Update to Node.js v12

[kumavis]

💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐

:::.    :::.    ...    :::::::-.  .,::::::            .:.   .:::.  
`;;;;,  `;;; .;;;;;;;.  ;;,   `';,;;;;''''             ;;  ,;'``;. 
  [[[[[. '[[,[[     \[[,`[[     [[ [[cccc    .:.   .:. [[  ''  ,[['
  $$$ "Y$c$$$$$,     $$$ $$,    $$ $$""""     Y$c.$$"  $$  .c$$P'  
  888    Y88"888,_ _,88P 888_,o8P' 888oo,__    Y88P    88 d88 _,oo,
  MMM     YM  "YMMMMMP"  MMMMP"`   """"YUMMM    MP     MM MMMUP*"^^

💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐 🦊 💐

I created this draft PR to help track barriers to v12 and test progress against CI

node v10 is about to go from 'active LTS' to 'maintenance'
a good time to move us on to the next LTS, v12

https://github.com/nodejs/Release#release-schedule

known issues

• dnode's dep weak does not work under v12, looks like we can disable it https://github.com/substack/dnode/blob/69c19c66d9fc078fe53dcd41788501ebd28c40e5/index.js#L26
• ethereumjs-wallet's dep scrypt is unmaintained, possible fix here Change scrypt.js to scrypt-shim ethereumjs/ethereumjs-wallet#108
• sesify has overly strict engines field
• works in development (not with scripts whitelist)
• e2e-chrome failure - ChromeDriver version

 SessionNotCreatedError: session not created: This version of ChromeDriver only supports Chrome version 79

• e2e-firefox failure ?

[2020-04-21 15:17:53.647][e2e] TypeError [ERR_MISSING_ARGS]: The "actual" and "expected" arguments must be specified

[kumavis]

noticed our install script actually had a failing command
improved here #8368

[kumavis]

( moved todo's in comments into the header )

[kumavis]

using the github url for ethereumjs-wallet doesnt work, likely some gitignore build file that it exports

/home/circleci/project/app/scripts/account-import-strategies/index.js
  2:20  error  Unable to resolve path to module 'ethereumjs-wallet'             import/no-unresolved
  3:23  error  Unable to resolve path to module 'ethereumjs-wallet/thirdparty'  import/no-unresolved

[kumavis]

couldnt get the build script to work either

[kumavis]

ok now we have a "normal" test failure in unit tests

  74 passing (151ms)
  1 failing

  1) permissions middleware
       eth_requestAccounts
         requests accounts for unpermitted origin, and approves on user approval:
      AssertionError [ERR_ASSERTION]: domain should have correct number of permissions
      + expected - actual
      -0
      +1
      at Context.equal (test/unit/app/controllers/permissions/permissions-middleware-test.js:542:14)

[kumavis]

rebased off of develop

[kumavis]

e2e-firefox failure: oh yeah this is definitely not correct, only one argument
https://github.com/MetaMask/metamask-extension/blob/develop/test/e2e/metamask-ui.spec.js#L639

@danjm what is this assertion supposed to test?
(edit: see #8388)
added here f81676f#diff-b00929238ad0453812af55a40336e889R518

TypeError [ERR_MISSING_ARGS]: The "actual" and "expected" arguments must be specified

curious how that test passes as is! is it matching false to undefined ?

[kumavis]

on this image

circleci@9e3b575b76ac:~$ google-chrome --version
Google Chrome 81.0.4044.122 

[kumavis]

test fix cherry-picked here #8388

[kumavis]

DevToolsActivePort file doesn't exist

😿

prev attempts:
#7991
#7996

      .addArguments(
        '--disable-dev-shm-usage',
        '--headless',
        '--no-sandbox',
      )

some recommendations in threads

do not run chrome as root.
Also, as a personal suggestion, if dealing with custom Chromium-like binaries, use Operadriver like so: driver = webdriver.Chrome("operadriver", options=options), you won't run into so many issues regarding versions.

1. Explicitly set the remote debugging port as an argument
2. Make your driver headless, using chrome Options

options.add_argument("--no-sandbox")
options.add_argument("--remote-debugging-port=9222")
options.headless = True
command_executor = "http://localhost:4444/wd/hub"
driver = webdriver.Remote(command_executor, desired_capabilities=options.to_capabilities())
driver.get("https://google.com")

[kumavis]

wao somehow e2e-chrome passed ??
edit: failed again, guess its a race condition

[kumavis]

ok back to the chromedriver problems

more thread finds

Turns out we were setting chromeOptions: instead of goog:chromeOptions:.
Not sure what the difference is, since chromeOptions: always worked for us in the past.

[kumavis]

 I would suggest you try to use Operadriver instead like so: Selenium::WebDriver::Chrome.driver_path = '/usr/bin/operadriver'.

It's Chrome's responsibility to create DevToolsActivePort file. So if it doesn't exist, it generally means Chrome didn't start up correctly. It's usually not possible for us to remotely find out why Chrome didn't start up.

https://bugs.chromium.org/p/chromedriver/issues/detail?id=3415&can=2&q=DevToolsActivePort

[kumavis]

Since the error is intermittent, add retry loop.

[kumavis]

new error:

       can send a simple transaction from one account to another:
     JavascriptError: javascript error: Cannot read property 'shadowRoot' of null
 
  (Session info: headless chrome=81.0.4044.122)
      at Object.throwDecodedError (node_modules/selenium-webdriver/lib/error.js:550:15)
      at parseHttpResponse (node_modules/selenium-webdriver/lib/http.js:563:13)
      at Executor.execute (node_modules/selenium-webdriver/lib/http.js:489:26)
      at processTicksAndRejections (internal/process/task_queues.js:97:5)
      at thenableWebDriverProxy.execute (node_modules/selenium-webdriver/lib/webdriver.js:699:17)
      at ChromeDriver.getExtensionIdByName (test/e2e/webdriver/chrome.js:56:12)
      at Function.build (test/e2e/webdriver/chrome.js:33:25)
      at buildBrowserWebDriver (test/e2e/webdriver/index.js:28:14)
      at buildWebDriver (test/e2e/webdriver/index.js:11:65)
      at withFixtures (test/e2e/helpers.js:20:24)
      at Context.<anonymous> (test/e2e/tests/simple-send.spec.js:14:5)

[kumavis]

seems like an improvement, maybe a retry wrapper will make the difference

[kumavis]

closing in favor of #9514