Using customHash has returned to prepending multiple times again

[typeoneerror]

(note: closed ember-cli/broccoli-asset-rewrite#43 in favor of this issue)

Updated to latest and assets are now getting "assets" three times when using customHash and prepend.

in:
<img src="assets/icons/stripe.svg">
out:
<img src="http://localhost:4200/assets/assets/assets/icons/stripe.svg">

  var fingerprintOptions = {
    enabled: true,
    extensions: ['js', 'css', 'png', 'jpg', 'gif', 'map', 'svg'],
    generateAssetMap: true,
    prepend: 'http://localhost:4200/',
    customHash: null,
    generateAssetMap: false
  };

Looks specific to customHash again (removing that fixes again) and may be related to #87

[typeoneerror]

FYI, the assetMap generated when generateAssetMap: true looks correct, so this may be an ember-cli issue.

[bichotll]

The same over here :/

[bichotll]

Hi @typeoneerror, any solution or at least workaround regarding this? ;)

I needed for a commercial project, I'm afraid :/

[typeoneerror]

@bichotll I didn't have any issues with production builds, it was only when I had customHash: null and generateAssetMap: false set in development build that it fails like this. So workaround is to just not use customHash at all locally and deal with the hash output on the dist files.

[rickharrison]

A failing test case would be really helpful in debugging this!

[grayt0r]

I ran into this issue and in the process of trying to create a failing test case realised, for me at least, that's it's actually caused by an open issue on broccoli-asset-rewrite: ember-cli/broccoli-asset-rewrite#39.

We reference the same relative image URL a few times in our CSS and each of those references appears to be rewritten multiple times, resulting in assets/ appearing in the path multiple times. This only happens when customHash: null is specified as if the file is fingerprinted then after the first rewrite the path will no longer match the fingerprint regex.

Happy to provide a failing test for the broccoli-asset-rewrite repo, but believe the test in the proposed PR to fix the issue linked above covers the same scenario.

[rickharrison]

Yes, that would be great if you had a failing test case over there.