l0ng-ai dbee51396f fix(agents): let #888's read marks decide a reattach this app already watched ...

With both fixes on main, the replayed status adopted as a baseline made
`status == last_agent_status` return before the read-mark lookup ran, so
every rebuild (workspace switch, tray reopen, both of which reattach)
skipped #888: an unread badge was dropped and a turn that finished while
away never badged.

A replay is now only a baseline when the pane has no mark, i.e. this app
never watched an agent in it (the restart case #890 is about). Marks are
kept for every status, not just Done, so a turn that was running when the
old view went still badges, and the restore path checks the mark was a
finished turn. Also adds `turns` to #890's unix-only remote.rs tests,
which no longer compiled against #888's field.

Claude-Session: https://claude.ai/code/session_01JRqYZ9E153WpSHGS2AW3BM

2026-09-15 23:06:53 +08:00

..

bin

fix(updater): read the designated requirement off the stream it is on (#708) (#717)

2026-08-25 16:16:08 +08:00

core

fix(update): read a 429 as a rate limit too

2026-09-10 18:20:48 +08:00

terminal

fix(agents): let #888's read marks decide a reattach this app already watched

2026-09-15 23:06:53 +08:00

ui

Merge pull request #888 from l0ng-ai/fix/870-hydrate-unread-badge

2026-09-15 23:02:25 +08:00

daemon.rs

chore: strip every comment from the Rust sources (#268)

2026-07-30 21:36:15 +08:00

main.rs

fix(macos): answer the Dock's reopen so a retired tty7 can come back

2026-09-14 00:18:07 +08:00