Fix: block-scoped-var issues (fixes #2253, fixes #2747)

[mysticatea]

I remade the rule with using escope.

[mysticatea]

Oops, this implement is checking let/const also, so I will fix a bit tomorrow.

[olalonde]

👍

[mysticatea]

OK, ready for review.

escope knows relations between variables and references, so the process for this rule became simple a bit.

---

I found this rule is not emulating behaviors no-redeclare and no-shadow.
I can add the behaviors easily, so I will send a PR later.

[mysticatea]

This test case make this rule ignore a switch block, why is that?

[michaelficarra]

Looks like a bug to me.

[nzakas]

Why is that a bug? Switch statements create a block but case statements do not.

[michaelficarra]

@nzakas There's a reference outside the switch.

[mysticatea]

Ah.... I'm sorry, I got a behavior which is same as the test case, on Chrome.
I had not known this behavior, switch statement seems to not make a block. I will study later....

[michaelficarra]

That also looks like a bug. Switch statements do make a block.

[mysticatea]

I found: http://www.ecma-international.org/ecma-262/6.0/#sec-switch-statement-runtime-semantics-evaluation

Yeah, switch statement should make a block.
And firefox does correctly.

[mysticatea]

I modified a bit for switch statement. The diff is here: mysticatea@cc1af6c

[nzakas]

@michaelficarra ah, right you are. Thanks, completely missed that.

[mysticatea]

I added tests for #2967.

And my misunderstanding was corrected in #3158, so I removed a logic which allows multiple blocks (e.g. if (true) { var a = 0; foo(a); } else { var a = 1; foo(a); }).
The logic of this rule became simple more.

---

I have a question.

• Why does this rule report the same warnings as no-undef?

For example:

> echo "var a = b;" | node .\bin\eslint.js --stdin --no-eslintrc --rule "block-scoped-var:2, no-undef:2"

<text>
  1:10  error  "b" used outside of binding context  block-scoped-var
  1:10  error  "b" is not defined                   no-undef

✖ 2 problems (2 errors, 0 warnings)

I guess the warnings can be removed.
Specifically, I guess this function can be removed.

[mysticatea]

Coverage decreased.... but block-scoped-var is 100%...

[gyandeeps]

this is new. I have never seen coveralls to display here.

[mysticatea]

When the tests of node v0.12 finished lastly, it seems to decrease coverage.
It's a mystery.

[mysticatea]

I complement my question.

It needs to rewrite some tests in order to remove the duplicate warnings of no-undef.
I want to do it, but can I do it?

[nzakas]

The function can be removed, but you should create a separate issue and pull request because it's a breaking change.

[mysticatea]

OK, thank you.

So this PR is ready for review again.

[nzakas]

LGTM