\chapter{Automated Manual Build}
\ChapterLabel{AutomatedManualBuild}
\index{Gitea Actions}\index{runner}\index{build artifact}\index{LaTeX}\index{continuous integration}

\begin{devopsNote}[Automated build verification]
The team demonstrated push-triggered PDF compilation, downloadable build
artifacts, and a controlled failure followed by recovery using a self-managed
Gitea Actions runner. The following sections record the observed run results
and traceable source revisions. Jenkins merge checks are covered separately
in Chapter 4.
\end{devopsNote}

\section{Source and trigger}
The team maintains the complete buildable LaTeX source, bibliography, figures,
and nonsecret build instructions in the Gitea repository:
\url{https://git.137-184-97-250.nip.io/stevens-smartpark/smartpark-app}.
The main document is \texttt{latex-manual/devopsAssignment.tex}; this chapter is
\texttt{latex-manual/devopsAutomatedManualBuild.tex}. The version-controlled
workflow is \texttt{.gitea/workflows/manual.yml}, and the build program is
\texttt{ci/build\_manual.py}.

The workflow uses the Git push event on \texttt{main},
\texttt{chapter3-build-automation}, and \texttt{chapter3-build-test}. The
actual demonstration pushes were made to \texttt{chapter3-build-test}.
Gitea Actions run 2 was triggered by a push from \texttt{vincentpenalosa}
and succeeded for commit
\texttt{9fec46a41ef1dcdffb40d0429e19475ba21af41c}:
\url{https://git.137-184-97-250.nip.io/stevens-smartpark/smartpark-app/actions/runs/2}.
Run 3 succeeded for the subsequent source commit
\texttt{e16a6deab277621f860c8a057a80c9737b5eb4b1}:
\url{https://git.137-184-97-250.nip.io/stevens-smartpark/smartpark-app/actions/runs/3}.
These runs demonstrate the configured source-push-to-workflow connection.

Git clients push over HTTPS to the public Gitea hostname on TCP port 443.
Caddy terminates TLS and forwards the request to Gitea on private container
port 3000. The runner VM reaches the public HTTPS Gitea hostname and does not
need a public inbound build, compiler, or management port. The runner checks
out source over the network and uploads the artifact through Gitea Actions;
no shared filesystem between the VM and Gitea is needed.

\begin{figure}[htbp]
\centering
\includegraphics[width=0.92\linewidth]{png/successful_run_1.png}
\caption{October 2026: push-triggered Gitea Actions run 2 on
\texttt{chapter3-build-test}, showing a successful job and one artifact.}
\label{fig:ch3-push-run2}
\end{figure}

\section{Runner setup}
The self-managed runner executes on Vincent's separate Ubuntu 22.04.5 LTS
VirtualBox VM (\texttt{x86\_64}), rather than on the shared DigitalOcean
Droplet hosting Gitea and Jenkins. The handoff records approximately 4 GB of
allocated memory and a 25 GB virtual disk; these are historical observations,
not current free-resource measurements. The installed Gitea runner version is
reported as 2.3.0. Its configured name is \texttt{vincent-ubuntu-manual},
with a repository-scoped registration for the SmartPark repository, label
\texttt{smartpark-latex:host}, and capacity one. The workflow selects the
label through \texttt{runs-on: smartpark-latex}.

The runner executes jobs in host mode, using TeX and Python utilities installed
on Ubuntu rather than launching Docker job containers. A dedicated unprivileged
\texttt{smartpark-runner} Linux account runs the service. The handoff identifies
\texttt{smartpark-runner.service}, configuration at
\texttt{/etc/smartpark-runner/config.yaml}, and a private runner identity at
\texttt{/var/lib/smartpark-runner/.runner}. A supplied service screenshot shows
the service active; successful jobs 2 and 3 additionally establish that the
runner executed workflow work. The VM must remain running and network-connected
to accept new jobs. No reboot-survival or current resource-availability test is
claimed without separate evidence.

\begin{figure}[htbp]
\centering
\includegraphics[width=0.92\linewidth]{png/runner_activated.png}
\caption{October 2026: Ubuntu service evidence for the separate
Gitea Actions runner VM. Service activity is not, by itself, proof of current
Gitea online status.}
\label{fig:ch3-runner-service}
\end{figure}

\section{Reproducible compilation}
On a clean Git checkout, the documented entry point is
\texttt{python3 ci/build\_manual.py}. This script copies the manual source to
a fresh temporary work directory and compiles the main file with
\texttt{latexmk -pdf}, stopping on LaTeX errors. It supports the document's
\texttt{minted} highlighting with Pygments and shell escape, and coordinates
bibliography, glossary, index, and repeated LaTeX passes. Source and workflow
edit rights must be restricted because enabling shell escape executes helper
programs. The expected deliverables are
\texttt{build/manual/SmartPark-Manual.pdf} and
\texttt{build/manual/build-info.txt}.

