page.widgets() generator SegFault 11

[dgallitelli]

Please provide all mandatory information!

Describe the bug (mandatory)

While iterating on the widgets, obtained via the page.widgets() generator, a Segmentation Fault error is generated. When trying to access the second widget, after doing whatever operation is expected on the widget, it generates a SegFault.

To Reproduce (mandatory)

import fitz
doc = fitz.Document(<anypdf>)
for page in doc:
    for w in page.widgets():
        print(w.field_value)
        # Note if using (break) here, no Seg Fault is caused. 

Expected behavior (optional)

No Seg Fault

Screenshots (optional)

N/A

Your configuration (mandatory)

• MacOS 10.14.6
• Python 3.7.3, same thing on Python 3.8 and on AWS Lambda (using a Layer on both p37 and p38)
• PyMuPDF version: 1.17 from pip

For example, the output of print(sys.version, "\n", sys.platform, "\n", fitz.__doc__) would be sufficient (for the first two bullets).

Additional context (optional)

This behaviour does not happen on PyMuPDF 1.16.9, while still happens on 1.16.8

[JorjMcKie]

Thanks for reporting this.
Confirming the observation. It is a new error in v1.17.0 (worked prior to that). I reproduced it on several Windows and Linux versions.
Annotations seem to work.

[JorjMcKie]

Found and solved the problem. Was caused by dropping same memory multiple times ...
Will publish this with the next version. Please let me know if there is any urgency for you.

[dgallitelli]

Hi there, thanks for checking out the bug.
I can use a Lambda Layer with version 1.16.8 (pre-built by KLayers) with no problem for now, therefore there is no urgency.

I haven't tried Annotations though, what's the difference between those and Widgets? I don't think I can use them for reading filled forms.

[JorjMcKie]

I haven't tried Annotations though, what's the difference between those and Widgets? I don't think I can use them for reading filled forms.

No, annots are something functionally entirely different. I mentioned them, because they share most of the PDF specifications with annots, and also >80% of the code in (Py-) MuPDF.

[JorjMcKie]

Fixed by v1.17.1 which is currently being uploaded to PyPI.