The existing documentation was redundant and confusing.

Closes #3679

Follow up to #3893.

Part of #3893

A little bit excessive, but this will really help debugging issues with
Blocking Proxy.

…nce (#3970)

…#3989)

Fixes #3986. Also error if jvmArgs isn't an array.

Fix for #3978.

Our initial plan to allow setting --disableChecks with an environment variable is insufficient, since the custom framework isn't even require()'d until after the config is parsed. This moves the unknown flag check into the runner, and gives frameworks a way to specify extra flags they accept.

Replace browser.rootEl with browser.setAngularRoot(), which changes the root element in a promise on the control flow. Note that browser.rootEl will immediately return the current value, but browser.setAngularRoot() will return a promise that resolves during the next step in the control flow.

Also update to BlockingProxy 0.0.3, which allows changing rootSelector.

…pter (#3988)

Closes #3894,
#3908, and
#3909

#3992)

Wrapping it in a `q` promise is blocking #3899

Closes #3902

Custom frameworks might not make this change but it'll be fine.  It'll only be a
problem in edge cases and they probably weren't returning the right promise
before anyway.

Closes #3898

…4013)

Closes #3985

Also allows `browser.restart` to work when the control flow is disabled, and
fixes it for forked browsers.

Closes #3899 and
#3896

…rk (#4015)

Closes #3900

…4018)

Closes #3903

…se from `waitForAngularEnabled` (#4021)

Minor breaking change since `waitForAngularEnabled` no longer returns a boolean

Part of #3904

Chaining `browser.get` has proved surprisingly complex, so I'll do that in a different PR

Also fixed a minor bug in `lib/clientsidescripts.js` while debuging

Closes #3904

… config (#4023)

Closes #3691

…ibility (#4006)

Add test cases to reproduce the missing element race conditions possible in
expected condition methods `visibilityOf`, `textToBePresentInElement`,
`textToBePresentInValue` and `elementToBeClickable`.

Add error handler `falseIfMissing` to all expected conditions that depend
on the presence of an element.

Expected conditions check the presence of an element before other checks,
but when an element is removed exactly in the moment after the `isPresent`
and before `isDisplayed` in `visibilityOf` the condition used to fail.

This solution does not handle missing elements in (`isEnable`, `isDisplayed`, `isSelected`) and focused only on expected conditions (see
#3972)

This problem was also referenced in
#3578 and
#3777

The second test case fails. Updating so that it passes.

Closes #3185

…ow is disabled (#4029)

Also added some tests to `spec/ts/noCF/smoke_spec.ts` double checking that the control flow is off

I decided to address this comment:

#4000 (comment)

While doing do I decided to take on this TODO:

https://github.com/angular/protractor/blob/ccf02ab5f1070f0d7124318dc0099252f3c747e2/lib/browser.ts#L38

One possible issue here is that `ProtractorBrowser` only copies over methods, so it doesn't actually
implement `WebDriver`'s interface.  This isn't a problem right now, since `WebDriver`'s interface
only has functions on it.  But it could be a problem in the future.

…art of webdriver (#3966)

Closes #3744

…#4038)

…#4037)

I also fixed a minor issue where `internalRootEl` wasn't being set when blocking proxy was being
used.  I also just cleaned up our internal uses of `this.rootEl`.

Closes #4032

This adds two options, both of which are implemented with Blocking
Proxy. --webDriverLogDir will create a readable log with timing
information of webdriver commands in the specified directory.

--highlightDelay will pause before clicking on elements or sending keys.
While paused, the element that's about to be affected will be
highlighted.

Running `webdriver-manager update` will now by default grab the latest
versions of all binaries (standalone, chromedriver, iedriver, gecko
driver). You can continue to pin to a specific versions using the
command line option. Example `webdriver-manager update --versions.chrome 2.20`. As of this release the latest versions are:

- gecko v0.14.0
- selenium-standalone 3.0.1 is the latest jar file; however, we recommend 3.0.0-beta4. See note below on Firefox support.
- chromedriver 2.27
- iedriver 3.0.0

A note on FireFox support: Selenium standalone 3.0.1 has a bug which prevents it from working with any version of FireFox. We have tested version 3.0.0-beta4 and know that it works with FireFox 51, and we expect that the 3.0.2 release will also work.

closes #4033

…se chaining (#4047)

I noticed I missed `waitForAngularEnabled` in
#4037.  This commit fixed that.

While I was at it I fixed a minor error where the promises implicitly created by
setting `rootEl` and `ignoreSynchronization` weren't getting chained properly.

