Fix Read the Docs build by replacing images that serve WebP under a .png name

[pulorsok]

Problem

Read the Docs has not published since 2026-07-31. The site still serves that build, so Sova, EventBot and the v26.8.1 version bump are all missing from https://quark-engine.readthedocs.io/en/latest/malware_report.html

Every Sphinx builder succeeds. The failure is in the PDF step:

[rtd-command-info] ... exit-code: 12
latexmk -r latexmkrc -pdf -f -dvi- -ps- -jobname=quark-engine -interaction=nonstopmode

! error:  (file dna1NWm_xl.png) (readpng): internal error
libpng error: Not a PNG file
!  ==> Fatal error occurred, no output PDF file produced!

Because formats requests pdf, epub and htmlzip, a failure in any one of them marks the whole build failed, and Read the Docs does not publish the HTML that built cleanly.

Cause

The seven cdn.imgpile.com images in the DroidKungFu and GoldDream sections now return WebP data:

$ curl -sI https://cdn.imgpile.com/f/dna1NWm_xl.png | grep content-type
content-type: image/png

$ curl -sL https://cdn.imgpile.com/f/dna1NWm_xl.png | file -
/dev/stdin: RIFF (little-endian) data, Web/P image

The URL, the extension and the Content-Type header all still say PNG. Browsers sniff the real format and render them, which is why the HTML pages look correct. LuaLaTeX selects its decoder from the file name, so libpng receives WebP bytes and aborts.

Nothing in this repository changed. latex_engine = "lualatex" dates from #491 (2023-03), formats from #676, and these images from #857 (2026-01). They rendered into the PDF correctly for months; the image host converted its stored files to WebP some time between 2026-07-31 and 2026-08-11.

I checked all 103 image URLs in the document. Only these seven are affected — the images on postimg, catbox and ibb are all genuine PNG or JPEG.

Fix

Converted the seven images to real PNG files, re-hosted them alongside the other report images, and updated the URLs. Content is unchanged.

Verification

All four builders that Read the Docs runs pass locally, using the same dependency sequence as the RTD build:

Builder	Result
html	build succeeded, 2 warnings
singlehtml	build succeeded, 1 warning
latex	build succeeded, 1 warning
epub	build succeeded, 3 warnings

Every one of the 100 images the LaTeX builder downloads is now a valid PNG (previously the first one aborted the run).

Suggested follow-up

The document depends on four third-party image hosts. Any of them can repeat this, and a single bad image silently blocks HTML publishing. Dropping pdf from formats, or committing the report images into the repository, would remove that coupling. I left both out of this PR to keep it focused on the regression.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 78.48%. Comparing base (5edfa4f) to head (4235feb).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #962   +/-   ##
=======================================
  Coverage   78.48%   78.48%           
=======================================
  Files          84       84           
  Lines        7692     7692           
=======================================
  Hits         6037     6037           
  Misses       1655     1655           

Flag	Coverage Δ
unittests	78.48% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.