Repository navigation
Conversation
- Persisted debug logs stay off until the user turns collection on. - Turning it off waits out an in-flight save, drops memory, then deletes the rows. A failed delete turns the toggle back on.
- The logger and buffer suites already cover the on path, so those checks added nothing. - German agreement and the Korean disable button now match the existing diagnostics wording.
Build package tests before starting the CI timeout. Add package Make targets, raw log capture, test selectors, and shared build flags with matching docs. Accept either offline-map destination state in navigation tests and combine the equivalent draft guards.
* fix: own each bluetooth connection's recovery Bind requests, links, sessions, and recovery tasks to their connection so stale work cannot publish ready, replace a session, or retire a newer link. Scope WiFi observations, identity import, and queued sends to the current connection. Preserve accepted DM acknowledgement deadlines across reconnects without sending the message again.
Put Change Device above Disconnect and restore the red Disconnect styling. Keep Stop Reconnecting first during recovery so it stays visible at large text sizes.
- Keep the connected radio's telemetry so Device Info can show the last reading after a reload fails. - Store that history with the radio and include it in backups, with a warning that GPS coordinates may be exported.
- Drop the Connection section because Device Info only appears for an already connected radio and the row always showed Connected. - Remove the strings that existed only for that section.
- Add DM retry settings with four initial attempts and four fallback attempts by default. - Use one DM retry budget and shared room send handling to keep retry behavior consistent. - Wait for command issuance and ACK tracking before admitting another DM or room post. - Persist quick replies for recovery and include retry preferences in backups.
- Clear temporary scope before the next command so it uses the radio default. - Wait for pending host replies before changing scope. - Replace the session if cleanup cannot be confirmed and keep accepted sends marked Sent.
- End the send pass when a later command times out or exhausts TABLE_FULL retries, and keep waiting for eligible ACKs. - Retry failed ACK owner writes through the ACK checker.
- Show 2/8 on the first retry and 8/8 on the last with the default limit. Use the same counts for VoiceOver.
* feat: add block/filter text rules for incoming messages - Filter channel, DM, and room messages using local rules. - Keep matches in Recently Blocked for review. - Handle missed room posts and interrupted radio cleanup.
- Avoid loading unrelated DMs when a sender repeats a timestamp. - Keep the exact UTF-8 check for replay.
- Keep the worker and continuation on one attempt so replies, timeouts, and cancellation can claim completion only once. - Test early and duplicate replies, replacement attempts, and cancellation during persistence.
- Store submission and wait state together so late send completion cannot replace a result. - Save the submitted path and hash size when the editor changes.
- Keep native picker ownership until dismissal so cancellation cannot admit a replacement picker. - Model pairing work and its single foreground retry with explicit phases.
- Keep one terminal writer so competing ACKs preserve delivery events and send admission. - Retain matched ACKs after failed saves and clean up owners installed during delivery. - Use the same settlement path for waiter and listener ACKs.
- Keep confirmed input and progress with each import. - Prevent old previews and success cleanup from changing new input.
- Replace paired guard flags with a typed claim to prevent inconsistent states. - Keep manual refresh inhibition independent of full sync. - Share gated contact fixtures and combine overlapping tests.
- Keep each destination's queue and active request under one owner. - Preserve FIFO order across retries and reject work after retirement.
- Keep Populate completion and Anchor consumption in ChatOpenAnchorState so completion cannot replay the Anchor. - Derive divider readiness from the current Bake to avoid positioning on cached rows from a previous open.
- Save reaction Messages and PendingSend rows together so reconnects retain queued work. - Persist badges after radio acceptance and keep reaction metadata consistent during deletion and backup.
- Share Message creation and enqueue failure handling between the composer and Siri. - Restore never-attempted queued DMs and increment their attempt counts in one save so recovery can retry persistence failures.
- Reset repeats before a new resend so early echoes remain recorded. - Keep repeats during recovery and reject mismatched echo timestamps.
- List each hearing so repeater and signal values can be compared. - Include repeater hashes and hop counts for VoiceOver.
- Attach each dialog to its button or menu so it opens from the correct control. - Preserve presentation ownership when the layout changes.
A reserved path-length byte, including 0xFF, was shown as a hop count taken from the lower 6 bits. decodePathLen already rejects that byte, so the hop label and the path map stay hidden when the byte does not decode.
