[usmanmateen]

Issue resolved by this Pull Request:
Resolves #4695

The includegraphics branch in handlers/macros.py passed the opened figure straight to ImageRef.from_pil, which can't write CMYK as PNG. The except Exception turned that into a debug log, so a CMYK JPEG figure was added as a picture without an image. This normalizes the figure with normalize_image_for_png first, the same way the XLSX (#4664), OpenDocument (#4666), DOCX and PPTX backends do. RGB, RGBA and other PNG-compatible figures pass through unchanged.

Checklist:

• Documentation has been updated, if necessary. (not needed)
• Examples have been added, if necessary. (not needed)
• Tests have been added, if necessary.

Review evidence:

• The reproduction from LaTeX: CMYK JPEG figures are dropped #4695 prints ConversionStatus.SUCCESS [True, False] on main and ConversionStatus.SUCCESS [True, True] with this change.
• test_latex_includegraphics_cmyk_jpeg converts a .tex file with one CMYK JPEG figure and checks that the picture has an RGB image with the expected colour. It fails on main (picture.get_image(doc) is None) and passes here.
• pytest tests/test_latex: 134 passed.
• ruff format --check, ruff check and tach check pass. ty check reports no diagnostics on the changed lines.

Not changed: a str source (convert("paper.tex")) still gives no figure an image, because the backend gets a stream and can't resolve relative files. That is #4694, which I'll fix in a separate PR.

🤖 Generated with Claude Code

[github-actions]

✅ DCO Check Passed

Thanks @usmanmateen, all your commits are properly signed off. 🎉

[mergify]

Merge Protections

🟢 Merge protection satisfied — ready to merge.

Show 1 satisfied protection

🟢 Enforce conventional commit

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?:

[github-actions]

AI triage (advisory)

Checked commit 46d0517068. The analysis read the diff only. It did not run code from this pull request.

This pull request normalizes CMYK JPEG figures before image conversion in the LaTeX backend. This change prevents dropped figures during document export.

CI safety: no concern found. A maintainer must still approve CI.

1 related pull request(s) in the same code

PR	State	Signal	Assessment
#4697	open	also references issue #4694	related: Both PRs update macro processing in the LaTeX backend, but PR 4697 resolves relative file paths for string inputs.

_{The ai:* labels are advisory. They do not approve CI and they do not add tests:full. A new push runs this check again.}

[github-actions]

AI first review (advisory): no blocking issue found

This review covers the full diff at 762d5019fb. It read the code only. It did not run tests or code from this pull request.

The PR fixes CMYK JPEG figures being silently dropped by the LaTeX backend. It applies normalize_image_for_png before ImageRef.from_pil, matching the pattern already used in the DOCX, PPTX, XLSX, and ODF backends. A regression test constructs a synthetic CMYK JPEG, converts it through the backend, and checks that the resulting picture has an RGB image. The change is small and low-risk; the main review points are the test robustness and one minor redundancy.

_{A maintainer makes the merge decision. This review never approves the pull request.}

[github-actions]

AI first review (advisory): no blocking issue found

This review covers the changes since 762d5019fb at 46d0517068. It read the code only. It did not run tests or code from this pull request.

This PR fixes a bug where CMYK JPEG figures in LaTeX documents were silently dropped because ImageRef.from_pil cannot write CMYK as PNG. The fix applies normalize_image_for_png before the ImageRef.from_pil call, matching the pattern already used in XLSX, ODF, DOCX, and PPTX backends. The incremental update addresses both nit findings from the previous review: it replaces the redundant isinstance guard and loose colour-threshold assertions with a single exact equality check.

_{A maintainer makes the merge decision. This review never approves the pull request.}

[usmanmateen]

cc @PeterStaar-IBM @DanielNg0729: same normalize_image_for_png fix as #4664 / #4666, this time for LaTeX \includegraphics figures. Happy to adjust anything.

[DanielNg0729]

Hi @usmanmateen, thank you for the fix! I noticed there are a few other fixes addressing the same normalize_image_for_png issue. If there are any other cases where this issue still occurs, would you mind including those fixes in this PR as well? That way, we can hopefully address the issue comprehensively in one place. Thanks!

[usmanmateen]

Thanks @DanielNg0729! I went through every ImageRef.from_pil call on main (3181d9f) to look for other places with the same problem. With this PR, LaTeX was the last backend that passed a user-supplied image to from_pil without normalizing it. The others are covered like this:

Already normalized on main: XLSX (#4664), ODF pictures through _image_ref_from_odf_image (#4666), DOCX and PPTX (#4598).

Covered by open PRs:

• fix: keep pictures whose mode cannot be written as PNG #4683 (@ryanmeowy): ImageResourceLoader.load_image_ref and the HTML embedded-image path. The loader is shared, so this also covers Markdown, AsciiDoc, JATS and ODF linked images.
• fix(iwork): keep the transparency of pictures #4686 (@wfatih): iWork _add_picture, which your feat(iwork): read pictures placed on a Numbers sheet #4700 also uses for Numbers sheet pictures.
• fix(odf): keep ODS pictures anchored to the sheet #4678 (mine): ODS sheet-anchored pictures now go through _image_ref_from_odf_image, so they are normalized too.

Not affected, because the image is never in a mode PNG can't store: PDF, Tectonic and LibreOffice chart renders (pypdfium2 output), the HTML page screenshot and video frames (both .convert("RGB")), image input files (_to_rgb in image_backend.py), and the service client, which re-embeds images the server already wrote.

One case I left alone: METS/GBS page scans are opened without conversion, but they are page images rather than pictures, and the test archives are 1-bit TIFFs, which PNG stores as is. I didn't find a real METS/GBS source in another mode, so I didn't change it.

So this PR doesn't need more changes for now. If you'd rather have the #4683 / #4686 changes in one place, I'm happy to coordinate with their authors, but I didn't want to copy their work here.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[usmanmateen]

run-tests-core was cancelled by the 45-minute job timeout, not by a test failure. The runner spent about 37 minutes in the setup step (Preconfiguring packages), so pytest started at 17:31 and was stopped at 63% (17:37). Every test that ran passed, including the first LaTeX tests. check and ci-status only fail because of that job. Could a maintainer re-run the failed jobs? I can't re-run them from a fork. Thanks!

[PeterStaar-IBM]

lgtm!