Some browsers have issues with stylesheets that contain multiple `@charset` definitions. The issue surfaces while using Sass since it inserts a `@charset` at the top of each file. Then Sprockets concatenates them together.
The `CharsetNormalizer` processor strips out multiple `@charset` definitions.
The current implementation is naive. It picks the first `@charset` it sees and strips the others. This works for most people because the other definitions are usually `UTF-8`. A more sophisticated approach would be to re-encode stylesheets with mixed encodings.
This behavior can be disabled with:
environment.unregister_bundle_processor 'text/css', Sprockets::CharsetNormalizer
# File lib/sprockets/charset_normalizer.rb, line 25 def evaluate(context, locals, &block) charset = nil # Find and strip out any `@charset` definitions filtered_data = data.gsub(/^@charset "([^"]+)";$/) { charset ||= $1; "" } if charset # If there was a charset, move it to the top "@charset \"#{charset}\";#{filtered_data}" else data end end
# File lib/sprockets/charset_normalizer.rb, line 22 def prepare end