Fix IP _resolve_hostname for Unix

[TheBlueOompaLoompa]

Fix the implementation of _resolve_hostname for the unix ip driver.
Can resolve .local domains now using mDNS.
Fixes #101453

[Faless]

Is there any reference for the requirement of hints.ai_socktype = SOCK_STREAM; specific to mDNS?
The manual just claims:

This field specifies the preferred socket type, for example SOCK_STREAM or SOCK_DGRAM. Specifying 0 in this field indicates that socket addresses of any type can be returned by getaddrinfo().

And domain resolution is also used in functions that open UDP sockets. It's not clear to me why this would fix the issue when 0 is not working.

[TheBlueOompaLoompa]

I specified hints.ai_socktype = SOCK_STREAM; initially because it was in an example. I built a test c++ project outside godot, and used the example from the manual.

You are correct that 0 works, and it might make sense to switch it to that. I tested it with just setting to 0 and it works just as well. I'm not sure what advantage any specific option would provide in this situation, or if it even matters at all.

I'm guessing the issue has to do with hints.ai_socktype technically being uninitialized, even though memset should have already set it to 0. I went in with a debugger to investigate, but hints.ai_socktypewas still 0 after getaddrinfo was called. At this point figuring it out would require stepping inside the getaddrinfo function, but that really doesn't seem like an option.

[Faless]

I'm guessing the issue has to do with hints.ai_socktype technically being uninitialized, even though memset should have already set it to 0. I went in with a debugger to investigate, but hints.ai_socktypewas still 0 after getaddrinfo was called.

Well, this doesn't make much sense, if the value is properly zeroed by memset, setting it to 0 via the member won't change what's processed by getaddrinfo, so it must produce the same result.

In fact, I just tested resolving a .local domain with 4.4.1 stable and it's working fine for me (Debian 12 Bookworm)

[Faless]

As per the discussion above, the SOCK_STREAM hint is incorrect, and it's not clear why explicitly setting it to 0 works for the OP since we already memset the whole struct to 0.

I've also not been able to reproduce the original issue in my network.

Further investigation and testing is needed.