Adrián Pérez de Castro authored Mar 15, 2017 and Michael Catanzaro committed Apr 10, 2017

In the logic to determine which filename to use when saving a non-HTML
resource the last component of the URI was being used as-is, which could
include percent-encoded characters. Calling soup_uri_decode() before
using the value ensures that we present readable file names to the user.

Also, this stops using g_file_get_basename() to extract the last
component of an URI because it would horribly break on platforms which
do not use a slash (“/”) as path separator.

This fixes bug #780086