[Electron] TypeError: Cannot read property 'randomBytes' of undefined

[spkjp]

We run into the following issue when using bcrypto 4.1.0 inside Electron:

    TypeError: Cannot read property 'randomBytes' of undefined

      42 |    */
      43 |   static getAddress (passphrase, pubKeyHash) {
    > 44 |     return Identities.Address.fromPassphrase(this.normalizePassphrase(passphrase), pubKeyHash)
         |                               ^
      45 |   }
      46 | 
      47 |   static getAddressFromPublicKey (publicKey, pubKeyHash) {

      at Object.randomBytes (node_modules/bcrypto/lib/native/random.js:20:18)
      at Function.randomBits (node_modules/bcrypto/lib/js/bn.js:3459:25)
      at Function.random (node_modules/bcrypto/lib/js/bn.js:3502:22)
      at ShortPoint._getBlinding (node_modules/bcrypto/lib/js/curves.js:785:24)
      at ShortPoint.precompute (node_modules/bcrypto/lib/js/curves.js:731:40)
      at SECP256K1.precompute (node_modules/bcrypto/lib/js/curves.js:160:12)
      at ECDSA.get curve [as curve] (node_modules/bcrypto/lib/js/ecdsa.js:52:19)
      at ECDSA.publicKeyCreate (node_modules/bcrypto/lib/js/ecdsa.js:228:20)
      at Function.fromPrivateKey (node_modules/@arkecosystem/crypto/src/identities/keys.ts:18:34)

console.log(process.versions)
 {
        node: '12.3.1',
        v8: '7.4.288.27-node.18',
        uv: '1.29.1',
        zlib: '1.2.11',
        brotli: '1.0.7',
        ares: '1.15.0',
        modules: '72',
        nghttp2: '1.38.0',
        napi: '4',
        llhttp: '1.1.3',
        http_parser: '2.8.0',
        openssl: '1.1.1b',
        cldr: '35.1',
        icu: '64.2',
        tz: '2019a',
        unicode: '12.1'
      }

From what I can tell loading the native module succeeds:
https://github.com/bcoin-org/bcrypto/blob/master/lib/random.js#L10

but calling randomBytes fails because crypto is missing. Manually forcing to ./node/random resolves it (can we get an environment variable for it? 😛 ).

I also tried a npm rebuild without success.

Many thanks in advance.

[pinheadmz]

Possible duplicate issue: #15

Looks like the recommendation for now is to fall back to the JavaScript methods.

[pinheadmz]

The environment variable is NODE_BACKEND=<js|node>

[spkjp]

Thanks for the reply. I tried playing around with NODE_BACKEND=js before and the issue still persists:

    console.log node_modules/bcrypto/lib/native/binding.js:9
      js
    console.log node_modules/bcrypto/lib/native/binding.js:11
      boom
    console.log node_modules/bcrypto/lib/random.js:11
      all ok

In other words, it throws the exception here:
https://github.com/bcoin-org/bcrypto/blob/master/lib/native/binding.js#L9

but does not enter the catch block here:
https://github.com/bcoin-org/bcrypto/blob/master/lib/random.js#L12

[chjj]

      at Object.randomBytes (node_modules/bcrypto/lib/native/random.js:20:18)
      at Function.randomBits (node_modules/bcrypto/lib/js/bn.js:3459:25)

This stack trace is confusing. Why does electron succeed at dynamically loading the binding? The random object is always on the exposed binding no matter what. Is electron does giving us an empty object for compatibility purposes?

but does not enter the catch block here

@supaiku0, what about NODE_BACKEND=node. Does that give any better results?

We only fall back to the JS random backend in the browser (which uses getRandomValues() instead of the crypto module).

I'm still fuzzy on what electron does/doesn't support and there doesn't seem to be a simple writeup anywhere. Personally I prefer not to touch electron, so I'm basically shooting in the dark here.

[chjj]

Also, is there an easy way to detect electron? Do they expose a property on the process object or something?

[tynes]

It looks like process.versions.electron is set when running in electron.

electron/electron#2288 (comment)

[spkjp]

@chjj @tynes This is a bit embarassing, but it's not actually Electron at fault here. I was only able to reproduce the error when running unit tests via jest and for some reason I mixed it up with Electron. I can no longer reproduce it after I wiped everything on my Linux system. It must have been a local issue on my side.

Sorry, for the confusion!