Base URL hostnames are overridden by certain paths when constructing a URL

[jonnydgreen]

• Version: v16.2.0
• Platform: Darwin bisterix.local 20.5.0 Darwin Kernel Version 20.5.0: Sat May 8 05:10:33 PDT 2021; root:xnu-7195.121.3~9/RELEASE_X86_64 x86_64
• Subsystem: url

What steps will reproduce the bug?

Running new URL('\\\\-', 'http://asd') in Node.js v16.2.0 shell will override the hostname of the arbitrarily chosen base URL.

How often does it reproduce? Is there a required condition?

Every time.

What is the expected behavior?

The URL is correctly constructed and the hostname is preserved:

URL {
  href: 'http://asd///-',
  origin: 'http://asd',
  protocol: 'http:',
  username: '',
  password: '',
  host: 'asd',
  hostname: 'asd',
  port: '',
  pathname: '///-',
  search: '',
  searchParams: URLSearchParams {},
  hash: ''
}

What do you see instead?

The hostname of the base URL is overridden by the path:

$ node
Welcome to Node.js v16.2.0.
Type ".help" for more information.
> new URL('\\\\-', 'http://asd')
URL {
  href: 'http://-/',
  origin: 'http://-',
  protocol: 'http:',
  username: '',
  password: '',
  host: '-',
  hostname: '-',
  port: '',
  pathname: '/',
  search: '',
  searchParams: URLSearchParams {},
  hash: ''
}

Additional information

This does not occur in Node.js 14.15.4 (I haven't tested in other versions). Happy to help resolve this, although I would need some pointers for where to start :)

[bl-ue]

It's the same behavior in the browser (Chrome 90) — it's probably actually correct functionality per the spec 🤔

[bl-ue]

I believe this is working as intended. According to the spec, the URL constructor take two parameters: url and optional base. If base is not undefined and url has no base (a base is like https://abc or \\host), then base gets applied to url to produce the final URL. To demonstrate:

// `base` (second param) is used here because `url` (first param) has no base (it's just a relative path)
new URL('/hello/world', 'http://def') // -> { href: 'http://def/hello/world' ... }

// `base` is ignored here because `url` already has a base
new URL('http://abc/hello/world', 'http://def') // -> { href: 'http://abc/hello/world' }

You might be getting confused because \\-—which is a normal Windows UNC path like \\hostname—is also interpreted as a base, thus your base parameter (http://asd) is getting ignored.

cc @aduh95

[jonnydgreen]

Ah that makes sense about the base interpretation, yeah - thanks for the clarification! :) If this is correct functionality per the spec for v16, would that mean it's incorrect for Node.js v14 in that case, since this does not get interpreted as a base?

[bl-ue]

Not sure what you mean about v14 — from v10.0.0 (when URL was introduced) to v16.3.0 (latest) the behavior is the same (I checked the latest of all major versions, 10, 11, 12, 13, 14, 15, 16 🙂).

[jonnydgreen]

Sorry, I should've provided an example! On v14, I ran the same command but got a different result, i.e. the path '\\\\-' was not interpreted as a base and http://asd was not ignored?

$ node
Welcome to Node.js v14.15.4.
Type ".help" for more information.
> new URL('\\\\-', 'http://asd')
URL {
  href: 'http://asd///-',
  origin: 'http://asd',
  protocol: 'http:',
  username: '',
  password: '',
  host: 'asd',
  hostname: 'asd', <-- Was expecting this to be '-' like in v16
  port: '',
  pathname: '///-',
  search: '',
  searchParams: URLSearchParams {},
  hash: ''
}

[bl-ue]

Ohhhh — I made a mistake. I didn't realize what you were saying :)

Yeah, something has definitely changed recently 😬 let me investigate.

[jonnydgreen]

Cool, thanks - happy to provide more testing/debugging output if you need :)

[bl-ue]

I'm guessing that the previous functionality (that which you just demontrated) was a bug, and fixed with v15.6.0 by @RaisinTen in #36613.

cc @jasnell @watilde

[jonnydgreen]

Thanks for the update, looking at that PR it looks like it's included in the v14.17.0 proposal? I've just run on this version and it works okay!

$ node
Welcome to Node.js v14.17.0.
Type ".help" for more information.
> new URL('\\\\-', 'http://asd')
URL {
  href: 'http://-/',
  origin: 'http://-',
  protocol: 'http:',
  username: '',
  password: '',
  host: '-',
  hostname: '-',
  port: '',
  pathname: '/',
  search: '',
  searchParams: URLSearchParams {},
  hash: ''
}

I guess we can close this now?

[bl-ue]

Ah yes, #36613 was backported to v14 with v14.17.0.

I guess we can close this now?

Yes, feel free to. Thank you for the issue! I'm glad I could help out ;)

[jonnydgreen]

Will do and thanks for the help and clarifications - I learned something new today about Windows UNC paths! :)

[bl-ue]

Yeah.

Now let me tell you something: you might be inclined to think that in https://google.com, https:// is the protocol. Actually, it's not. https: is. The actual URL is //google.com, which is simply a UNIX UNC path, because, you know, UNIX uses / for slashes and Windows uses \\ for slashes, so //google.com == \\google.com :)

[jonnydgreen]

I did not know that either, that's really interesting and things make a lot more sense now! 🚀 - thank you very much!