Refactor URI overrides

[Shadowghost]

Refactor how URI overrides are loaded and created. It should now respect JELLYFIN_PublishedServerUrl again.
The override set via JELLYFIN_PublishedServerUrl supercedes any other override set via the UI or config file (and overrides them).

Changes

• Fix GetAllBindInterfaces -> fixes Kestrel binding on startup
• Do not exclude interfaces from unicast binding if the interface is not multicast capable
• Properly handle multicast binding for Linux, Windows and macOS
• Exlcude localhost manually from multicast on macOS due to MulticastInterface set to loopback index does not work on OSX dotnet/runtime#24340
• Refactor URI override handling
• Respect JELLYFIN_PublishedServerUrl
• Introduce new class PublishedServerUriOverride
• Subnet parser now properly creates the catchall subnets if 0.0.0.0 and ::0 are input without subnet mask
• Cleanup NetworkManager logging and only show network information on startup (on network change printing is now debug-only)

Issues

• Fixes [Issue]: Docker: JELLYFIN_PublishedServerUrl gets ignored #10005
• Fixes [Issue]: Jellyfin unstable not binding to localhost #10070

[satmandu]

Is there a good way to generate a docker image from this PR for testing?

I don't have a build env setup to build Jellyfin, and we need to test it with Docker anyways...

[Shadowghost]

I don't think so. But you can test it without docker too (I'll do it myself if I find the time...)

[PrplHaz4]

Is there a good way to generate a docker image from this PR for testing?

I don't have a build env setup to build Jellyfin, and we need to test it with Docker anyways...

These instructions served me well in the past to test without standing up the whole jf toolchain:
https://jellyfin.org/docs/general/contributing/development#building-and-testing-inside-a-docker-container

[satmandu]

This PR fixes #10005 for me.

[Bond-009]

Failing test on windows

[xUnit.net 00:00:00.57]     Jellyfin.Networking.Tests.NetworkExtensionsTests.TryParse_ValidHostStrings_True(address: "www.google.co.uk") [FAIL]
  Failed Jellyfin.Networking.Tests.NetworkExtensionsTests.TryParse_ValidHostStrings_True(address: "www.google.co.uk") [149 ms]
  Error Message:
   Assert.True() Failure
Expected: True
Actual:   False
  Stack Trace:
     at Jellyfin.Networking.Tests.NetworkExtensionsTests.TryParse_ValidHostStrings_True(String address) in D:\a\1\s\tests\Jellyfin.Networking.Tests\NetworkExtensionsTests.cs:line 30
   at InvokeStub_NetworkExtensionsTests.TryParse_ValidHostStrings_True(Object, Object, IntPtr*)
   at System.Reflection.MethodInvoker.Invoke(Object obj, IntPtr* args, BindingFlags invokeAttr)
Results File: D:\a\_temp\VssAdministrator_fv-az644-602_2023-08-15_01_15_02.trx

Failed!  - Failed:     1, Passed:   122, Skipped:     0, Total:   123, Duration: 1 s - Jellyfin.Networking.Tests.dll (net7.0)

[Shadowghost]

Failing tests are caused by a DNS resolution error when resolving www.google.co.uk. Not sure how to handle this, they are running fine locally.

[cvium]

Failing tests are caused by a DNS resolution error when resolving www.google.co.uk. Not sure how to handle this, they are running fine locally.

We shouldn't have non-integration test making external calls

[Shadowghost]

The difference between this PR and earlier is that we're now explicitly calling Dns.GetHostAddresses() with IPv4 or IPv6 parameter instead of using the generic overload. According to the .NET docs this seems to be unnecessary though, so I can revert it and the test will pass again.
I'd revert the change anyway but should I remove the test case too?

[satmandu]

The refactor is working fine for me.

[satmandu]

@Shadowghost Any luck on getting this merged?

[Shadowghost]

I still want to fix a bug concerning multicast/broadcast binding on Windows. But it is a bit annoying to debug and I don't have that much time right now...

[satmandu]

@Shadowghost Any chance of leaving that Windows issue to a later PR, since this already fixes two issues? (I'm currently building your branch rebased against master, but I would love to not have to do that.)

[satmandu]

Thanks for rebasing! I don't know what happened on my end, but the rebased version of course works well.

[Shadowghost]

@satmandu @cvium bugs should be fixed. I tried a lot of combinations but apparently Windows and Linux need different handling of broadcast and multicast if you don't want to bind on all interfaces...

[crobibero]

works for me on macOS