Also fixed minor (so minor I think it was impossible to trigger) where
browser.plugins_ could be undefined.

Also added the ability to run `generate-docs.sh` against a custom
branch/tag/commit by doing `./scripts/generate-docs.sh branch_name`.  Still
defaults to the version from package.json though.

Also updated release.md to tell people to check that doc generation/the website
is working BEFORE doing a release.  Needing a change in the codebase is more
likely now that we're compiling down to es5 only this one time, and ideally
any changes to the codebase would happen before release.

Also including a minor change in the codebase where we wrap a promise from
blocking proxy in a webdriver promise.  This probably should have been done the
whole time, but is unlikely to matter either way at runtime, since blocking
proxy wouldn't be returning a webdriver promise regardless.  Mostly did I did
this to fix typechecking.

- move noCF tests to install and fix reference to protractor
- changed element.ts to not use keyof
- remove gulp task tsc:spec

…ainst (#4066)

Closes #4054

* Removed redundent information in `/docs/plugins.md`.  Now redirects to
  `/lib/plugins.ts`.
* Comment cleanup in `/lib/plugins.ts`
* Export some necessary types in `/lib/index.ts`

* Moved our `/spec/install/noCF/` tests back to `/spec/ts` and restored
  `ts_spec_config.json`.
  * It turns out that typescript@~2.0.0 does support `-p config_file.json`.
    The thing it doesn't support is pointing `include` to a folder.  So rather
    than moving everything in #4062, we just needed to use globs.
* Split `smokeSpec.ts` into `is_disabled_spec.ts` and `element_spec.ts`.
* Renamed `noCFSmokeConf.ts` to `noCFBasicConf.ts` so it is more clearly
  analogous to `basicConf.js`.
* Added test using blocking proxy for `noCFBasicConf.ts`.

Also added generated `exampleTypescript/` files to `.npmignore`

* Added information about `SELENIUM_PROMISE_MANAGER` to `docs/control-flow.md`,
  including pointing to `/spec/ts/` for examples
* Added `docs/async-await.md`, which redirects to
  `exampleTypescript/asyncAwait/README.md`.
* Updated `exampleTypescript/asyncAwait/README.md`, including pointing to
  `/spec/ts/` for more examples.
* Added `docs/typescript.md`, which redirects to `/exampleTypescript/`.
* Added information about `@types/jasminewd2` to `exampleTypescript/README.md`.

Website updates to come in a future change.

Closes #3692.

Also added npm command `compile_to_es5` to make running the website tests easier

…4081)

Had to fix link parsing in website generation while I was at it.

Also updated docs/toc.md, which was very out of date.

* Use ES6/TS native Promise in place of Q.
* Removing superfluous Q imports. Breaking tap into error-rethrowing segment as well.
* Updated arbitrarily-failing lint code

The `err` object doesn't have the `code` property any more (Selenium Server Standalone 3.3.1 + IEDriver win32 3.3.0), so we need a new way to detect those errors. See #841

This reverts commit 4e73d5e.

Change the async/await example to use ts-node. Also, when using
SELENIUM_PROMISE_MANAGER=false, you can debug tests using
`node --inspect` and the chrome debugger.

Also, update the docs around timeouts to include Angular vs AngularJS

Also clang-format with 1.0.49

Anchor names on Selenium wire protocol wiki page changed, rendering our
links outdated. Fix them.

Fixes #4196

This fixes a problem I encountered similar to #4196 - where `stackTrace` is listed as an option but an error is given saying it's an "unknown extra flag"

By default, it wasn't returning anything. Now it returns a promise that
resolves to internalAngularAppRoot. Fixes #4233

Chromedriver requires that certain options always be passed as an array.
Optimist passes --single-option as a string instead of an array which is
invalid. This ensures that we always pass an array, even if a single
option is passed via the cli.

Fixes #4050

Also, start using Yarn, it's pretty cool.

Disable Firefox directConnect test until #4253 is fixed.

paths.

Fixes #4011

…ar js

Closes #4261

The Protractor documentation says that By.js is inherited from Selenium Webdriver, and the types for that define the function signature for `js` like I have done in this PR.

https://github.com/DefinitelyTyped/DefinitelyTyped/blob/master/types/selenium-webdriver/index.d.ts#L2441

Fixed incorrect usage of the cucumber dry run option.

Fixed incorrect example of setting window size with headless Chrome. When `800x600` is used it is ignored. The correct syntax is `800,600`.

…to nowhere

Fixes #4371

If local driver provider is used, `seleniumAddress` appears in
config only after `setupDriverEnv()` is resolved.