Jabref Daily: Undo, Autosave, and the Cost of Shared State
Today's activity centers on a cluster of correctness fixes around per-library state — undo history, autosave, and startup labeling — all rooted in the same class of bug: state that should be per-library was being shared globally. A dense batch of UI polish and repo-tooling PRs rounds out the day.
Duration: PT2M27S
Episode overview
This episode is a short developer briefing from Jabref Daily.
It explains recent repository work in plain language.
- Show: Jabref Daily
- Published: 2026-09-06T04:00:34Z
- Audio duration: PT2M27S
Transcript excerpt
This excerpt keeps the crawler page concise. Listen to the episode or use the RSS feed for the full update.
Good morning. It's September 6th, and today's Jabref activity has a clear throughline: bugs caused by state that was shared across libraries when it should have been separate.
The headline is PR 16857 from calixtus, which gives each open library its own undo manager and journal. Previously, one shared undo stack meant undoing in one library could apply an edit to the wrong library entirely, and saving one library could mark a second, still-modified library as saved. That's a…
The same theme shows up in autosave. PR 16826 from koppor found that since an earlier change-filter refactor, the autosave manager had stopped registering for library changes — meaning autosave silently did nothing since that refactor landed. The fix also made the pending-save flag atomic, so a change arriving…
A second theme is parser correctness. Subhramit's PR 16851 fixes spurious warnings in the BST virtual machine when checking for a missing crossref field, and PR 16865 fixes the BibTeX parser so percent-sign comments don't leak into parsing logic. Both are small, targeted correctness fixes to how JabRef reads and…
On the UI side, several usability fixes landed: tooltips in the main table now only appear…
Rep…