The supplied successful artifact metadata records \texttt{latexmk} 4.76,
\texttt{pdfTeX} 1.40.22 (TeX Live 2022/dev/Debian), and Pygments 2.11.2.
The handoff also reports \texttt{makeglossaries} 4.49 and packages including
\texttt{texlive-latex-extra}, \texttt{texlive-fonts-recommended},
\texttt{lmodern}, and \texttt{python3-pygments}. The build script removes old
deliverables, checks command exit status and the expected PDF, and treats
unresolved reference/citation summaries as errors. Therefore an old PDF
cannot silently stand in for a newly successful build. The compiler runs on
the runner VM, not in the Gitea application container.

Run 2's successful \texttt{manual} job took 41 seconds; run 3 also completed
successfully. Their run pages and downloadable \texttt{build-info.txt} files
provide the logged build context:
\url{https://git.137-184-97-250.nip.io/stevens-smartpark/smartpark-app/actions/runs/2}
and
\url{https://git.137-184-97-250.nip.io/stevens-smartpark/smartpark-app/actions/runs/3}.
Run 4 is the controlled negative compilation test, with a failed workflow
status. The run 4 screenshot records the negative test status; the job console
is the authoritative source for the compiler diagnostic.

\section{PDF version traceability}
The team maintains the human-readable manual version separately from the
machine-generated build identifier. The documented manual version for these
runs is \texttt{1.9-draft}. For Gitea Actions builds, the identification page
adds the triggering commit's short revision and the workflow run number.
The metadata is generated during the temporary build and is not pushed back
to the branch, avoiding a self-triggering push/build loop. Local builds use
\texttt{local}; local modified trees can display \texttt{dirty}, so a local
build is not substituted for Actions evidence.

The downloaded PDF for run 2 visibly identifies
\texttt{1.9-draft+build.2.9fec46a41ef1}, source revision
\texttt{9fec46a41ef1}, and workflow run \texttt{2}. The downloaded PDF for
run 3 displays \texttt{1.9-draft+build.3.e16a6deab277}, source revision
\texttt{e16a6deab277}, and workflow run \texttt{3}. The matching full
commits are recorded in Section 6. The respective build metadata timestamps
are October 10, 2026 at 02:12 UTC and 02:26 UTC. The identifiers printed in the respective downloaded PDFs match the
workflow run numbers and source revisions.

\begin{figure}[htbp]
\centering
\includegraphics[width=0.90\linewidth]{png/successful_run_2.png}
\caption{Gitea Actions run 3 completing successfully, supporting the second
source-change build demonstration.}
\label{fig:ch3-run3-success}
\end{figure}

\section{Downloadable artifact}
The workflow names artifacts using the workflow run and triggering source
commit. Run 2 uploaded a named downloadable artifact:
\url{https://git.137-184-97-250.nip.io/stevens-smartpark/smartpark-app/actions/runs/2}.
Its downloaded ZIP contains \texttt{SmartPark-Manual.pdf} and
\texttt{build-info.txt}. The 82-page PDF was extracted and opened
successfully. Its build identifier is
\texttt{1.9-draft+build.2.9fec46a41ef1}, matching run 2 and commit
\texttt{9fec46a41ef1dcdffb40d0429e19475ba21af41c}. Run 3 also
produced a distinct successful PDF and matching metadata, rather than
reusing the previous output.

A team member obtains an artifact by opening the matching Gitea Actions run,
choosing the named entry under Artifacts, downloading and extracting the ZIP,
and visually inspecting the PDF and its identification page. The workflow
requests a 30-day artifact retention period. The server's actually applied
retention policy was not independently verified. Artifacts are downloadable
build deliverables, not permanent backups.

\begin{figure}[htbp]
\centering
\includegraphics[width=0.92\linewidth]{png/successful_run_1.png}
\caption{October 2026: successful run 2 showing the named downloadable
manual artifact in the Gitea Actions interface.}
\label{fig:ch3-artifact-run2}
\end{figure}

\section{Change, failure, and repair evidence}
The team performed the following source-change and repair sequence on
\texttt{chapter3-build-test}. Expected results were compared with the
actual Gitea Actions status, with links to each run.

\begin{enumerate}
\item \textbf{First successful source change:} run 2, ``Document first manual
build demonstration,'' compiled commit
\texttt{9fec46a41ef1dcdffb40d0429e19475ba21af41c}, generated the PDF
identifier \texttt{1.9-draft+build.2.9fec46a41ef1}, and uploaded one named
artifact. Expected: build and artifact succeed; actual: success.
\url{https://git.137-184-97-250.nip.io/stevens-smartpark/smartpark-app/actions/runs/2}.

