macOS broadcast improvements

[themarpe]

Purpose

Fixes IPv4LL edge case on macOS when running with another ETH interface alongside normal WiFi connection. Also fixed CI that was failing before. PR split into two commits - one related to broadcast fix and one to CI improvements

Dependencies & Potential Impact

macOS discovery has potential impact

Deployment Plan

Planned for RVC2&4 IPv4LL improvements

@jakaskerl

[moratom]

Thanks @themarpe ! Appreciate the CI/integration fixes to make it running.

Is my understanding correct here?

• the PR makes sure discovery actually sends a broadcast packet on all interfaces on MacOS
• the behavior before that was that the same interface (that MacOS somehow picks) was used multiple times for the discovery
• Linux works as expected without explicitly binding to an interface

LGTM if tested otherwise.

[themarpe]

Is my understanding correct here?

• the PR makes sure discovery actually sends a broadcast packet on all interfaces on MacOS
• the behavior before that was that the same interface (that MacOS somehow picks) was used multiple times for the discovery
• Linux works as expected without explicitly binding to an interface

1. Yes, the change actually makes sure that per iteration of the loop of the broadcast packet being sent to specific interface, is actually sent to that subnet - by - binding to interface first (macOS only diff)
2. Yes, UDP broadcast in certain occasions (WiFi with internet + IPv4LL ETH), on macOS wasn't routed to IPv4LL subnet, even though it targeted IPv4LL subnet (169.254.x.y/16). Because we iterate through interfaces, we can then also bind to an interface to make sure it goes out on that interface & its subnet. (has to do with how macOS sets default routing in this occasion)
3. Correct, Linux works as expected without binding to an interface

---

Yes, tested and approved by @jakaskerl as well on 2 mac hosts (existing & fresh)