PdfReader.Open throws "Page has no MediaBox" for pages that inherit /MediaBox from an ancestor `/Pages` node

[professor-k]

Summary

PdfReader.Open fails on any PDF whose page(s) do not carry a direct /MediaBox but inherit it from a parent /Pages node. /MediaBox is an inheritable page attribute (PDF 32000-1:2008, §7.7.3.4, Table 31), so this is a valid, common construction — e.g. every PDF produced by ComponentOne (GrapeCity) C1.C1Preview PdfExportProvider is laid out this way.

The page object is constructed (which reads/validates /MediaBox) before the inherited attributes are applied, so the lookup fails.

Version

• Broken: PDFsharp 7.0.0-preview-1 (NuGet), net8.0. The same code is present on main (HEAD).
• Works: PDFsharp 6.2.4 and 1.50.5147 — both open the same file as a 612×792 page.

So this is a regression introduced by the 7.0 page-tree rewrite; the inherited-attribute handling that the 6.x reader did correctly was

Repro

Attached: c1-inherited-mediabox.pdf — a trivial one-page document produced by ComponentOne PdfExportProvider. Its page tree is:

10 0 obj <</Type /Pages /Count 1 /MediaBox [0 0 612 792] /Resources 11 0 R /Kids [12 0 R]>>
12 0 obj <</Type /Page /Parent 10 0 R /Contents 2 0 R /Resources 11 0 R>>

/MediaBox is on the /Pages node (obj 10) and inherited by the page (obj 12), which has none of its own.

using var fs = File.OpenRead("c1-inherited-mediabox.pdf");
var doc = PdfReader.Open(fs, PdfDocumentOpenMode.Import); // throws

Expected

The page opens with size 612×792, resolving /MediaBox via inheritance (as 1.50 and other readers do).

Actual

System.InvalidOperationException: Page has no MediaBox.
   at PdfSharp.Pdf.PdfPage.Initialize(Boolean setupSizeFromMediaBox)
   at PdfSharp.Pdf.PdfPage..ctor(PdfDictionary dict)
   at PdfSharp.Pdf.ElementsBase.CreateContainer(Type type, PdfContainer oldContainer, Boolean createIndirect)
   at PdfSharp.Pdf.PdfPages.TraversePageTree(List`1 pages, PdfPageTreeNode treeNode, InheritedValues inheritedValues)
   at PdfSharp.Pdf.PdfPages.FlattenPageTree()
   at PdfSharp.Pdf.IO.PdfReader.OpenFromStream(...)
   at PdfSharp.Pdf.IO.PdfReader.Open(...)

Root cause

In PdfPages.TraversePageTree (src/foundation/src/PDFsharp/src/PdfSharp/Pdf/PdfPages.cs):

var page = (PdfPage)kid.Elements.CreateContainer(typeof(PdfPage), kid, true); // (1) constructs PdfPage -> throws
page.ApplyInheritedValues(ref inheritedValues);                              // (2) inherited /MediaBox applied here, too late

CreateContainer (1) runs the PdfPage(PdfDictionary) constructor, which calls Initialize(setupSizeFromMediaBox: true). In PdfPage.Initialize (.../Pdf/PdfPage.cs):

var rectangle = Elements.GetRectangle(InheritablePageKeys.MediaBox, false);
if (rectangle.IsEmpty)
    throw new InvalidOperationException("Page has no MediaBox.");

GetRectangle(..., false) reads only the page's own dictionary (no inheritance), so for an inheriting page it is empty and Initialize throws — before ApplyInheritedValues (2) ever runs.

Suggested fix

Apply inherited values before the page's size is set up — e.g. call ApplyInheritedValues prior to (or as part of) Initialize, or have Initialize/GetRectangle consult inherited attributes (walk /Parent) when the page has no direct /MediaBox, or defer the MediaBox validation until after inheritance has been promoted in TraversePageTree.