Check minified code for syntax errors (e.g. by trying to construct an AST)

[mgol]

* Description*

Enabling strict mode uncovered an UglifyJS issue (mishoo/UglifyJS#1052) where it was creating code with SyntaxErrors. Perhaps we should try to generate the AST of the generated minified code to check if it doesn't contain syntax errors.

[markelog]

Validate AST? Sounds awfully like a linter to me. I think @gibson042 meant to just check if we can build an AST.

[mgol]

@markelog updated the title & description.

[markelog]

And it also sounds like, that this could be (if proven viable) implemented on uglify side, like self check of sorts which could be run if correct option is passed (since it would reflect on the performance).

@mishoo what do you think?

[mishoo]

@markelog Not sure I understand what's your question, but I do agree that @mgol reported a valid issue on UglifyJS. We already have a PR that fixes it; I didn't have time to look into it yet, though — please test if possible.

[markelog]

Not sure I understand what's your question

We have an idea, which we currently considering to implement in jquery build process - after uglify minifies the source, run it through parser once again and try to build an AST, if AST is builded, then we could be sure that there is no obvious syntax errors.

Yes, we do run minified version through test suite on various set of browsers with browserstack, but because mentioned error was noticed only at that point, we almost reached our cap on that service, which is not preferred, proposed approach might caught such errors on early stage if automated.

And yes, most errors will not be revealed at that stage, but it seems some of them will, like mishoo/UglifyJS#1052?

I'm wondering if second parser run could be introduced on uglify side as a self check and overall, if you think this idea worth something.

Other option is to run linter on minified version, with limited amount of enabled rules, but running linter on minified version sounds kinda... or is it?

[kzc]

A second uglify parse would not have caught this issue:

"use strict";

!function() {
    if (window) function i() {}
}();

as uglify parses it without error:

"use strict";!function(){if(window)function i(){}}();

A full-fledged linter from a different parsing code base or testing with PhantomJS is your best bet.

[mgol]

This issue doesn't seem actionable now, it seems it requires further research. Should we move it to the roadmap?

[markelog]

Should we move it to the roadmap?

Done

[mgol]

Re-opening since it's now possible to do such a verification via a plain ESLint run.

[mgol]

PR: #4594