Coverage remained the same at 63.804% when pulling ba5b1da on medikoo:fix-env-resolution into 4a6c85c on visionmedia:master.

I think your client-side bundler is misbehaving. In the package.json file we have "browser": "./src/browser.js" which bundlers like browserify and webpack use as the entry point by default. The index.js file is intended for Node.js and Electron processes to use.

I think your client-side bundler is misbehaving. In the package.json file we have "browser": "./src/browser.js"

Not all bundlers support browser tag, and personally I prefer to rely on cleaner (imo) approach as feature detection directly in a module (it doesn't require any custom handling from a bundler)

I'm happy to see that you do feature detection in debug, still it doesn't seem right. Do you see any harm in change I proposed? Does it break anything?

I'm happy to see that you do feature detection in debug, still it doesn't seem right. Do you see any harm in change I proposed? Does it break anything?

It seems ok over first glance. I am skeptical though because we've learned through history of this module that tweaks to these parts of code can break edge cases. I suppose we could merge though and wait for any complaints to pop up. I still think you should use a better bundler though 😉

I still think you should use a better bundler though 😉

Actually in this bundler it's more a feature than a bug :) Idea is to work with portable CJS modules pure, least custom way (best if no customizations involved). Think of it as strict mode for CJS ;-)

I suppose we could merge though and wait for any complaints to pop up

I'll be grateful. The issue could appear only if some would run it in env where typeof process === "undefined" but still want the require('./node.js') to be loaded. As that contradicts each other (./node.js uses internally process object) change seems 100% safe.