Listen hashchange if popstate = false

[PaulMaly]

Seems we should add an event listener to hashchange if popstate option equals false and hashbang option equals true. In this case, right now back navigation via history.back() not working.

[coveralls]

Coverage remained the same at 92.531% when pulling 502db29 on PaulMaly:patch-3 into 7f8e01d on visionmedia:master.

[PaulMaly]

@matthewp Any news here? Is this patch reasonable?

[matthewp]

It's hard to keep all of this in my head correctly. We've had a lot of modification of this part of the code. I'm a little bit afraid that we're playing bug whack-a-mole. We fix one thing but break something else. To prevent that going forward I feel we really need tests for each PR. So if you can add a test for this case then I think it's good to go.

[PaulMaly]

@matthewp I believe it's a very obvious case to be tested. Seems we should trigger _onpopstate somehow any way in all possible cases. There's the case when _onpopstate never be triggered:

page.start({
    popstate: false,
    hashbang: true,
});

Because we've only two places where _onpopstate method could be triggered:

if(this._popstate)

and

if(this._hashbang && hasWindow && !hasHistory)

So, in a case when we've history and don't want to popstate, _onpopstate method will never be triggered.

It's very bad because if in this case, we'll press the browser back button, the router won't be informed about location changed.

[PaulMaly]

@matthewp Any thoughts?

[matthewp]

Not on your last comment, no, sorry. But, what needs to happen is we need a test that verifies that a hashchange listener is added when { popstate: false }, that much makes sense to me.