Rework zeroconf for unicast scanning

[bdraco]

I've tested this on both my large production setups without issues with many airplay, home pods, and apple tv devices. I think it will solve home-assistant/core#80215

It would be great if someone with an older apple tv could test this.

needs python-zeroconf/python-zeroconf#1192

needs home-assistant/core#95714

fixes home-assistant/core#80215

[codecov]

Codecov Report

Patch coverage: 100.00% and project coverage change: -0.12% ⚠️

Comparison is base (dfcc82b) 88.61% compared to head (d5ea8d8) 88.50%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2083      +/-   ##
==========================================
- Coverage   88.61%   88.50%   -0.12%     
==========================================
  Files         163      163              
  Lines       10991    11001      +10     
==========================================
- Hits         9740     9736       -4     
- Misses       1251     1265      +14     

Files Changed	Coverage Δ
pyatv/core/scan.py	97.83% <100.00%> (+0.55%)	⬆️

... and 3 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[bdraco]

We should try load_from_cache first like we do in aiohomekit

[bdraco]

the behavior should be the same exact we don't create hundreds of tasks now if they are just going to return right away because its in the cache

[bdraco]

and if its not in the cache we do create the tasks

[bdraco]

I'm pretty sure this will solve home-assistant/core#80215 and it doesn't regress any of my production set ups. It would be nice to get a test from someone else though

[bdraco]

Since this path is used for HA we could do a release, and bump HA at the top of the month to get a long feedback period

[postlund]

Just let me know when you are satisfied here and want this merged. I can probably make a release if needed.

[bdraco]

I think the code is fine.

I'm internally debating if a test that patches to the code to make sure we do the correct zeroconf calls makes sense or if there is a better way to do it. I think patching is probably the only realistic option since otherwise we end up with an unreliable test that needs specific network conditions

[bdraco]

even with checking the cache first, there are still too many requests (this is an existing problem)

python-zeroconf/python-zeroconf#1205

[bdraco]

fixed that issue. will look at the tests a bit more later

[postlund]

I suggest patching for tests as well. That's easier and more reliable.

[bdraco]

Happy with the performance now. Will bump the zeroconf version and add the tests

[bdraco]

@postlund should be good to go now.

[postlund]

@bdraco Great! Heading to bed now, but I'll look it over tomorrow and make a new release for you.

[bdraco]

perfect. thanks!