ePubLift/epublift cli-v4.1.0 on GitHub

epublift repair and the upgrade run epubsana 0.30.0. New repairs:
- Every XHTML 1.x named entity is mapped, not 117 of the 248 (&emsp;
  and the rest of the Latin-1, symbol and special sets).
- A link nested inside another link in an EPUB 3 document is unwrapped,
  as it already was in EPUB 2.
- Spaces in content-document URLs are percent-encoded.
- An XHTML file declared as text/html is declared as
  application/xhtml+xml (needs a decision).
- A <script> whose file is not in the book is deleted, with the
  document's scripted property when nothing else needs it (needs a
  decision: the four shelf cases all name Kobo's kobo.js).
The upgrade relabels XHTML files declared as text/html with
epubsana's fix.html_media_type, applied without asking like the empty
<title> fix. EPUB 2 only warns about the label; EPUB 3 rejects it, so
every upgrade of such a book failed validation. The file must parse as
XHTML, or the label stays. epubveri never fully checked a file labelled
text/html, so the relabel can bring defects the book always had to
light: on the test shelf, 95 hidden errors in one book and 2 in
another. The deleted-script fix stays out of the
upgrade: it answers no EPUB 3 requirement, and repair asks for it.
Measured on our 562-book test shelf against cli-v4.0.0: upgraded books
that validate 293 → 301; after repair -y 378 → 398. Of 252 valid
books, 249 stay valid, as before.

The upgrade no longer adds errors of its own in four shapes found on
the test shelf:
- A package document that writes its namespace with a prefix
  (<ns0:package>, from some converters) gets its added meta and
  item elements with the same prefix; before, they fell outside the
  namespace and broke the package (6 of 562 books).
- A <guide> reference without a title becomes a landmark labelled with
  its type ("Cover") instead of an empty link; a table-of-contents entry
  with a blank label gets "Untitled", as a missing one already did
  (7 books).
- A <guide> that lists the same reference twice yields one landmark
  (1 book).
- Links in the generated nav.xhtml percent-encode spaces in file names
  (My%20Book_split_004.html); before, they were copied from toc.ncx
  as invalid URLs (4 books).

Full Changelog: cli-v4.0.0...cli-v4.1.0