fix(arborist): use hosted-git-info to correctly parse resolved git urls (#8356)

When custom hosted git url are used to install dependencies it's not
installing that dependency when the url is of
`git+ssh://[email protected]:a/b/c.git` format instead of
`git+ssh://[email protected]/a/b/c.git`.

#### Cause:
` new URL(resolved)` throws an error which is getting caught and in-turn
results in removal of that node.
This behaviour was working as expected before this
#8185 which replaced
`hgi.parseUrl(resoved)` call to `new URL(resolved)`

#### Fix:
keep the `hgi.parseUrl` call to correctly parse the git url as
mentioned/explained in this PR #5758

Fixes: #8331

Author: milaninfy

workspaces/arborist/lib/arborist/reify.js

@@ -9,6 +9,7 @@ const debug = require('../debug.js')
 const { walkUp } = require('walk-up-path')
 const { log, time } = require('proc-log')
 const rpj = require('read-package-json-fast')
+const hgi = require('hosted-git-info')
 
 const { dirname, resolve, relative, join } = require('node:path')
 const { depth: dfwalk } = require('treeverse')
@@ -886,7 +887,7 @@ module.exports = cls => class Reifier extends cls {
     // Shrinkwrap and Node classes carefully, so for now, just treat
     // the default reg as the magical animal that it has been.
     try {
-      const resolvedURL = new URL(resolved)
+      const resolvedURL = hgi.parseUrl(resolved)
 
       if ((this.options.replaceRegistryHost === resolvedURL.hostname) ||
            this.options.replaceRegistryHost === 'always') {

workspaces/arborist/test/arborist/reify.js

@@ -3343,6 +3343,24 @@ t.test('should preserve exact ranges, missing actual tree', async (t) => {
     },
   })
 
+  const customGitSshPackument = JSON.stringify({
+    _id: 'gitssh',
+    _rev: 'lkjadflkjasdf',
+    name: 'gitssh',
+    'dist-tags': { latest: '1.1.1' },
+    versions: {
+      '1.1.1': {
+        name: 'gitssh',
+        version: '1.1.1',
+        dist: {
+          // this is a url that `new URL()` cant parse
+          // https://github.com/npm/cli/issues/5278
+          tarball: 'git+ssh://[email protected]:a/b/c.git#lkjadflkjasdf',
+        },
+      },
+    },
+  })
+
   const notAUrlPackument = JSON.stringify({
     _id: 'notaurl',
     _rev: 'lkjadflkjasdf',
@@ -3359,6 +3377,36 @@ t.test('should preserve exact ranges, missing actual tree', async (t) => {
     },
   })
 
+  t.test('valid custom hosted git url', async (t) => {
+    const testdir = t.testdir({
+      project: {
+        'package.json': JSON.stringify({
+          name: 'myproject',
+          version: '1.0.0',
+          dependencies: {
+            gitssh: '1.1.1',
+          },
+        }),
+      },
+    })
+
+    tnock(t, 'https://registry.github.com')
+      .get('/gitssh')
+      .reply(200, customGitSshPackument)
+
+    const getLogs = warningTracker()
+
+    const arb = new Arborist({
+      path: resolve(testdir, 'project'),
+      registry: 'https://registry.github.com',
+      cache: resolve(testdir, 'cache'),
+    })
+    await arb.reify()
+    // since it's not throwing an error on invalid url and returning undefined
+    // which trashes the node, so here we can only check if it has no warnings
+    t.strictSame(getLogs(), [], 'did not get warnings')
+  })
+
   t.test('host should not be replaced replaceRegistryHost=never', async (t) => {
     const testdir = t.testdir({
       project: {