Fix sources: ["0"], output the correct source

[xg-wang]

Case 1, sourcemap missing
In terser-js, sources are always 0 if old sourcemaps are not provided.

The value passed for sourceMap.url is only used to set //# sourceMappingURL=out.js.map in result.code. The value of filename is only used to set file attribute (see the spec) in source map file.

In broccoli-uglify-sourcemap we know in this case we are generating sourcemap for the file we are processing, changing 0 to the actual file gives us the correct source.

Case 2, multiple sourceMap comments
source-map-url only matches the very first magic comment, that let us mistakenly thinks the js file doesn't have a valid sourcemap.

Fix #130

[xg-wang]

These are 2 cases I met in some apps where sorucemaps have sources: [0], which makes reading the original file difficult.
There might be another issue from broccoli-concat where it leaves multiple magic comments in a js file.

Should supersede #58

[xg-wang]

Need review @stefanpenner @Turbo87

[xg-wang]

As suggested by @stefanpenner I created 2 benchmarks comparing

• regular regex test
• global regex test

The result on my mac:

  while ((matchedUrl = srcURL.getFrom(src)) !== null) {
    urls.push(matchedUrl);
    src = srcURL.removeFrom(src);
  }

run.js n=100 relativePath="ember.prod.js": 25.288791793887814
run.js n=100 relativePath="ember.prod.bad.js": 35.53592907140567

  while (match = srcRegExpg.exec(src)) {
    urls.push(match[1] || match[2] || '');
  }

run.js n=100 relativePath="ember.prod.js": 26.578133126260052
run.js n=100 relativePath="ember.prod.bad.js": 1,097.2937083164868

So add another commit here to use getFrom and removeFrom seems more reasonable.

[stefanpenner]

released as v3.1.1 🎉

[robclancy]

Just spent so long thinking I had configured something wrong trying to sort this out. And turns out a fix has been here all along :/