42. 【IEEE Paper】 Running a Control Systems LaTeX PoC to the End with CI | GitHub Actions

tags:LaTeX,IEEE,ControlSystems,AcademicWriting,GitHubActions,CI

📌 Position of This Article

This article is a continuation of the following series:

• 39: Fixing the LaTeX environment first on VS Code
• 40: Defining a minimal LaTeX structure to finish an IEEE Control Systems paper
• 41: Reaching the final PDF using that structure

In this article, as the final step, I document the following PoC:

“Can the same LaTeX structure be compiled to the end on CI (GitHub Actions)?”

🎯 What This Article Covers

This article does not discuss:

• Control theory
• Experimental results
• Research novelty

The goal is strictly this:

Can an IEEE Control Systems paper be compiled end-to-end with LaTeX on CI without structural failure?

🤔 Why Go as Far as GitHub Actions?

Even if a PDF compiles locally, many papers stop at the next stage:

• Environment-dependent failures
• BibTeX errors on CI
• No logs, making debugging impossible

In other words:

“It works on my machine, but is not reproducible.”

This PoC verifies that the same LaTeX structure works on:

• Local environments
• GitHub Actions (Ubuntu)

with no special handling.

🧱 Design Policy

This CI setup prioritizes the following:

• IEEEtran + XeLaTeX as a fixed assumption
• Always collecting logs, even on failure
• Uploading the final PDF as an artifact on success

The goal is to avoid the situation where CI fails silently.

🧩 Full GitHub Actions Workflow

Below is the workflow used in this PoC.

name: Build LaTeX PDF (Humanoid TCST)

on:
  push:
    paths:
      - 'papers/2025_humanoid_tcst/**'
      - '.github/workflows/latex_humanoid_tcst.yml'

jobs:
  build:
    runs-on: ubuntu-latest
    steps:
      - name: Checkout repository
        uses: actions/checkout@v4

      - name: Compile LaTeX document
        id: compile
        uses: xu-cheng/latex-action@v2
        with:
          root_file: main.tex
          working_directory: papers/2025_humanoid_tcst
          latexmk_use_xelatex: true
          continue_on_error: false

      # --- Always show logs ---
      - name: Show LaTeX logs (tail)
        if: always()
        run: |
          echo "==== list ===="
          ls -la papers/2025_humanoid_tcst || true
          echo "==== main.log (last 200 lines) ===="
          tail -n 200 papers/2025_humanoid_tcst/main.log || true
          echo "==== latexmk.log (last 200 lines) ===="
          tail -n 200 papers/2025_humanoid_tcst/latexmk.log || true

      - name: Upload PDF artifact
        if: success()
        uses: actions/upload-artifact@v4
        with:
          name: humanoid_tcst_pdf
          path: papers/2025_humanoid_tcst/main.pdf

      - name: Upload logs artifact
        if: always()
        uses: actions/upload-artifact@v4
        with:
          name: latex_logs
          path: |
            papers/2025_humanoid_tcst/*.log
            papers/2025_humanoid_tcst/*.aux
            papers/2025_humanoid_tcst/*.blg

🔍 Key Points of This Workflow

① Explicit XeLaTeX Usage

latexmk_use_xelatex: true

This fixes the assumptions for:

• IEEEtran
• newtx text/math fonts

directly on the CI side.

② Always Collect Logs

if: always()

ensures that:

• main.log
• latexmk.log

are always available.

Without logs, debugging LaTeX failures on CI is nearly impossible.

③ Upload the PDF as an Artifact

uses: actions/upload-artifact@v4

allows the following to be preserved:

• Final compiled PDF
• Build logs

This serves as proof that:

“This LaTeX structure successfully compiles on CI.”

🧠 What This PoC Demonstrated

• IEEEtran compiles cleanly on CI
• Appendix and Biography do not break the build
• LaTeX can be fully automated if the structure is correct

🧾 Series Summary (39–42)

• 39: Fixed the environment and workflow first
• 40: Defined a minimal LaTeX structure for completion
• 41: Reached the final PDF
• 42: Verified the same structure on CI (GitHub Actions)

Only after this foundation is stable does it make sense to refine the paper content.

🏁 Closing Remarks

Many papers stop not because of theory or experiments, but because:

“The structure is not reproducible.”

By fixing the LaTeX structure and running it on CI:

• Environment differences
• Reproducibility issues
• Individual dependency

can all be eliminated at once.

I hope this PoC helps anyone struggling to finish an IEEE Control Systems paper with LaTeX.

This article documents a LaTeX / CI proof-of-concept and does not aim to evaluate or claim research contributions.