ROB: Handle /Pages node without /Kids during flattening

[gaoflow]

Summary

Reading a PDF whose page tree has a /Pages node typed /Type /Pages but without a /Kids entry (e.g. a malformed document advertising /Count 0 and no children) raises a bare KeyError: '/Kids' instead of being handled gracefully. This is the issue reported in #3811.

Cause

In _flatten, the node type is decided like this:

if PagesAttributes.TYPE in pages:
    t = cast(str, pages[PagesAttributes.TYPE])
elif PagesAttributes.KIDS not in pages:   # only reclassifies when /Type is absent
    t = "/Page"
else:
    t = "/Pages"

The existing fallback only treats a node as a single page when /Type is missing. When /Type is explicitly /Pages but /Kids is absent, the code still enters the /Pages branch and iterates pages[PagesAttributes.KIDS], which raises KeyError.

Fix

Treat a missing /Kids as an empty array, so a /Pages container with no children simply contributes no pages. len(reader.pages) then returns 0 for such a document instead of crashing. Documents that do have /Kids are unaffected (the key is still looked up exactly as before, preserving indirect-reference resolution).

Reproduction

from pypdf import PdfReader
# /Pages object: << /Type /Pages /Count 0 >>  (no /Kids)
reader = PdfReader("file.pdf")
print(len(reader.pages))   # before: KeyError: '/Kids'; after: 0

Tests

Added test_flatten__pages_without_kids, which removes /Kids from a real document's /Pages node, sets /Count 0, and asserts len(reader.pages) == 0. It fails with KeyError: '/Kids' on main and passes with this change. Existing multi-page documents still report the correct page counts.

Closes #3811

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.73%. Comparing base (52545c5) to head (5c04709).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3825   +/-   ##
=======================================
  Coverage   97.73%   97.73%           
=======================================
  Files          55       55           
  Lines       10417    10418    +1     
  Branches     1931     1931           
=======================================
+ Hits        10181    10182    +1     
  Misses        130      130           
  Partials      106      106           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[stefan6419846]

I do not think that this comment contributes enough to be useful here.

[stefan6419846]

This can be written in a simpler way:

Suggested change

	kids = (
	kids = pages.get(PagesAttributes.KIDS, ArrayObject())

[stefan6419846]

While we are at it (and although not required here), I would recommend replacing this cast and increase the resilience here.

What I mean is that we should use an empty ArrayObject if the kids are a NullObject and raise a proper exception if we see anything different from an ArrayObject for the iteration.