Repository navigation
Integrate upstream 12.1.0 prerelease into the Apple test version - #25
Closed
AlexSchmidt1999 wants to merge 8 commits into
Closed
AlexSchmidt1999 wants to merge 8 commits into
AlexSchmidt1999 wants to merge 8 commits into
Conversation
The caller list was written as an exhaustive set, so it goes stale the moment a gate joins. StressWidgetProducer's memo became the fifth. Refs ryanbr#2710
Route the validated WHOOP 5/MG event 109 reading into Apple live state so the Devices card matches Android without new BLE traffic. Clear stale readings on pack removal and disconnect.
The MG ECG probe could not observe a completed reading on any strap or firmware, and its report looked correct while it happened. @meta1971's field timing on ryanbr#891 is the evidence: across five MG sessions the terminal frame (classifier state 2, progress 100) landed 38 to 39 s after the first live frame, against a 30 s listen window held independently by both platforms. The verdict sheet appearing at 30 s made it worse than a missed deadline, since Stop is the next control a person reaches for. Second defect, same report: the candidate cap kept the FIRST twelve frames and R17 arrives about once a second, so the retained evidence was seconds 1 to 12 while result, average HR and the reason mask populate only on the terminal frame. The packet count stayed honest; the retained evidence did not. Whoop5EcgProbe now owns the window, the caps and the two decisions, with a declared twin on each side, so the numbers cannot drift apart again. A capture run listens 60 s, wrist-select and stop keep 30 s, a terminal frame is retained past the cap within a bounded reserve, and it re-arms the run for 10 s more rather than ending it: the variability field reads 0xffff for 2 to 9 s after that frame, so stopping on it would pin the unset value as the reading's own. A terminal frame also now ends a run EARLY, so the report quotes the time listened rather than the window asked for. The client wiring is BLE code and cannot be covered without a strap. The rules it calls are pure and pinned on both platforms, including the window against the measured 39 s, and a mutation back to the old behaviour kills the two cases that name the two defects on both sides. Tools/parity_twin_map.json carries the derived counts for the four new declarations (functions 4534 to 4538) so the authority keeps reproducing. Refs ryanbr#891
Found re-reviewing the window change, which is what created it. A stop re-arms the probe run, and re-arming cleared the candidate lines and the packet count along with the steps. At a 30 s window the verdict sheet had almost always rendered first, so little was lost. At 60 s it has not: the reading completes at 38 to 39 s while the sheet still reads "waiting", so the obvious next action is to stop a session the user can see has finished, and that discarded every R17 line including the terminal frame this change exists to retain. So a stop issued while the capture's window is still open carries the lines, the count and the run's START TIME. The start time travels with them because the report states "packets seen in Ns", and a count from a 47 s capture under a 30 s stop window would attribute those packets to a span that did not collect them. Only while the window is open. Once the verdict has rendered the lines are already out, and a later stop is a separate run. A fresh capture still clears, or a previous reading's frames would surface as the new one's, and wrist select still clears because its report is about a different command. Refs ryanbr#891
Second re-review pass, on my own first pass. `ecgProbeDeadline` is a MARKER: `ecgProbeArmed` tests it for nil and nothing in the file compares the stamp to the clock. So stamping it with `now + window` encoded the window a second time, where only the verdict timer's delay can act on it, and `ecgStartCapture` had to pass the same number twice for the code to read correctly. The window now lives in `scheduleEcgProbeVerdict` alone, which also renews the marker, and `beginEcgProbeRun` goes back to arming the run. Arming stays there rather than moving into the scheduler because the sends in between can draw a COMMAND_RESPONSE and the triage is gated on it. This is the shape Kotlin already had, where `ecgProbeListening` is a plain flag and the window exists only in the posted delay, so the two clients now make the same four window decisions in the same four places. Refs ryanbr#891
Marketing version moves with it, since this beta is of 12.1.0 rather than another 12.0.0 staging build. Refs ryanbr#891
Review the project hash for the prerelease version; capabilities and endpoints are unchanged.
Owner
Author
|
Closing this older fork test PR. Its full head ( |
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
5 participants
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.
Integrate the published upstream 12.1.0 prerelease (
testing-latest, pinned toeae23433) into the existing Apple test version, preserving the fork's assistant, training, widget, and performance work.The seven upstream commits include battery-pack charge display, ECG probe corrections, and the upstream Android stress-cache fix. Android files are imported unchanged as part of the authorized full sync; local builds and tests cover Apple only. Keep the fork's higher iOS build counter and review only the project-version hash in the source-security baseline.
Validation on the combined candidate: 4 ECG probe Swift tests; macOS app compilation and 44 focused app tests, including battery-pack routing, sync presentation, and training; iOS Release compilation including widget/watch targets. No ECG command was sent to a strap. Real battery-pack/ECG behavior remains untested on hardware. GitHub checks must be assessed separately from these local results.
This is a fork-only test integration. No upstream PR or main-branch promotion is requested.