Do not fail parse_resolv_conf on invalid hostname

[schultetwin1]

On macOS, user's can set their hostname to something invalid such as
matt.schulte's macbookpro. This will cause parse_resolv_conf to fail
because "into_resolver_config" will think the hostname contains the
system domain and then fail to parse that system domain.

Instead, we should treat the hostname as if it contains no domain (the
same as if the hostname contained no ".").

This fixes #1739

[bluejekyll]

If I’m understanding the change here, we’re ignoring a malformed name. I guess I don’t know what we want in that case. Im a little surprised that Apple doesn’t escape those strings.

[Noah-Kennedy]

@bluejekyll windows doesn't either iirc

[schultetwin1]

If I’m understanding the change here, we’re ignoring a malformed name. I guess I don’t know what we want in that case. Im a little surprised that Apple doesn’t escape those strings.

That's correct but I look at it a little differently.

I would say we are treating malformed names the same way we are treating names without a "." in them. So if your hostname is "mattscomputer" or "matt's computer", or "matt.schulte's computer" this now treats them all the same way. Previously, "mattscomputer" or "matt's computer" would result in None for the domain and parse_resolv_conf would not fail. "matt.schulte's computer" would result in an parse_resolv_conf failing.

[bluejekyll]

Would you have time to add a test (with some documentation) to that effect? I’m ok with merging this, but I worry that we might regress this in the future as it’s not an obvious case we’re trying to support.

Thanks for the explanation and the PR!

[schultetwin1]

Would you have time to add a test (with some documentation) to that effect? I’m ok with merging this, but I worry that we might regress this in the future as it’s not an obvious case we’re trying to support.

Future regressions are a good point. I can add some documentation but unfortunately I'm not sure if there is a minimal way to add testing here. The get_system_domain eventually calls into a crate that will call a system call to get the hostname. I'm not sure how I could control the return of that syscall without modifying the actual machine's hostname (which I'd like to avoid if possible).

[Noah-Kennedy]

@djc any reason this isn't merged?