Fixes runtime exception when comparing inequal lists with over 100 elements

[robinheghan]

This has the same solution as #1018, but uses fuzz testing instead of unit tests. This PR is also not conflated with upgrading the testing code for 0.19.

[jerith666]

Sorry for the long delay in responding here, but I finally had a chance to run the tests at this commit (with the 0.19.1 updates from #1059), and they don't pass for me:

elm-test 0.19.1
---------------

Running 1268 tests. To reproduce these results, run: elm-test --fuzz 100 --seed 193469607216503 /home/matt/git/elm/core/tests/tests/Main.elm

↓ Test.Equality
↓ Equality Tests
↓ List equality
✗ Simple comparison

    This test failed because it threw an exception: "TypeError: Cannot read property '$' of undefined"



TEST RUN FAILED

Duration: 65658 ms
Passed:   1267
Failed:   1

[jerith666]

I also tried running the tests added by your 3520419 commit with the fix from my 6eff09b commit, and the tests pass.

[jerith666]

Finally, I also tried running the tests added by my 6eff09b against the fix from your 3520419; the tests also fail in that case.

[jerith666]

I think what this all means is that it's the typeof x !== 'object' || x === null || y === null check that prevents Tuple2s containing undefined from being pushed onto the stack, not the x === y check.

[robinheghan]

Hmm. So the typeof check should be moved up to before the depth check?

[jerith666]

I think so, yes. My fix moves both the x === y and the typeof ... checks above the depth check.

I should say that I don't have any deep knowledge of how Elm objects get represented as JS objects that's informing me here, I'm just going by what I saw in the browser debugger when I was exploring the original failures from the SSCCE I posted in #1011. That's maybe why I got a bit more elaborate in my test cases than you, as well?

I'm happy to open a new PR with my fix rebased onto the current tip of master if that helps.

[jerith666]

Yep, that leaves Utils.js identical to what I had in #1018, so it should work!

[harrysarson]

Should we add the extra tests to prevent regressions?

[jerith666]

I'm always a fan of more tests -- esp. for something as critical as elm/core! :)

I can make a PR with the combined set of tests if you'd like.

[jerith666]

My new PR is #1062