Fix clobbering of overridden canvas methods

[STRML]

The method may be overridden again after we override,
as is done in the `hidpi-canvas-polyfill`. So if we
restore the method to the original as we see it, we'll
wipe out the polyfill.

I'd like to write tests for this, but I see very little related to canvas fingerprinting, and what is there is pretty inscrutable (where is fingerprinting.html?)

Fixes #2657, fixes #1226, fixes #1854.

[ghostwords]

Hello, and thank you for the report and the PR! I will take a proper look next week.

fingerprinting.html lives in https://github.com/EFForg/privacybadger-test-fixtures. (We should probably document where fixtures live somewhere in our tests doc.)

Let me know what other questions you have. Thanks again!

[STRML]

Thanks @ghostwords. Perhaps this should be same-repo, and we just run a local webserver rather than hit the github.io page? Adding or modifying a test requires now coordinating two PRs, which is quite difficult. Regardless, I'll make a PR there that correlates to this one, but it will need to be accepted and the tests re-run here.

[ghostwords]

Yeah, the coordination is a bit of a pain. We used to host test fixtures on various public JS playground/gist hosts, but for reasons like reliability and control over what actually gets served we migrated to the two GH Pages repos we have now. #928 has a bit of history and links to some relevant commits.

A local fixtures server would be nice, but we need support for things like https and distinct-to-Privacy Badger hostnames (to simulate first and third-party interactions). (Localhost domains are probably not a good fit since they should probably be ignored: #817)

[STRML]

That's fair enough. So we'd have to run a system-level dns resolver and point the local browsers to it... quite a song & dance. Doubt it's worth the effort.

[STRML]

Build failing, I'll ping when this is ready for merge.

[ghostwords]

I should have some time to look into what's going on with the test next week, if that's helpful.

[STRML]

It would be, thanks. When running locally I have no issue.

[ghostwords]

I think the test doesn't work on Travis because it doesn't run in a HiDPI environment. We could fake it with something like window.devicePixelRatio = 2;. Still looking though.

[ghostwords]

Looks like that was it, the polyfill wasn't kicking in on Travis. I added the devicePixelRatio override, the test should pass now.

[ghostwords]

Nice, thank you! Looks good so far outside of some nits. Let me know if I should take care of them myself.

[STRML]

@ghostwords We're green, thanks a lot for the help and for catching the fixture issue, I wouldn't have caught that.

I can clean up the history if you like, or if you're just going to squash it should be no problem.

[ghostwords]

Thank you for the contribution! I'll finish testing next week, expect to merge then in time for the next Privacy Badger update.

I won't squash merge as I would rather preserve the individual commits. You're welcome to clean up history, but it's not necessary.

[ghostwords]

Thanks again!