A parked retry was sending attempt 0 again, and a row could become sent before the radio accepted the command. - Store the next attempt byte and continue a parked send from that byte. - Keep the new timestamp when the write goes out and no reply comes back, and also when the radio says its message table is full. - Put a Send Again back to delivered when the radio refuses the command, or when the write never leaves the phone. - Mark the row sent only after the radio accepts the command.
- The channel send now runs in sendQueuedChannel. - The channel gate returns whether the row is gone, the bump failed, the radio accepted the send, or the radio failed after a known attempt count. - The attempt-count lock and QueuedChannelPreflightFailure are gone.
- Direct Messages stores a flood attempt count of 0 when Fall Back to Flood is off, and shows Flood Attempts only while that switch is on. - Path Fallback keeps sign-in, repeater and room commands that expect a reply, and contact telemetry on the known route. - Flood Routing on the sign-in screen still clears that node's route, and the switch stays locked for the whole attempt and after the route is cleared. - A missing Path Fallback key stays on, and a backup stores an explicit false without writing the key when it was absent.
Use device localisation for temperature/distance/pressure units
A short real timeout can expire before a test injects its reply, which failed the package test job. - CLI, command, region, and lifecycle tests sleep on TestClock - The clock advances only when the test expects a timeout
- Signing, resync, remote admin, and sensor tests use TestClock - The app test clock reports the next sleep and can advance to it - A sensor reply advances only the short CLI poll
A stored message can be an MECP wire string. The bubble, the chat list, and notifications show the reading from the language packs. The stored text stays the wire string, so copy, reply, and send again still use it. The message details row shows that wire string. Settings acknowledgements include the CC BY 4.0 license for the packs.
The simulator gets a #emergency hashtag channel so you can see those readings without a radio. One row uses MECP/5/M01, which stays ordinary text.
Sensor labels, filter-rule labels, and radio cleanup were still the English source in every other locale. - Translate those strings for de, es, fr, it, ko, nl, pl, pt, ru, uk, and zh-Hans - Remove the TODO comments on the strings that now have a translation
- Inches of mercury keep two digits. Millibars and hectopascals use one digit, and millimeters of mercury use a whole number. - Tests now pass a fixed locale, so a UK reading stays in Celsius and a Liberian altitude stays in metres.
The chat and room bubbles draw MECPCatalog.reading when the stored text decodes. A reading has no translation row. Copy, reply, and send again keep the wire string.
- Rebuilding a private key clears the connected device before the replacement link is published, including while the connection is still connecting. - The open import page stays until the request publishes the replacement or disconnects. - A missing device with no rebuild still leaves device settings.
potential fix for idle macos crash
- The home-indicator spacer kept the window inset from when it attached, so a later inset left the latest message and the input bar slightly covered. - The chrome now writes that inset into the spacer, and the keyboard offset follows the height already laid out.
- A narrow Dynamic Island was showing only a down arrow and an ellipsis, so the compact rate is the digits alone. - Lock Screen and the expanded island stay on two rows, with the battery under the radio name and the rate labeled packets/min. - An unread count takes the top row and pushes a smaller rate to the row below.
- The Lock Screen puts the battery in the name column, so it no longer hides a copy of the antenna. - One packet-rate style replaces the overlapping flags, and a stale activity resolves to 0 before the labels draw it. - Compact draws one antenna. The minimal digits use the same size as the limited-width circle.
- A ZephCore backup counted the virtual contact toward the contact limit and could write it back on restore. - Export leaves it out of the file, and import skips the live key and the file identity's key.
- Live Activity text transitions slide the packets/min number on each update - The stale date now matches the packet window so the rate does not drop to 0 early
RunningBoard kills an iOS app on Mac when Metal's lock.mdb is held at suspend and the file has no suspend attribute. - Set the attribute at launch and on each scene change. - Flush the debug log directly. The background task around the flush ended immediately and did not release the lock.
- The pair button stayed on Connecting through a long sync and looked frozen. - It now follows the status pill, and the extra Connecting... string is gone.
- ZephCore answers a tagged command without echoing the tag, so the waiter timed out and the next scoped channel send reconnected. - A tag-less reply counts while attribution is still clean. After a timeout, an untagged reply is still ignored.
- A pairing that never finishes leaves the pair screen busy and blocks the next scan. - Cancel replaces Connect via WiFi while that attempt is busy, and skipping the device cancels it too.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.