Repository navigation
Spanish questionaire and test farm changes - #1183
Merged
Merged
Conversation
…ll screen Each minute's dumpsys meminfo reading now also stores the App Summary rows (native heap, graphics, Java heap, code, system, private other, stack), and the report shows them under the PSS total with each part's change over the current run of the app, so growth can be pinned on one part. bring_to_front made a pop-up task unresizeable only for as long as it took to bring it forward. The A54 remembers the task was a pop-up, so the dark mode switch that follows put it straight back into one and Restore ran in a small window. The task is now left unresizeable until the app next restarts. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ex6Fdp5ftuhE1ZEyhZb7WQ
A native speaker confirmed Detallado / Simplificado / Esencial / Silencioso: «Esencial» sounded formal at first, but fits a level that only gives the callouts that matter for a mental map, and its synonyms were worse. The questionnaire lists it under the decisions and now only asks "anything else". On the Spain-only place words (ES-N1), «Estacionamiento» is understood on both sides but formal in Spain, while many Latin American speakers don't know «Aparcamiento». We chose to keep Spain's words: they're clear in context and the search synonyms already accept both. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ex6Fdp5ftuhE1ZEyhZb7WQ
The A50's USB port can't keep up with the farm's load, so its battery drifts down. At battery_low_percent (10) a cycle now stops the app and turns the screen off, and waits until the battery is back to battery_resume_percent (80), logging recharge and recharged events. If it hasn't risen after an hour it logs one error and keeps waiting. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ex6Fdp5ftuhE1ZEyhZb7WQ
Maestro names its log folders by local time and keeps the newest few by name. When the clocks went back on 25 Oct, each new folder sorted as the oldest and was deleted as soon as it was made, so every Maestro run crashed opening its own lock file for an hour and Restore failures paused two phones. The farm now sets TZ=UTC before anything else, which Maestro inherits, and its own log, run folders and timestamps stay in order too. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ex6Fdp5ftuhE1ZEyhZb7WQ
The change column covered the whole run of the app, so a phone up for a day showed a day's drift and early start-up growth hid what's happening now. It now covers at most the last 4 hours of the current run, comparing the mean of the first and last 15 minutes of on-screen readings, since single readings swing by tens of MB from minute to minute. Runs shorter than 45 minutes show no change. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ex6Fdp5ftuhE1ZEyhZb7WQ
- The A54 could be left as a Samsung pop-up window after a monkey soak that Restore still passed; in the small window the home buttons are off screen, so every flow of the cycle failed at its first tap. The farm now takes the app out of a pop-up before every flow, not only when Restore fails. - The Moto could be left with the map full screen, which hides Home's search bar and which Back doesn't close. LaunchHome now closes it with the same button, and only presses Back while it still isn't on Home. - A route left playing hides Home's full screen map button, so Callouts failed; it now stops any route or beacon first. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ex6Fdp5ftuhE1ZEyhZb7WQ
…ow fixes - The monkey can delete the offline map from the app's Offline maps screen, after which Places Nearby is empty and nearly every flow fails until the phone is set up again. After each monkey soak the farm now checks for it and pushes it back; where the shell can't see the app's files (the Cubot), the app checks and downloads it again, which takes a minute when it's still there. - On a small screen Home scrolls and could be left scrolled down, with Places Nearby, Markers and Routes off the top; LaunchHome now scrolls back up. - DownloadExtract waited for the progress percentage to go, but on the Cubot it sat at 100.0% after a moment without it; it now waits for the map list. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ex6Fdp5ftuhE1ZEyhZb7WQ
… can't leak Turning Wi-Fi off wasn't enough. With scanning always available, which is on by default, the phone still scans for location, and Google's fused location matched the networks it had cached and reported the farm's real position, 68 km from the mock one, every few seconds. Soundscape uses fused location, so set-up's marker A, saved at the current location, landed there and the Markers and RouteCreation flows failed. The A52s, the one phone with scanning off, never had it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ex6Fdp5ftuhE1ZEyhZb7WQ
…tion can't leak Turning off Wi-Fi and Bluetooth scanning (3685387) didn't stop it: the fused provider kept mixing the farm's real position into the mock fixes on every phone with Play services. Location Accuracy is what feeds it, and the shell can't write that setting, so prepare() opens its settings page and taps the switch. With it off, no fix away from the mock point turned up in about 1,000 samples across a full set-up of the four phones. A single missed mock fix is routine on the Cubot and the Moto and the next lands a second later, so only log a run of five. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ex6Fdp5ftuhE1ZEyhZb7WQ
…t tap On the Cubot J10, PlacesNearby failed about one run in three, always on placesNearby-8, with the farm's screenshots on a place's Location Details. Screen recordings with the touch trail on showed two causes: - Just after the app starts, a Places Nearby folder opens empty and its places appear about 25 seconds later. Swipes and taps made meanwhile land on whichever place appears under them. WaitForScreen has a new PlacesList screen, the first place listed (its description has commas; a folder's is just its name), which the flows that open All Places now wait for. - Now and then a Maestro swipe reaches the Cubot as a press without its moves, which taps the row under it. SwipeAndTap takes an optional escape id: if that screen shows after a swipe, Back returns to the list and the scrolling carries on. PlacesNearby passes Location Details'. Without escape the check is skipped, so other callers don't pay for it. With both, PlacesNearby passed 10 runs in a row on the Cubot. RouteWalk also stops a route left playing first: a farm restart in the middle of its walk left Route Details showing Stop Route. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ex6Fdp5ftuhE1ZEyhZb7WQ
…onkey - The helper was a bare broadcast receiver, so its process only lived for a fix. The Cubot killed it after nearly every one (about 3,000 times in two hours) and the Moto often, so each fix forked a new process; on the Moto the helper hit an ANR, which made monkey abort a soak. A foreground service (KeepAlive) now holds the process, started with the app-op grant. - monkey runs with --ignore-timeouts, so another app's ANR can't cut a soak short. Soundscape's ANRs are still counted from the dropbox. - monkey runs on the phone, so stopping the farm left it going and its remaining events landed in the next flows after a restart. prepare() now stops it, and a soak cut short by the farm stopping isn't recorded as a failure. - record_flows in a phone's [devices] entry screen-records those flows and keeps the video of a failure with its logs. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Ex6Fdp5ftuhE1ZEyhZb7WQ
This branch was successfully deployed
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.