[Bug]: decodeScan - unexpected MCU data error for scanned pdf documents

[DongusJr]

Attach (recommended) or Link to PDF file

scanned-document-error.pdf

Web browser and its version

Chrome
Version 129.0.6668.58 (Official Build) (arm64)

Operating system and its version

MacOS - Sonoma 14.6.1

Is the bug present in the latest PDF.js version?

Yes

Is a browser extension

No

Steps to reproduce the problem

with https://mozilla.github.io/pdf.js/web/viewer.html open developer tools in your browser of choice (I used Chrome).
Open the attached document and notice that in the developer tools console we get the error
Warning: decodeScan - unexpected MCU data, current marker is: ffff around 600~ times

What is the expected behavior?

Not getting warning logs for decodeScan

What went wrong?

In the viewer.html, this does not cause any noticeable disruption to the document.
However, in a React application that uses pdf.js (in my case it was version 4.5.136 with legacy), this error sometimes causes extreme load times and in worst case scenarios, completely blank pdf or crashes in the browser.
This seems to happen more with larger documents.

I'll leave a React pdf component that reproduces this error (This code is just to give an idea how the pdf is rendered in our application):

import {
  getDocument,
  GlobalWorkerOptions,
} from 'pdfjs-dist/legacy/build/pdf.mjs';
import pdfWorker from 'pdfjs-dist/legacy/build/pdf.worker.mjs?url';
import React, { useEffect, useRef } from 'react';

type SimplePdfViewerProps = {
  contentBase64: string;
};

export const SimplePdfViewer = ({ contentBase64 }: SimplePdfViewerProps) => {
  const canvasRef = useRef<HTMLCanvasElement>(null);

  const base64ToByteArray = (byteData: string) => {
    const bytes = window.atob(byteData);
    const byteLength = bytes.length;
    const byteArray = new Uint8Array(new ArrayBuffer(byteLength));
    for (let i = 0; i < byteLength; i++) {
      byteArray[i] = bytes.charCodeAt(i);
    }
    return byteArray;
  };

  useEffect(() => {
    if (typeof window !== 'undefined' && 'Worker' in window) {
      GlobalWorkerOptions.workerSrc = new URL(
        pdfWorker,
        import.meta.url,
      ).toString();
    }

    const loadDocument = async (arr: Uint8Array) => {
      const pdf = await getDocument({
        data: arr,
      }).promise;

      const page = await pdf.getPage(1);

      const canvas = canvasRef.current;

      if (!canvas) return;

      const context = canvas.getContext('2d');

      if (context === null) return;

      const canvasViewport = page.getViewport({ scale: 1 });
      canvas.height = canvasViewport.height;
      canvas.width = canvasViewport.width;

      await page.render({
        canvasContext: context,
        viewport: canvasViewport,
      }).promise;
    };

    const typedArr = contentBase64 && base64ToByteArray(contentBase64);
    if (typedArr) {
      loadDocument(typedArr);
    }
  }, []);

  return (
    <canvas
      ref={canvasRef}
      style={{
        width: '100%',
      }}
    ></canvas>
  );
};

Additional context

I have encountered this issue with few documents, I am not sure if the problem lies in the pdf document it self or if this is a bug in pdf.js.

I have tried validating the pdf to check if it is corrupt but everything seems fine.

The warning is thrown in src/core/jpg.js, which might be because the pdf document is scanned.

Any idea what precisely causes this?

[calixteman]

You can change the verbosity level:

pdf.js/web/app_options.js

Lines 429 to 433 in 567df42

	verbosity: {
	/** @type {number} */
	value: 1,
	kind: OptionKind.API,
	},

Do you have a pdf you can share with us causing a crash or blank page ?
If you can't make it public you can send it to me on my pro email (I'm the triage owner of the pdf component: https://bugzilla.mozilla.org/enter_bug.cgi?product=Firefox&component=PDF%20Viewer).

[DongusJr]

I have a larger file here:
crash-error-document-larger.pdf

However, when testing this file I noticed that it was not crashing the browser in the test application that I made to reproduce this issue. However, it crashes the original solution that originally lead me to investigating this problem.
I will have to debug this, Ill add a comment if I figure out why it only crashes in the original application.

one more thing,
as I am not knowledgable about the inner workings of rendering pdfs/jpgs. What generally causes decodeScan - unexpected MCU data error? I have noticed that this warning always appears on documents that crash/don't render