ePubLift/epublift web-v1.20.0

on GitHub

Fixed

• Repair no longer deletes a book's description. The Repair mode dropped a
  dc:description whose text sat in child elements
  (<dc:description><p>…</p></dc:description>) as if it were empty: it read
  only the element's first text node, which there is just a line break. Three
  of the 544 test-shelf books lost a description of 800 to 1,400 characters this
  way, and epublift repair up to cli-v2.3.0 did the same. An element that holds
  markup is now never empty, and a no-break space counts as content, as epubcheck
  counts it. The Repair mode still runs this older repair; it moves to epubsana,
  which cli-v3.0.0 runs, in a later release.

Changed

• Validate mode runs epubveri 0.20.0 in the browser, the same validator
  as cli-v3.0.0: a book with an image whose header stops before its width and
  height now fails with PKG-021, as it does in epubcheck. On the 544-book test
  shelf no report changes, and the browser and epublift check give identical
  findings, in identical order, for every book.

Run it

docker pull ghcr.io/epublift/epublift-web:1.20.0

Multi-arch (amd64 + arm64). The web service ships as a container image
only; for the command-line tool, see the latest cli-v* release.