Fatal crash calling page.getPixmap() with certain pdf file pages

[derwyddon]

Please provide all mandatory information!
In some pages of certain pdf files the getPixmap crashes without showing any nformatio.

You could use these files from the IMF:
https://www.imf.org/~/media/Websites/IMF/imported-flagship-issues/external/pubs/ft/weo/2013/01/pdf/_c1pdf.ashx
https://www.imf.org/~/media/Websites/IMF/imported-flagship-issues/external/pubs/ft/weo/2013/01/pdf/_c3pdf.ashx

to reproduce the crash

Describe the bug (mandatory)

The _c1pdf.pdf file crashes when processig page 1 and the _c3pdf.pdf file crashes processing page 2 (starting from page 0).

To Reproduce (mandatory)

Step 1) Open the file with fitz.open
Step 2) Using a bucle to call getPixmap for every doc Page
Step 2.a) The specific parameters used where mat = fitz.Matrix(2.0, 2.0) and alpha=False, but the getPixmap crashes even with the default parrameters: getPixmap() and not only with getPixmap(mat,alpha=False)

Expected behavior (optional)

Getting the pixmap or a Exception

Your configuration (mandatory)

• Windows 10 64bit v1809
• Python v3.6.3
• PyMuPDF v1.17.5

The output of print(sys.version, "\n", sys.platform, "\n", fitz.__doc__) :
3.6.3 (v3.6.3:2c5fed8, Oct 3 2017, 18:11:49) [MSC v.1900 64 bit (AMD64)]
win32

PyMuPDF 1.17.5: Python bindings for the MuPDF 1.17.0 library.
Version date: 2020-08-06 06:31:06.
Built for Python 3.6 on win32 (64-bit).

[JorjMcKie]

I don't know why you see no information, but those PDFs do contain errors, which also are revealed when subjecting them to e.g. mutool draw _c1pdf%d.png _c1pdf.pdf.
When I do page.getPixmap() under IDLE I see a lot errors, 2 or more per page around "recursive colorspace" definitions. When setting off ICC color management, fitz.TOOLS.set_icc(False), the situation is better and _c3pdf.pdf is at least processing each page - although not without still some errors.
Investigating further ...

[JorjMcKie]

Found the point where an issue is generated.
Is this urgent for you? In this case I can make a pre-vesion and upload a wheel to here.

[JorjMcKie]

what the hell - here is a pre-version:
PyMuPDF-1.17.6-cp36-cp36m-win_amd64.zip
Change the extension to "whl" and execute python -m pip install -U PyMuPDF-1.17.6-cp36-cp36m-win_amd64.whl

[derwyddon]

Thank you.
In order to continue working, I have moved the getPixmap version to another thread to be able to control the unstability. I am processing a set of pdf files (looking for the most heterogeneous pdf files you could found into the "jungle") and I have found some addtional files with the same behaviour (the "crash"), but I do not know if the problem comes from the same origin exactly. There are a lot of pd files that ara really bad built and contains a lot of different errors (as you mention).
Then, it is not urgent for me now, but I wil try the wheel and I will let you know if any of the files that I have found continues to have problems with the patch.

[JorjMcKie]

resolved in 1.17.6