MinaChan1120 · 30 Jan 2026, 12:19 UTC
Time request, response arrival, display update and file-write completion separately, using an observation path that does not itself wait on the suspected file destination.
21 pointsDeveloper added those timings to the diagnostic replay. The response arrives promptly, then the display update waits behind the file write. It repeats when the replay writer is deliberately delayed.
7 pointsWhat does the screen show during that wait? If the last reading looks current, that needs fixing alongside the blocked write.
11 pointsAnd who needs the detailed log continuously? It may be useful, but the level you enabled for diagnosis need not become permanent by accident.
15 pointsWe had a screen that kept its comforting old number while the diagnostic folder stalled. People read it as live because nothing said otherwise. Put the reading's age where they look, not in the very log that's stuck.
22 pointsBruno, old value stays unchanged with no age. Aaron, maintenance wants normal summaries plus detailed logging when investigating. We've agreed those modes explicitly now.
23 pointsWhen they decouple writing, ask what happens at the buffer limit and on write failure; moving the stall off the display should not create unbounded memory use or invisible lost records.
21 pointsEmma's right. I would rather see 'record incomplete' than a responsive screen quietly throwing away the evidence I later need.
11 pointsThat can be separate from unavailable reading. Otherwise I'll chase the network when it's the log destination that's failed.
10 pointsDeveloper has bounded buffering with an explicit incomplete-record notice on overflow or write failure. Reading age is separate. Slow-writer replay keeps the display updating now, but we still need the failure checks.
23 pointsTry it with no successful read yet as well. An empty startup and a stale last value can make very different screens confusing.
15 pointsThe first-read failure says unavailable without a made-up previous value. Success followed by failure ages the retained reading. Buffer-overflow and write-failure cases each show record incomplete, and the replay identifies the missing interval.
21 pointsHas maintenance tried the installed revision from the usual shortcut?
14 pointsYes. Installed comparison and maintenance walkthrough passed, including the delayed-writer case. The normal shortcut uses the revised monitor. Thank you, Emma; separating the timings found the blockage without altering the robot connection.
19 pointsThis discussion is closed to new replies after six months without activity. Last activity: .