IPv4

Theoretically, when we get IPConfigurationBusy, we must check if the corresponding value is true and rebind only if it is. But at the moment, it seems to be too much of a change for a single MR. And without this pedantic change, the results are quite promising.

The change is: I added a filtering step to dynamic_store_callback in telio-sockets/src/protector/apple.rs. It checks if the change key is one of the above. And it checks if the interface is primary.

☑️ Definition of Done checklist

• Commit history is clean (requirements)
• README.md is updated
• Functionality is covered by unit or integration tests

[nord-llt]

All committers have signed the CLA.

[tomasz-grz]

Some minor notes, otherwise +1

[tomasz-grz]

Nice! Please clean up the commit history before merging.

Also if you add a comment in the .unreleased/NAPPCORE-2747_broadcast_only_primary_interface_events file, regarding the changes, it should appear in the release changelog, so that the Apple Team is inform.... oh wait.. 😅

[oleksii-skidan]

Hey @tomasz-grz, what do you mean by "clean up commit history"? Do you suggest squashing it all?

And regarding the file in the .unreleased directory. I noticed that they are all empty. Is there a template or structure to the contents of the file? I mean, should I have a one-liner describing what it is or write an essay? 😆

[tomasz-grz]

@oleksii-skidan Yes squash them, or you can "Squash and Merge" (Just don't tell anybody I said that) 😀

Not sure why check-dod is not happy...
Can you try adding this to the bottom of your PR description?

### :ballot_box_with_check: Definition of Done checklist
- [x] Commit history is clean ([requirements](../blob/main/docs/git_commit_messages_requirements.md))
- [x] README.md is updated
- [x] Functionality is covered by unit or integration tests

[tomasz-grz]

@oleksii-skidan regarding the .unreleased, usually they are empty when the changes don't affect the apps/platforms so much.

If you add a simple oneliner like this the ticket/PR will be listed in the changelog.

[oleksii-skidan]

Hey @tomasz-grz, sorry for the delay :) I squashed it, updated the PR description, and added a one-liner to the changelog.

[oleksii-skidan]

Also, our QA tested this change and it works as expected 🎉

[oleksii-skidan]

Hey @tomasz-grz and @LukasPukenis, could you please take another look at this PR, approve it, and merge? 🥺