v4.1.27 — the duplicates popup stops naming deleted books, marking unread clears properly, editing a book no longer stalls the site
Latest
The duplicates popup stops naming books you already deleted, marking a book unread clears everything it should, and editing a book no longer stalls the site for everyone else.
docker pull ghcr.io/new-usemame/calibre-web-nextgen:v4.1.27
Fixed
The "Duplicates found" popup kept naming books you had already deleted. If you removed duplicate copies in Calibre itself rather than in the web app, the popup carried on listing them, while the Duplicates page and a fresh scan both correctly reported nothing. It was reading a list saved at the time of the last scan and never re-checking it against your actual library, so it was the one place still remembering the deleted books. It now re-checks before it shows: a group whose books are gone, archived, or hidden from you drops out, and a group that merely lost one copy shows the real remaining count instead of the stale one. The sidebar duplicate badge reads the same number, so it was wrong in the same way and is fixed too. Reported by @blahblah57 (#1167).
A duplicate you dismissed could come back on its own. Dismissals were matched against a label built from the title and author of whichever copy happened to sort first, so editing a book's metadata, or importing another copy, quietly changed the label and the dismissal stopped applying. Dismissals now hold onto a stable identity that metadata edits don't move.
Marking a book unread left its dates, and your e-reader's position, behind. The percentage cleared, but "Started reading" and "Last synced" stayed on the page, and "Last synced" jumped forward to the moment you pressed the button, so a book you had just marked unread looked like it had synced seconds ago. The reading position the device holds was also left in place, which meant a Kobo could quietly restore the exact spot you had just cleared on its next sync. Marking unread now clears both dates and the device resume point. Books already left in that state by an earlier version display correctly again, with no migration. Reported by @uschi1 (#627).
Saving a book, or switching on a metadata source, made the server stop answering everyone. Not just the tab doing the work: every other person's page load hung for as long as the cover download or the metadata lookup took, which on a slow cover host is up to 30 seconds. Both actions now do their network work off the request handler, so the rest of the site keeps responding while they run. Measured on a real server during a 1.5 second cover download, other page loads went from 1 request served with a 1,254 ms worst case to 201 served with an 18 ms worst case. Reported upstream at Calibre-Web-Automated#1111 and calibre-web#2963.
Changed
The admin Version Information table now reports the Calibre and Kepubify you are actually running. Both rows showed a value stamped into the image when it was built, so if the binaries had been replaced or the converter path pointed somewhere else, the numbers on the page were not the ones in use. The Statistics page already read the real binaries, so the two pages could show different versions on the same install with no way to tell which one was right. Both rows now read the binary itself, and one that can't be found or can't be run says which of the two it is instead of "Unknown". Contributed by @chloeroform (#1277).
Russian is complete again. Three phrases in the custom-columns section of the edit-metadata screen were still in English on an otherwise fully Russian page: the "Custom columns" heading, the "Not set" placeholder shown for an empty column, and the hint telling you a field takes comma-separated values. Contributed by @standhaftsohnsergius (#1269).
Something still broken? Open an issue — it gets read.
