Bradford C. Smith

This does appear to be a bug in the closure-compiler parser. Since we use `goog.module()` / `goog.require()` instead of ES modules in Google, fixing this is likely to be a...

There's a check that runs early during compilation to ensure that all referenced variables are declared somewhere. However, when `RemoveUnusedCode` later runs and tries to find the declaration of `MediumPicker`...

I don't see any smoking guns there. I think we'll need a minimal repro in order to diagnose. I suggest you start by replacing the imported file with one that...

@robfig you could try passing the `--print_source_after_each_pass` option to the compiler. It might then become apparent which pass is removing the definition of MediumPicker, or else that none is and...

This does sound like the problem was a wrong entry point being provided to the compiler. I'm glad you figured it out.

OK, honestly, I'm not sure. I'm flying blind here in several respects. 1. I don't know exactly what options you're passing to the compiler. 1. I don't know exactly what...

Thanks a lot for the excellent repro. It took me less than 5 minutes to find the source of the problem with that in hand. Change `--entry_point corp.abc` to `--entry_point...

The best advice I can give with the current state of affairs is to explicitly put `@nocollapse` on every static method of a class if you intend to pass that...

By using `--print_source_after_each_pass` I found that after transpilation and inlining the compiler converts the code above to `new function(){}(1);` for ES5 output or `new class {constructor(){}}(1);`. Then the `peepholeOptimizations` pass...

@niloc132 I've just read through your draft change diff. My primary concern is that we had a design goal to stop depending on maven for building, but this seems to...