yarn 1.x.x invalid url with node 18.17.0 VS 18.16.1

[FrederickEngelhardt]

Version

18.17.0

Platform

Mac OS (aarm64 & intel)

Subsystem

No response

What steps will reproduce the bug?

Using node 18.17.0 and installing yarn 1.x.x will cause this error.

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

1. Every time after installing 18.17.0 from 18.16.1. So far, this issue is specific to a monorepo being used.
2. I'm having a hard time replicating it on init projects that use the npm registry. The project with the issue uses a artifactory jfrog registry which points to a couple other internal artifactory registries.

Other tips to replicating this issue:

• If you are swapping node from a minor version make sure to reset your yarn cache to replicate the issue 100% of the time. IE yarn cache clear. Project level cache should be enough.

Example commands

nvm install 18.17.0; nvm use 18.17.0; npm i -g yarn; yarn cache clean; yarn;

What is the expected behavior? Why is that the expected behavior?

yarn should install dependencies and resolve those urls.

What do you see instead?

Error and unexpected error has occurred "Invalid Url".

Additional information

If this is an expected behavior change 18.16 - 18.17.0 (breaking yarn package manager on 1.x.x) we can close this issue.

Yarn 1.x.x is still used in many projects so this may affect a bunch of users.

Temp workarounds

1. Stay on Node 18.16 and lock the version. IE with .nvrmc or _node-version change it from 18 to 18.16

Possibly related URL issues:

• URL.parse fails on URLs with comma separated hosts #48850

[anonrig]

Can you share the URL causing this error? cc @nodejs/url

[FrederickEngelhardt]

 verbose 0.688693792 TypeError [ERR_INVALID_URL]: Invalid URL
     at new NodeError (node:internal/errors:405:5)
     at Url.parse (node:url:445:17)
     at urlParse (node:url:167:13)
     at GitResolver.isVersion (/Users/<path>.yarn/releases/yarn-1.22.19.js:38701:21)
     at getExoticResolver (/Users/<path>.yarn/releases/yarn-1.22.19.js:29731:18)
     at /Users/<path>.yarn/releases/yarn-1.22.19.js:38152:78
     at Generator.next (<anonymous>)
     at step (/Users/<path>.yarn/releases/yarn-1.22.19.js:310:30)
     at /Users/<path>.yarn/releases/yarn-1.22.19.js:321:13
 error An unexpected error occurred: "Invalid URL".
 info If you think this is a bug, please open a bug report with the information provided in "/Users/<path>yarn-error.log".

[aduh95]

Looking at the stack trace, it seems to be related to the legacy Url class, not the WHATWG URL class.

The related code on Yarn source is at https://github.com/yarnpkg/yarn/blob/158d96dce95313d9a00218302631cd263877d164/src/resolvers/exotics/git-resolver.js#L38

/cc @arcanis

[aduh95]

The throw is located at

node/lib/url.js

Lines 444 to 446 in a39b8a2

	if (this.hostname === '' || forbiddenHostChars.test(this.hostname)) {
	throw new ERR_INVALID_URL(url);
	}

The hostname is set at

node/lib/url.js

Line 433 in a39b8a2

this.hostname = toASCII(this.hostname, true);

And toASCII is imported from internal/idna:

node/lib/url.js

Line 34 in a39b8a2

const { toASCII } = require('internal/idna');

Which itself reexports it from internal/url:

node/lib/internal/idna.js

Lines 3 to 4 in a39b8a2

	const { domainToASCII, domainToUnicode } = require('internal/url');
	module.exports = { toASCII: domainToASCII, toUnicode: domainToUnicode };

The list of commits that have touched internal/url since v18.16.0 is not small:
https://github.com/nodejs/node/commits/a39b8a2a3e8414c8a54650bdfe4a46998282d88a/lib/internal/url.js

/cc @nodejs/url

[akkunchoi]

$ node --version && node -e "console.log(require('node:url').parse('x://0.0,1.1/').host)"
v18.16.1
0.0,1.1

$ node --version && node -e "console.log(require('node:url').parse('x://0.0,1.1/').host)"
v20.5.0
0.0,1.1

$ node --version && node -e "console.log(require('node:url').parse('x://0.0,1.1/').host)"
v18.17.0
node:url:445
          throw new ERR_INVALID_URL(url);
          ^

TypeError [ERR_INVALID_URL]: Invalid URL
    at new NodeError (node:internal/errors:405:5)
    at Url.parse (node:url:445:17)
    at Object.urlParse [as parse] (node:url:167:13)
    at [eval]:1:33
    at Script.runInThisContext (node:vm:123:12)
    at Object.runInThisContext (node:vm:299:38)
    at node:internal/process/execution:79:19
    at [eval]-wrapper:6:22
    at evalScript (node:internal/process/execution:78:60)
    at node:internal/main/eval_string:28:3 {
  input: 'x://0.0,1.1/',
  code: 'ERR_INVALID_URL'
}

$ node --version && node -e "console.log(require('node:url').parse('x://0.0,1/').host)"
v18.17.0
0.0,1

$ node --version && node -e "console.log(require('node:url').parse('x://a.a.a.a,b.b.b.b/').host)" 
v18.17.0
a.a.a.a,b.b.b.b

It seems to result in an Invalid URL error under the following conditions

• v18.17.0
• When specifying a URL with a hostname that includes numbers, dots, and a comma

I use a comma in the hostname when specifying multiple hosts in MongoDB.
https://www.mongodb.com/docs/manual/reference/connection-string/#standard-connection-string-format

[aduh95]

Starting bisecting

[richardlau]

I guess this is a duplicate of #48850?

[aduh95]

0dc485e is the first bad commit /cc @anonrig

[arcanis]

Seems it's not a problem with Yarn itself but, out of curiosity, why is your custom registry hostname containing commas? I didn't even know this was supported 😅

[lemire]

The ada URL parsing library itself has no beef with commas, as you can see in the playground...

https://www.ada-url.com/playground?url=x%3A%2F%2F0.0%2C1%2F