appending a page from a pdf with non terminal acroform fields results in multiple page objects in the result

[typetetris]

Using

hierarchical.pdf
having two pages with acroform fields, which are grouped under a single non terminal
acroform field

in

Python 3.13.5 (main, Jun 11 2025, 15:36:57) [Clang 19.1.7 ] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import pypdf
>>> writer = pypdf.PdfWriter()
>>> writer.append('hierarchical.pdf', pages=[0])
>>> with open("result.pdf", "wb") as output:
...     writer.write(output)
...
(False, <_io.BufferedWriter name='result.pdf'>)
>>> quit()

and looking into result.pdf, one can see, that there are multiple page objects and
the old page tree root is included, but not referenced anywhere.

My conjecture would be:

I suspect that the acroform fields for the page not appended have not been removed
because of the common non terminal parent. This might have lead to the widgets annotations
of the "to be removed" page have not been removed and therefore the page being kept,
which also leads to keeping the old page tree root, as it is referenced from the page.

Environment

Which environment were you using when you encountered the problem?

$ python -m platform
# macOS-26.4-arm64-arm-64bit-Mach-O

$ python -c "import pypdf;print(pypdf._debug_versions)"
# pypdf==6.10.1, crypt_provider=('cryptography', '45.0.4'), PIL=11.3.0

Code + PDF

hierarchical.pdf

Python 3.13.5 (main, Jun 11 2025, 15:36:57) [Clang 19.1.7 ] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import pypdf
>>> writer = pypdf.PdfWriter()
>>> writer.append('hierarchical.pdf', pages=[0])
>>> with open("result.pdf", "wb") as output:
...     writer.write(output)
...
(False, <_io.BufferedWriter name='result.pdf'>)
>>> quit()

Share here the PDF file(s) that cause the issue. The smaller they are, the
better. Let us know if we may add them to our tests!

My test pdf
hierarchical.pdf
was derived from https://github.com/mozilla/pdf.js/blob/9159afd633bf02ebcb38766d75b508c0d8ee7c39/test/pdfs/form_two_pages.pdf and therefore probably has an Apache-2.0 license on it.

Traceback

None

[stefan6419846]

Thanks for the report.

This seems to be some weird constellation with the page defined twice and referenced from two different /Pages objects. As this does not seem to cause any further direct issues apart from the weird constellation, I have not looked further into where this goes wrong.

You are of course invited to further investigate this issue and propose a corresponding PR.

[typetetris]

I think it might be an issue for some users, that if you append a single page, you get two in the result, even if one isn't visible in a viewer.

The contents would still be recoverable for people with technical knowledge.

I don't know how widespread the use of hierarchical acroform fields is.

Will see, If I can look further into it, depends on RL.