segfault when getting the number of pages

[lsamper]

Pikepdf throws a segfault when getting the pages attribute directly

>>> import pikepdf
>>> pikepdf.Pdf.open("test_facture_template_word.pdf").pages
Segmentation fault (core dumped)

However, when using a temporary variable, it works

>>> import pikepdf
>>> pdf = pikepdf.Pdf.open("test_facture_template_word.pdf")
>>> pdf.pages
<pikepdf._core.PageList len=1>

It fails on the last version of pikepdf, with any PDF file

Steps to reproduce:

$ python3.8 -m venv ~/.virtualenvs/segfault_pikepdf
$ source ~/.virtualenvs/segfault_pikepdf/bin/activate
$ pip install pikepdf
Collecting pikepdf
  Downloading pikepdf-7.1.1-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (2.3 MB)
     |████████████████████████████████| 2.3 MB 2.7 MB/s 
Collecting lxml>=4.8
  Downloading lxml-4.9.2-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.manylinux_2_24_x86_64.whl (7.1 MB)
     |████████████████████████████████| 7.1 MB 11.3 MB/s 
Collecting deprecation
  Using cached deprecation-2.1.0-py2.py3-none-any.whl (11 kB)
Collecting packaging
  Using cached packaging-23.0-py3-none-any.whl (42 kB)
Collecting Pillow>=9.0
  Using cached Pillow-9.4.0-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (3.3 MB)
Installing collected packages: lxml, packaging, deprecation, Pillow, pikepdf
Successfully installed Pillow-9.4.0 deprecation-2.1.0 lxml-4.9.2 packaging-23.0 pikepdf-7.1.1
$ python
Python 3.8.10 (default, Nov 14 2022, 12:59:47) 
[GCC 9.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import pikepdf
>>> pikepdf.__version__
'7.1.1'
>>> pikepdf.__libqpdf_version__
'11.2.0'
>>> pikepdf.Pdf.open("Microsoft_invoice_2.pdf").pages
Segmentation fault (core dumped)

[lsamper]

This segfault occurs with versions 7.1.0 and 7.1.1

With version 7.0.0, the code doesn't work but throws an exception:

>>> pikepdf.Pdf.open("filename.pdf").pages
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
pikepdf._core.PdfError: attempted to dereference an uninitialized pikepdf.Object

With version 6.2.9, it works.

[mara004]

Is it possible that this has to do with the v7 change that dependent objects no longer keep their parent alive?

[jbarlow83]

Yes this is due to the v7 change. The expected behavior is now something like the 7.0 behavior. In any case, I won't have time to fix this for a while. Set pdf = pikepdf.Pdf.open(...) and all will be well.