Add mocks to NodeJS "timers" module

[swenzel-arc]

Purpose (TL;DR) - mandatory

Fix issue #466

Solution

It works like this:

Try to import timers module as timersModule. If it doesn't work, that's okay.

Upon FakeTimers.install(), iff we're installing on the globalObject and timersModule was successfully imported:
For each property that was mocked on the globalObject iff it also exists on timersModule:
Safe its name and original value from timersModule to a new property on the Clock object.
Then set it to the same property as the faked property on globalObject.

Upon FakeTimers.uninstall(), iff the object holding timersModule's originals is truthy:
For each entry set the corresponding property back to the original.

[fatso83]

I think you could probably afford adding a line to the README that we touch the "timers" module, but otherwise this generally looks fine! Just a few minor things.

[fatso83]

Stuff the tests inside of this into a suitably named new describe block in fake-timers-test.js. We usually reserve the issues* files for regression test cases.

[fatso83]

Just use Mocha's built-ins to selectively run it in supported environments

        before(function () {
            if (!timersModule) {
                this.skip();
            }
        });

Enables a bit better reporting :)

[fatso83]

They are not actual mocks, though, so could trim that bit off.

To be consistent with the methods prop, just name it timersModuleMethods

[swenzel-arc]

All done :)
Btw. do you think we should add a config flag that allows to skip altering the timers module?

[benjamingr]

Btw. do you think we should add a config flag that allows to skip altering the timers module?

No but that's a good point, it should respect the regular "should I mock setTimeout/setInterval or whatever" option (toFake)

[benjamingr]

Good job!

Does it correctly not mock methods if they're not part of toFake?

[swenzel-arc]

Does it correctly not mock methods if they're not part of toFake?

Yes it does :)

[fatso83]

Out as 10.2.0

[fatso83]

Seems like #470 made a good candidate for why this could have been behind a flag. Some weird regression hitting us.