Fix PDF OCR import by switching to the PDFBox 3 Loader API

[Mercurlc]

Problem

Importing a PDF through the Tesseract OCR wizard fails at runtime with NoSuchMethodError on PDDocument.load(...).

The current implementation in PDFToImageRenderer still uses the PDFBox 2 API, while the runtime uses PDFBox 3, where PDDocument.load(...) was removed.

Root cause

application-ocr-tesseract-filter uses the old PDF loading API:

• PDDocument.load(...)

In PDFBox 3 this API was removed and replaced with:

• org.apache.pdfbox.Loader.loadPDF(...)

This causes PDF OCR import to fail before the OCR step starts.

Fix

This PR updates the PDF rendering code in PDFToImageRenderer to use the PDFBox 3 Loader API.

Changes included:

• replace PDDocument.load(...) with Loader.loadPDF(...)
• add an explicit pdfbox dependency in application-ocr-tesseract-filter
• exclude commons-logging to satisfy XWiki dependency rules

Validation

Tested on a local XWiki Docker setup with the Tesseract OCR Application installed.

Result:

• image OCR still works
• PDF OCR import now succeeds
• the previous NoSuchMethodError no longer occurs

Notes

This change is limited to the PDF loading/rendering path in the Tesseract filter module.

[tmortagne]

Note that this change would make impossible to install this extension in anything older than XWiki 17.10.1, while it's currently documented as supporting XWiki 8.4+.

I guess the cleanest would imply upgrading the parent version to 17.10 (which is based on 17.10.2) and, given the size of the jump, probably release a 2.0.0 version of the extension. If someone really need a for XWiki < 17.10, it's always possible to create a 1.x branch and continue release 1.1.x bugfix versions.

WDYT @aubincleme ?