fix(browser): allow updating total specs count

[matz3]

This change allows providing the total specs count after the tests have
been started.
The count will be updated in case it has been provided already.
This can be uselful for some adapters where the total specs count can
not be determined before starting the first test.

Regarding tests
I wanted to update the tests and/or create a test case but the onInfo tests are not clear to me:

• 'browser_dump' doesn't seem to be a valid event
• I'm not sure whether the behavior of the tests is really expected or just works by chance.

So I would be happy to get feedback from you before writing a test case.

karma/test/unit/browser.spec.js

Lines 102 to 137 in c277a6b

	describe('onInfo', () => {
	beforeEach(() => {
	browser = new Browser('fake-id', 'full name', collection, emitter, socket)
	})
	it('should emit "browser_log"', () => {
	const spy = sinon.spy()
	emitter.on('browser_log', spy)
	browser.state = Browser.STATE_EXECUTING
	browser.onInfo({log: 'something', type: 'info'})
	expect(spy).to.have.been.calledWith(browser, 'something', 'info')
	})
	it('should emit "browser_info"', () => {
	const spy = sinon.spy()
	const infoData = {}
	emitter.on('browser_info', spy)
	browser.state = Browser.STATE_EXECUTING
	browser.onInfo(infoData)
	expect(spy).to.have.been.calledWith(browser, infoData)
	})
	it('should ignore if browser not executing', () => {
	const spy = sinon.spy()
	emitter.on('browser_dump', spy)
	browser.state = Browser.STATE_CONNECTED
	browser.onInfo({dump: 'something'})
	browser.onInfo({total: 20})
	expect(browser.lastResult.total).to.equal(0)
	expect(spy).not.to.have.been.called
	})
	})

[johnjbarton]

• 'browser_dump' doesn't seem to be a valid event

I guess the test is attempting to cover:
https://github.com/karma-runner/karma/blob/master/lib/browser.js#L63

However it tests something "not happening" and many things do not happen when code is changed. Maybe the event was originally browser_dump and was renamed, but the test was not corrected. Please correct it if you can.

[johnjbarton]

If you guard the update to results with state EXECUTING then the tests will pass. You can add a similar test with mock state EXECUTING to verify the change.

[matz3]

In order to fix the test should ignore if browser not executing I would suggest to wrap all those if-clauses with a check for state EXECUTING.
Should this.refreshNoActivityTimeout() be still called then?

[johnjbarton]

I think we should log events in any state, they could be error messages for corner cases. So remove the line in the test that checks the .dump path.

The this.refreshNoActivityTimeout() should always be called, it signals that the browser is still on the line.

[matz3]

Makes sense. I've pushed my changes.

[johnjbarton]

Thanks!

[matz3]

Thanks for the quick handling 👍