The Arabic PDF bug was never in my code — it was the library version

I spent months writing bidi reconstruction logic for reversed Arabic text in PDF extraction. The actual fix was deleting all of it and upgrading two dependencies. Here is how to tell which one is lying to you.