richard authored Jan 18, 2018 and Georg Koppen committed Feb 21, 2018

After much debugging and investigation, it seems that the required
information needed to drive the first-party domain cannot be accessed in
the XmlHttpRequest creation path.  The JS context the part of pdf.js making
the range requests runs with does not have a reference to parent window and
associated LoadInfo information (which includes the requesting first-party
domain).

To fix the issue, we can easily disable support for range-based requests
via the pdfjs.disableRange property.  However, the side-effect here is
that pages can not be read as they load; the entire pdf must be
downloaded before it can be read and interacted with.

This patch updates each platforms extension-overrides.js to change this
pref.