Windows fixes and iface indexes

[relativityspace-jsmith]

Hello! After what we talked about last week, I prototyped a way to add interface indices to the netifaces2 API. Along the way, I also found some bugs on Windows and am working through fixing them.

Changes made so far:

• Add a new netifaces.interfaces_by_index() API function. This is similar to netifaces.interfaces(), except that it returns a dictionary of interface index to interface name instead of just the interface name.
• Implement interfaces_by_index() for Windows and POSIX in the rust code. This is my first time ever writing any Rust, so please be gentle and let me know how I messed up :P
• Add regression tests for two issues I noticed on Windows
• Add "Developing Locally" section to the README. These are the commands I figured out to develop this project locally, no idea if they're the correct ones...
• I could NOT resist renaming some of the linux_xxx functions to posix_xxx ones, because I'm kinda triggered by the fact that they say linux but work for mac as well
• Bugfix: On Windows, IPv6 addresses were returned under the wrong key in the ifaddrs() dict
• Bugfix: On Windows, some interfaces would report being assigned 0.0.0.0 as an IP, when in actual fact they did not have an IP address at all

[SamuelYvon]

I could NOT resist renaming some of the linux_xxx functions to posix_xxx ones, because I'm kinda triggered by the fact that they say linux but work for mac as well

I like that thinking ;) Gotta make sure it's fully posix-compliant however. I'll setup a more complete test setup

[relativityspace-jsmith]

Thanks for the feedback! Any way you could approve my workflow run so I can see how badly I broke the CI?

[relativityspace-jsmith]

Also, re the Windows memory issue, I do not see the issue on my machine and will keep an eye out for any machines that do have it!

[relativityspace-jsmith]

Also, should mention, there is still one serious issue with netifaces-2 that would keep me from using it with multicast_expert: currently, loopback interfaces are not returned by netifaces.interfaces() on Windows (though they are on other platforms). I suspect this is related to using GetAdaptersInfo() in this library verseus netifaces using GetAdaptersAddresses(), because the original netifaces is able to find the loopback interface somehow.

Will have to look into this in a subsequent MR.

[SamuelYvon]

Also, re the Windows memory issue, I do not see the issue on my machine and will keep an eye out for any machines that do have it!

I was never able to reproducer either :/ the code looks right also, but apparently is not. See issue https://github.com/SamuelYvon/netifaces-2/issues/21

loopback interfaces are not returned Will have to look into this in a subsequent MR.

I would greatly appreciate! If you don't have the time I can also take a look this weekend but I'm not sure how much time I actually have to put towards this

[relativityspace-jsmith]

Workflow approve again? It looks like it needs approval each time X_X

[relativityspace-jsmith]

Also, side note, I'm having an issue that's really annoying. It seems like ruff is enforcing a line length of <=88 characters, but black is not configured to reformat down to that line length. So, running black doesn't make my line lengths correct and I have to enforce that manually. Any idea how to fix this? I tried setting the black line length in pyproject.toml but that didn't fix it.

P.S. I would be really happy if we could increase the line length to be a bit more generous, like 120 chars. 88 chars feels like I'm being squeezed into a tiny box...

[SamuelYvon]

Yep, I agree. I'll fix that ASAP

[SamuelYvon]

If you rebase on dev you'll have the changes.

[relativityspace-jsmith]

    hooks:
      - id: black
        args: [--config=./pyproject.toml]

ohhh, thaaaaat's what I was missing 🤦‍♂️

[SamuelYvon]

Workflow approve again? It looks like it needs approval each time X_X

Should be more lenient now

[relativityspace-jsmith]

OK I think this is ready for merge!

[SamuelYvon]

Thanks for the hard work!