known_host host identity fingerprint duplication

[legraphista]

Referring to issue #107
I have looked through the code and found that the only way that this error can happen

/app/node_modules/neo4j-driver/lib/v1/internal/ch-node.js:214
      for (var i = 0; i < pending.length; i++) {
                                 ^

TypeError: Cannot read property 'length' of null
    at /app/node_modules/neo4j-driver/lib/v1/internal/ch-node.js:214:34
    ...

is if the trust certificate callback is called twice:

    this._handleConnectionError = this._handleConnectionError.bind(this);

-> this._conn = connect(opts, function () {
      if (!self._open) {
        return;
      }

     <...>
      var pending = self._pending;
      self._pending = null;
      for (var i = 0; i < pending.length; i++) {
        self.write(pending[i]);
      }
    }, this._handleConnectionError);

because it it setting self._pending to be null.

The trust certificate function can be called twice if the loadFingerprint callback is called twice.
After looking at the known_hosts file

localhost:7687 fabacd62dbd94...
localhost:7687 fabacd62dbd94...

We see that there are duplicate entries, causing this to fire more than once:

require('readline').createInterface({
    input: fs.createReadStream(knownHostsPath)
  }).on('line', (line)  => {
    if( line.startsWith( serverId )) {
-->   found = true;
-->   cb( line.split(" ")[1] );
    }
  }).on('close', () => {
    if(!found) {
      cb(null);
    }
  });

But why are there duplicates?
Under the circumstances that the known_hosts file exists but is empty (manually created by someone) and if the app is calling for two (or more) driver sessions in the same process tick

const session = driver.session();
const session2 = driver.session();

The driver doesn't have time to update the known_hosts file and thus creates two (or more) entries of the same thing.
After that, the next session that is created:

setTimeout(() => {
    const session3 = driver.session();
}, 1000);

will throw the error.

[zhenlineo]

Thanks so much for helping us investigating on this issue! So sorry we are running a bit low bandwidth to helping with the related issue #107 now.

We will come back to this and get it fixed next week. You are most welcome to send us a PR if you have got an idea in mind! If it is possible, please include a test to verify the PR have the correct fix and prevent the same problem from happening again.

Thanks again!
Zhen

[legraphista]

Sure thing @zhenlineo. I will shortly come up with a PR.

[legraphista]

Fixed with #110