UnescapeURLWithAdjustmentsImpl has rules based on characters belonging to certain unicode classes. Unfortunately, the code can't depend on icu currently, since net/ needs to work with and without icu enabled, so has to have its own character tables. We should look into resolving this - we can make a special version of the method that unescapes everything unconditionally for net/ consumers that want that behavior (data URLs), and see if those that need it for display can either be made not to use it (Like QUIC logging code), or be excluded from the binary for non-icu builds (Like FTP directory listings, which Cronet probably doesn't need).
Comment 1 by mmenke@chromium.org
, Apr 6 2018