| name | pdf-diff |
| description | Visual PDF regression test comparing current Beamer output against a baseline branch. Use when checking if slide changes introduced visual regressions. |
| argument-hint | [PaperName] [branch, default: main] |
| allowed-tools | ["Read","Bash","Glob"] |
Visual PDF Regression Test
Compare the current Beamer PDF output against a baseline branch to detect visual regressions.
Prerequisites
diff-pdf must be installed (brew install diff-pdf)- The paper's
.tex file must compile successfully
Steps
Step 1: Parse arguments
- First argument: PaperName (required, e.g.,
SUNY, DreamZero)
- Second argument: baseline branch (optional, default:
main)
Step 2: Compile current version
cd /Users/iyunseong/maangeek/paper2pr/Slides
TEXINPUTS=../Preambles:$TEXINPUTS xelatex -interaction=nonstopmode $PAPER.tex
cp $PAPER.pdf /tmp/${PAPER}_current.pdf
Step 3: Get baseline version
cd /Users/iyunseong/maangeek/paper2pr
git stash --include-untracked
git checkout $BRANCH -- Slides/$PAPER.tex
cd Slides
TEXINPUTS=../Preambles:$TEXINPUTS xelatex -interaction=nonstopmode $PAPER.tex
cp $PAPER.pdf /tmp/${PAPER}_baseline.pdf
git checkout - -- Slides/$PAPER.tex
git stash pop
Step 4: Run diff-pdf
diff-pdf --output-diff=/tmp/${PAPER}_diff.pdf /tmp/${PAPER}_baseline.pdf /tmp/${PAPER}_current.pdf
echo "Exit code: $?"
- Exit code 0: PDFs are visually identical
- Exit code 1: Visual differences detected (diff PDF saved)
Step 5: Report results
- If identical: "No visual regressions detected."
- If different: "Visual differences found. Diff PDF saved to /tmp/${PAPER}_diff.pdf"
- Open the diff PDF:
open /tmp/${PAPER}_diff.pdf
- Differences are highlighted in red
Step 6: Cleanup
rm -f /tmp/${PAPER}_current.pdf /tmp/${PAPER}_baseline.pdf
Keep the diff PDF for review if differences were found.
Notes
- This compares full visual output, including fonts, spacing, and figure rendering
- Single-pass XeLaTeX is sufficient for visual comparison (no bibtex needed)
- For multi-page diffs,
diff-pdf --skip-identical shows only changed pages
