v4.1.34 — PDFs open past page one on iPad, the reader learns to search and take notes, and reload-metadata asks first

PDFs open past the first page on an iPad, the reader learns to search inside a book and keep your notes, and a mis-click on "Reload metadata from disk" stops silently overwriting metadata you curated.

docker pull ghcr.io/new-usemame/calibre-web-nextgen:v4.1.34

Reading

PDFs open past page one on an iPad. In the new interface a PDF showed its first page and nothing else on iPadOS, in both Safari and Firefox, while the same book was fine on a Mac and on Android. The reader was handing the file to whatever PDF viewer the browser ships, and on iPhone and iPad that viewer only ever draws one page inside an embedded frame. PDFs now open in the viewer the classic interface has always used, which brings PDF text search, thumbnails and annotations to the new reader with it. Reported by @chloeroform (#1584).

You can search inside the book you are reading. Click the search button in the reader toolbar and type: results show the sentence around each match with your term marked, grouped by chapter, and clicking one takes you there without moving the place you were reading. Long books are searched a chapter at a time so the page stays responsive. Neither reader could do this before.

Your highlights and notes are listed inside the reader. Seeing what you had marked meant leaving the book for the Highlights page and losing your place. There is now a highlighter button in the top bar with a count, opening a list of every highlight with its note; picking one takes you to that passage. Highlights made on a Kobo or in KOReader are listed too, and now say which device they came from.

You can write notes while reading. Select a passage and the popup offers Add note alongside the colours, and tapping an existing highlight lets you add, edit or remove its note — highlights carrying one are drawn with a dashed outline. Write a note at the top of the panel opens a blank note for a thought about the book as a whole, which previously had to be pinned to a sentence that was not really the point.

Opening a highlight no longer loses your place, or marks the book read. Tapping a highlight jumped to that passage and then saved that as your reading position. The same save reports how far through you are, and the server treats 99% as finished — so glancing at a highlight near the end could mark the whole book read and pass that on to a connected Kobo or Hardcover account. Jumping to a highlight is now treated as looking, not reading.

The reader can go full screen, in black, in one column or two. Three things the classic reader had and the new one did not. Black is its own page theme again with a true black page — anyone who had chosen it was quietly given the warm dark theme instead — and a single-column preference is now honoured rather than saved and ignored. The five toolbar buttons went from 34 to 44 pixels, and typing in a reader panel no longer loses the cursor mid-word.

Library

"Reload metadata from disk" asks before it overwrites your edits. It sits in the same row as the download buttons on a book's page, so reaching for a download and landing one button over rewrote the title, author and series from whatever the file said — with no undo and no warning. It now asks first and names the book. Revoking an app password asks too: those buttons render as a column of identical trash icons, and a misclick cut off whichever device was still using it. Reported by @JamesHACS (#1496).

Kindle books get their high-resolution cover. The Amazon high-res lookup was keyed only on ISBN, and a Kindle edition usually has an ASIN instead — so the books most likely to need a better cover were the ones the lookup could never reach. A stored Amazon identifier now works as a key, which also covers 979-prefixed ISBNs that were a dead end. The cover picker also names the image's real source when a card's picture came from somewhere other than the provider that supplied the metadata. Reported by @briffaantoine (#304).

A metadata source that is throttling you now says so. Searching for the same book twice could find it on the first click and not on the second. Goodreads and bol.com answer a real no-match with an ordinary empty page, so a refusal was read as "this book does not exist" and reported as "No results for this query". Looking at a book's other editions also stopped sending a Hardcover id to every other provider, which made Goodreads in particular look broken. Reported by @briffaantoine (#303).

Sync and devices

Two new cover fill styles that fill the frame instead of adding a border. Every existing style pads the cover out to your device's shape, leaving a mirrored, blurred or coloured band down the sides. Stretch to fill scales the cover and accepts a little distortion; Crop to fill keeps the proportions honest and trims a strip off the two long edges. The six original styles are untouched and Edge mirror is still the default. Requested by @mgrimace (#1280).

In an OPDS reading app, Discover is called Discover. The entry was labelled "Random Books" while the sidebar, the new interface and the link itself all said Discover — and on a translated server that one entry stayed in English. It now reads Discover, translated, in all 28 languages. Reported by @chloeroform (#1097).

Admin

Send everyone to your single sign-on without giving up the password form. If you run exactly one OAuth provider, the login page can take people straight to it — but that jump was welded to "Disable Standard Login", so switching it on switched off password login for everyone, including you if the provider went down. They are now separate settings: turn on Start the only OAuth provider automatically under Admin → Security, and the password form stays reachable at /login?local=1 as a way back in. Off by default. Contributed by @lduesing (#1488).

Fixed elsewhere

Errors explain themselves instead of turning into a blank 500. Some failures replaced their own explanation with a crash and took down the page handling them. Uploading a book with an unwritable ingest folder was the clearest case: the app had the right sentence ready — check your /cwa-book-ingest volume permissions — but crashed while writing the log line, so nobody ever saw it and the upload returned a 500. The same fault sat on the reload-metadata and ingest-folder-creation paths. Reported by @Thovi98 (#1556).

Reporting a problem fills the report in for you, including the error itself when there is one. Nothing is sent by your library: the report is written in your browser and handed to you as a link, so you can edit or delete any of it first. Your address, library name, file paths and book titles are never included.

Smaller controls are easier to hit. The ☰ menu button on phones, Delete format, Revoke on an app password and the Kobo/OPDS shelf checkboxes keep their appearance but respond over a larger area. The notice bar across the top was one fixed dark-teal band whichever theme you picked, sitting on Light and Sepia as a near-black slab; it now comes in a version made for each theme.

The Highlights page admits it holds your notes too. A book with three highlights and two notes reported "5 highlights". It now reads Highlights and notes throughout, with counts that say what they are counting.

Translations

Russian went from 2,618 to 2,719 translated phrases, German from 1,891 to 2,071, and eight more settings now read in Dutch. Many of these were entries gettext had guessed from a similar English sentence and marked provisional — a provisional entry is dropped when the catalogue is compiled, so they had been showing in English while a translation sat in the file unused. One had guessed Import as "Wichtig:" ("Important:"). Contributed by @standhaftsohnsergius (#1554), @chaosblog (#1549) and @VHE1987 (#1553).

---

Something still broken? Open an issue at https://github.com/new-usemame/Calibre-Web-NextGen/issues — reports here get looked at.