[Bug]: The content stream being parsed is incorrect. Instead of obtaining the updated version of the pdf increment, the old version was taken.

[klwjlalala]

Attach (recommended) or Link to PDF file

多分辨率的自动MRI图像配准.pdf
This pdf has an increentially updated xref table, but it does not delete the old obj objects, resulting in two OBJs with the same serial number (such as 60obj) inside. According to the pdf specification, the latter obj should be taken, but in reality, only the first one is taken.

Web browser and its version

Firefox version 142

Operating system and its version

windows 11

PDF.js version

The latest version of the master branch

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

Yes

Is a browser extension

No

Steps to reproduce the problem

After loading the pdf, view the objects on each page and then compare them with the objects in the pdf bytes.

What is the expected behavior?

Consistent with the underlying pdf, after the incremental update of the first page, five contents objects should be associated. However, what is obtained now is that the old version only associated one object.

What went wrong?

The fault-tolerant mechanism for non-standard PDFS is insufficient, and the PDFbox library in java can parse them correctly.

Link to a viewer

No response

Additional context

No response

[calixteman]

Could give us the object number which is not correct ?
You're talking about 60obj, I checked the object 60 0 R which is a ref on 112 which is the length of the stream pointed by 59 0 R and I get the exact same thing from the PDFBox debugger.
Anyway if there's a difference in rendering with an other software, please provide some screenshots. If there are some wrong data somewhere please give us precise information in order to quickly identify what's wrong. Here you just mentioned a piece of code but we don't know what we should looking for.

[klwjlalala]

There are two 6 0 obj in the pdf I provided. One is the initial one, and only [7 0 R] is in the contents. The second one is the incremental one, and [55 0 R 7 0 R 57 0 R 59 0 R 63 0 R] is in the contents. Because the position given by xref for the incremental part is incorrect and the format is rather chaotic, the incremental part was not parsed. Instead, the first page object with contents as [7 0 R] was taken for rendering.