Issue 427 - Desugar arrow functions to normal or bound functions

Correctly deciding if the bind call is necessary is kind of hard, so I decided for some kind of best effort approach.

[tschuster, 2014-05-13 18:07:37]

On 2014/05/10 12:46:28, tschuster wrote:
r?

[Wladimir Palant, 2014-05-15 06:53:44]

Please make sure that (foo) => 1, (foo) => {return 1;} and (foo) => {return
this;} are properly tested via autotest. Other than it's mostly nits - it's up
to you whether you want to reject new Function() and eval() calls.

http://codereview.adblockplus.org/5417063522762752/diff/5629499534213120/scri...
File scripts/abprewrite.js (right):

http://codereview.adblockplus.org/5417063522762752/diff/5629499534213120/scri...
scripts/abprewrite.js:473: if (["this", "Function", "eval", "=>"].some(search =>
body.contains(search))) {
I think this would be better here:

  if (/\b(this|Function|eval)\b/.test(body))

While this is still a hack, we should look for keywords rather than random
substrings.

And of course we should never generate code dynamically - already because CSP
restrictions in Chrome don't allow it. So I wonder whether creating
modifyCallExpression()/modifyNewExpression() functions that will reject new
Function(...) and eval(...) would be a better course of action here.

http://codereview.adblockplus.org/5417063522762752/diff/5629499534213120/scri...
File scripts/astDecompile.js (right):

http://codereview.adblockplus.org/5417063522762752/diff/5629499534213120/scri...
scripts/astDecompile.js:253: str = "(" + ast.params.map(decompileAST).join(",")
+ ")";
Declare this variable?

Also a nit: .join(", ") for consistency with function declarations and for
better readability.

[tschuster, 2014-05-18 10:55:23]

I just decided to ignore eval and Function for now, I can open a new bug if we
really want to ban them. Good call on that RegExp.

[tschuster, 2014-05-24 11:10:10]

r?

[Wladimir Palant, 2014-05-25 08:23:11]

LGTM, I somehow missed the update here.