\item \textbf{Second successful source change:} run 3, ``Document second
manual build demonstration,'' compiled the next source change at commit
\texttt{e16a6deab277621f860c8a057a80c9737b5eb4b1}. The downloaded PDF
identifies \texttt{1.9-draft+build.3.e16a6deab277}. Expected: a different
build/PDF tied to this source revision; actual: successful run and matching
PDF metadata.
\url{https://git.137-184-97-250.nip.io/stevens-smartpark/smartpark-app/actions/runs/3}.

\item \textbf{Controlled compiler failure:} run 4, ``Demonstrate controlled
LaTeX compilation failure,'' used the deliberate undefined command
\verb|\SmartParkIntentionalBuildFailure|. Expected: nonzero LaTeX build status
and no successfully published PDF. The supplied screenshot confirms failed
workflow status and shows no artifact in that view. The run URL links to the associated commit and compiler job log.
\url{https://git.137-184-97-250.nip.io/stevens-smartpark/smartpark-app/actions/runs/4}.

\item \textbf{Repair:} run 5, ``Revert Demonstrate controlled LaTeX compilation
failure,'' restored a working document. Expected: successful compile and
artifact upload; actual: screenshot shows a successful job and one artifact.
The run page identifies the repair revision and downloadable artifact.
\url{https://git.137-184-97-250.nip.io/stevens-smartpark/smartpark-app/actions/runs/5}.
\end{enumerate}

\begin{figure}[htbp]
\centering
\includegraphics[width=0.88\linewidth]{png/latex_compilation_failure.png}
\caption{October 2026: controlled LaTeX compilation failure recorded
by Gitea Actions run 4. The matching job log records the compiler failure.}
\label{fig:ch3-run4-failure}
\end{figure}
\begin{figure}[htbp]
\centering
\includegraphics[width=0.88\linewidth]{png/repaired_build.png}
\caption{October 2026: Gitea Actions run 5 succeeding after the
controlled error was reverted, with one downloadable artifact.}
\label{fig:ch3-run5-repair}
\end{figure}

\section{Runner and secret boundary}
The Chapter 3 runner uses Vincent's separate Ubuntu VirtualBox VM, not the
shared DigitalOcean Droplet. Its configured repository-scoped registration,
\texttt{smartpark-latex:host} execution label, capacity one, and unprivileged
\texttt{smartpark-runner} service account limit the jobs' intended access.
Host-mode jobs execute directly as that account and are \emph{not} isolated
from each other by disposable Docker containers. Consequently, workflow and
source changes targeting the runner should be restricted to reviewed team
contributors. A push-only trigger and label do not themselves establish an
access-control boundary.

The service configuration resides at
\texttt{/etc/smartpark-runner/config.yaml}; the private registration identity
resides at \texttt{/var/lib/smartpark-runner/.runner}. Runner registration
tokens, personal SSH keys, passwords and repository secrets must not appear
in Git, job logs, screenshots, or generated PDFs. The operational Passwords
table should contain only ownership and storage hints. No privileged Docker
socket access was granted to the Chapter 3 job environment on the shared
Droplet; doing so would expose the other Docker-hosted Gitea/Jenkins services
to substantial risk. Only outbound HTTPS access to Gitea is required.

The actual editor/branch permissions should be verified against repository
settings; this report does not claim that every policy was technically
enforced just because it is recommended. The service evidence is shown in
Figure~\ref{fig:ch3-runner-service}; nonsecret configuration lives under
\texttt{ci/} in the Gitea repository.

\section{Manual quality and controls}
The version-controlled source, reproducible workflow, host-mode runner,
compiler checks, visible build identifiers, successful PDF artifacts, and
controlled failure/repair are documented above. The manual version used for
the demonstrated builds is \texttt{1.9-draft}. The final released manual version is maintained by the team separately
from the build identifiers recorded for historical runs.

For operational records, the Hosts table identifies the Ubuntu runner VM,
its owner, role, outbound HTTPS access, and availability conditions. The
Passwords table tracks only the runner credential custodian and storage hint,
not the token or private runner identity. The team tracks Chapter 3 changes
in the manual history and retains source, build, and artifact links for
auditability. Index terms include runner, Gitea Actions, build artifact,
and reproducible compilation.

\textbf{Troubleshooting:} For queued jobs, check VM power/network access,
runner service status, repository scope, label matching, and capacity. For
checkout failures, check HTTPS/DNS and repository authorization. For compiler
failures, read the first actual TeX error, check paths and dependencies, and
rerun after a committed repair. For missing artifacts, check that
\texttt{build/manual/SmartPark-Manual.pdf} exists and inspect the upload step.
A green compiler step alone does not establish successful artifact upload.
For failure/repair evidence, keep the exact commit, run link, timestamp, and
outcome synchronized across the manual and screenshots.

\textbf{Final submission check:} Commit the integrated LaTeX and PNG files,
verify a new successful Gitea Actions build, download its PDF artifact, open
and inspect every figure and build identifier, and submit that downloaded PDF.
The submitted deliverable is the PDF downloaded from the final successful
Gitea Actions run.