Fix sessionStorage exceeds quota error in Safari

[timlindvall]

Fixes issue #86. We're adding an attempt to write to sessionStorage before we set the hasSessionStorage flag. This catches scenarios where sessionStorage is available but cannot be written to.

I'm also contemplating adding another try/catch block in setState() on the theory that the writeability of sessionStorage might change from page load to when we try writing to it later (if in the interim lots of data ends up getting written to sessionStorage). Thoughts?

Testing Done

• Ran /demo tour in Safari private mode. Verified state was being written to cookie and that no console errors were thrown.
• Existing Mocha tests pass.

[kate2753]

The fix looks good to me. I could not find any API that we can rely on to check available storage space, so try\catch seems like a way to work around this.

I agree, we need to have try\catch in "setState" method as well, since availability of sessionStorage may change.

[kate2753]

Code looks good to me. Any luck with unit testing this?

[timlindvall]

No luck, I'm afraid. I've created a potential unit test that passes (https://gist.github.com/zimmi88/80e9d72df993802efa8f) but, because we aren't tearing down and re-initializing Hopscotch between each test, once we trip the sessionStorage flag, we don't have the means by which to flip the switch back to normal. This causes later tests, which assume we're writing to sessionStorage, to fail.

Let's consider adding an issue that minimizes these dependencies between tests. We could also consider building into the build process special targets for unit tests to run against that make the internal methods of Hopscotch available for tests (possibly part of work for #72).

[kate2753]

Sounds good. I'll merge this in tomorrow by EOD if there are no objections.