Fix PDF spec compliance, git line-ending corruption, and add xref validation test

[Copilot]

The PDF renderer produced invalid files in two ways: the /Kids array embedded an inline page dictionary (violating the PDF spec, which requires indirect references), and writer.WriteLine() emitted platform-specific line endings causing xref byte offsets to be wrong when git normalized line endings on checkout.

PDF structure fix

/Kids must contain only indirect object references. Restructured from 3 objects to 4:

• Object 2 (Pages): /Kids [ 3 0 R ] — indirect reference instead of inline dict
• Object 3 (Page): new indirect Page object with /Contents 4 0 R
• Object 4 (Content stream): renumbered from 3→4; length now uses Encoding.ASCII.GetByteCount(content)

Cross-platform line endings

writer.WriteLine() emits Environment.NewLine, which is \n on Linux and \r\n on Windows. Since all other PDF output uses explicit \r\n, replaced with writer.Write("\r\n") for consistent output everywhere.

Git binary attribute

.gitattributes had * text=auto with .pdf only marked diff=astextplain — not binary. Git was stripping \r from PDF files on Linux checkout, making xref offsets wrong by 1 byte per newline. Changed to:

*.pdf  binary
*.PDF  binary

Approved snapshots

All 6 approved PDF snapshots regenerated. They now have consistent CRLF endings and correct xref offsets on all platforms.

Xref validation test

Added pdf_xref_table_is_valid to PdfByteQRCodeRendererTests — generates a PDF, locates startxref from the end, parses the full xref table, and asserts that every in-use entry's byte offset points to the correct N 0 obj line.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Fix PDF creation: /Kids must contain indirect references</issue_title>
<issue_description>Per CodeRabbit:

⚠️ Potential issue | 🔴 Critical

PDF Pages/Kids must reference an indirect Page object (not an inline dictionary).

Kids arrays shall contain indirect references to Page objects. The current inline page dictionary under /Kids violates the spec and may break validators/readers. Create a separate Page object and shift the content stream to the next object.

Apply this minimal restructuring:

-            "/Kids [ <<\r\n" +                                                      // Array of page objects - begin inline page dictionary
-                "/Type /Page\r\n" +                                                 // Declares this as a page
-                "/Parent 2 0 R\r\n" +                                               // References parent Pages object
-                "/MediaBox [0 0 " + pdfMediaSize + " " + pdfMediaSize + "]\r\n" +   // Page dimensions [x1 y1 x2 y2]
-                "/Resources << /ProcSet [ /PDF ] >>\r\n" +                          // Required resources: PDF operations only (no images)
-                "/Contents 3 0 R\r\n" +                                             // References content stream (object 3)
-                ">> ]\r\n" +                                                        // End inline page dictionary and Kids array
+            "/Kids [ 3 0 R ]\r\n" +                                                 // Kids must be indirect Page refs
             ">>\r\n" +
             "endobj\r\n"
         );
 
-        // Content stream - PDF drawing instructions
+        // Content stream - PDF drawing instructions
         var scale = ToStr(imgSize * 72 / (float)dpi / moduleCount);
         var pathCommands = CreatePathFromModules();
         var content = "q\r\n" +
           ...
           "Q";
 
-        writer.Flush();
-        xrefs.Add(stream.Position);
-
-        // Object 3: Content stream - contains the drawing instructions
-        writer.Write(
-            ToStr(xrefs.Count) + " 0 obj\r\n" +
-            "<< /Length " + ToStr(content.Length) + " >>\r\n" +
-            "stream\r\n" +
-            content + "endstream\r\n" +
-            "endobj\r\n"
-        );
+        // Object 3: Page (must be indirect)
+        writer.Flush();
+        xrefs.Add(stream.Position);
+        writer.Write(
+            ToStr(xrefs.Count) + " 0 obj\r\n" +
+            "<< /Type /Page\r\n" +
+            "/Parent 2 0 R\r\n" +
+            "/MediaBox [0 0 " + pdfMediaSize + " " + pdfMediaSize + "]\r\n" +
+            "/Resources << /ProcSet [ /PDF ] >>\r\n" +
+            "/Contents 4 0 R\r\n" +
+            ">>\r\n" +
+            "endobj\r\n"
+        );
+
+        // Object 4: Content stream - contains the drawing instructions
+        writer.Flush();
+        xrefs.Add(stream.Position);
+        writer.Write(
+            ToStr(xrefs.Count) + " 0 obj\r\n" +
+            "<< /Length " + ToStr(Encoding.ASCII.GetByteCount(content)) + " >>\r\n" +
+            "stream\r\n" +
+            content + "endstream\r\n" +
+            "endobj\r\n"
+        );

Also applies to: 117-124

🤖 Prompt for AI Agents
</issue_description>

<agent_instructions>Confirm this problem and resolution is correct and apply an appropriate fix if necessary. Do not necessarily use the suggested code unless it is correct. Ensure tests pass and/or fix them as necessary.</agent_instructions>

Comments on the Issue (you are @copilot in this section)

• Fixes Fix PDF creation: /Kids must contain indirect references #687

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[gfoidl]

👍🏻