Broken XREF table incorrectly repaired

[guoqikai]

Description of the bug

page.rotation returns incorrect value

How to reproduce the bug

For this particular file:
ss.pdf
pdf = pymupdf.open(...) print(paf[0].rotation) -> 270
While the correct rotation should be 180.

PyMuPDF version

1.25.5

Operating system

MacOS

Python version

3.12

[JorjMcKie]

The value in the page definition is 270 and reported correctly as such:

[guoqikai]

The value in the page definition is 270 and reported correctly as such:

Hi, thanks for the response. Maybe the issue is not the rotation but something related to it. Let me clarify a bit:
I'm trying to get the image from the page object
pix: pymupdf.Pixmap = page.get_pixmap( # type: ignore colorspace="BGR", dpi=sampling_dpi, annots=False )

The result image is rotated 90 degrees

The page looks normal in a pdf viewer
Setting the rotation manually will fix the problem
page.set_rotation(180)

[JorjMcKie]

Well, pixmap creation of course respects the value of 270. Given a MediaBox with width 792 and height 612 plus a rotation of 270, The outcome of course must inevitably be a Pixmap with width 612 and height 792.

There is no way to do anything else. Garbage in - garbage out.

[guoqikai]

Well, pixmap creation of course respects the value of 270. Given a MediaBox with width 792 and height 612 plus a rotation of 270, The outcome of course must inevitably be a Pixmap with width 612 and height 792.

There is no way to do anything else. Garbage in - garbage out.

This pdf file can be displayed properly without the rotation on every pdf viewer though. Are there any other parameters that could cause the page rotation?

[guoqikai]

@JorjMcKie Also, when looking into this pdf file, I found there were multiple page objects that have the /Rotate 270 line

4 0 obj
<< /Type /Page
/Parent 2 0 R
/Resources 3 0 R
/MediaBox [0 0 792 612]
/Rotate 270
/Contents 5 0 R

endobj

13 0 obj
<< /Type /Page
/Parent 2 0 R
/Resources 12 0 R
/MediaBox [0 0 792 612]
/Rotate 270
/Contents 14 0 R

endobj

19 0 obj
<< /Type /Page
/Parent 2 0 R
/Resources 18 0 R
/MediaBox [0 0 792 612]
/Rotate 270
/Contents 20 0 R

25 0 obj
<< /Type /Page
/Parent 2 0 R
/Resources 24 0 R
/MediaBox [0 0 792 612]
/Rotate 270
/Contents 26 0 R

endobj

However when loop over the pages with pymupdf only the first page has rotation equal to 270

[guoqikai]

@JorjMcKie I've tested the file against other libraries and PyMuPdf is the only one that has this issue. Below is the code to reproduce it.

from pdf2image import convert_from_path
import pypdfium2 as pdfium
import pymupdf

i = 0
pages = convert_from_path("ss.pdf", 500)
page = pages[0]
page.save(f"pdf2image_output_{0}.jpg", "JPEG")


# Load a document
pdf = pdfium.PdfDocument("ss.pdf")

# Loop over pages and render
page = pdf[i]
image = page.render(scale=4).to_pil()
image.save(f"pypdfium2_output_{i}.jpg")




pdf = pymupdf.open("ss.pdf")
page = pdf[i]

pix: pymupdf.Pixmap = page.get_pixmap(  # type: ignore
    colorspace="BGR"
)

pix.save(f"pymupdf_output_{i}.jpg")

Only PyMuPdf will produce a portrait image while other output the image in landscape
Please reopen the issue

[JorjMcKie]

It is not PyMuPDF, but MuPDF which is responsible for handling this.
I will open a MuPDF bug report and see what the MuPDF team has to say.

[JorjMcKie]

MuPDF bug link:https://bugs.ghostscript.com/show_bug.cgi?id=708456