[r-kamei]

Summary

Fix #2932.

visitor_text() could receive a stale text matrix when a new text
fragment begins with a synthesized leading space.

Tests

• Added minimal reproducer PDF: visitor_text_position.pdf
• Added regression test:
  test_visitor_text_uses_current_text_matrix

Fails before the fix, passes after the fix.

AI Usage

Development assistance: Microsoft Copilot (GPT-5).

[r-kamei]

I noticed that the failing CI job is tests/test_reader.py::test_read_form_416.
That test downloads a PDF from the FDA website via get_data_from_url() and fails while constructing PdfReader
with:

PdfStreamError: Stream has ended unexpectedly

The failure appears to be outside the code path modified by this PR, which only affects visitor_text() text-matrix handling.

[stefan6419846]

Regarding #4062 (comment): In the past, this was a temporary flakiness of CI (like with some other files). As it relied on external URLs and thus can hopefully be made more reliable, I am currently changing this to use a local copy. After this, CI should hopefully pass again.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.11%. Comparing base (d80ccfc) to head (2c13aa0).
⚠️ Report is 3 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4062   +/-   ##
=======================================
  Coverage   98.10%   98.11%           
=======================================
  Files          59       59           
  Lines       11480    11476    -4     
  Branches     2148     2147    -1     
=======================================
- Hits        11263    11260    -3     
  Misses        122      122           
+ Partials       95       94    -1     

☔ View full report in Codecov by Harness.
📢 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.

[r-kamei]

The remaining failure is a timeout in test_flatedecode__decode_png_prediction__speed, which appears unrelated to the changes in this PR.