diff --git a/docs/parity.md b/docs/parity.md
index 9ed1dc4..0cff493 100644
--- a/docs/parity.md
+++ b/docs/parity.md
@@ -46,7 +46,7 @@ No “Must” row is unowned. The owner is the story that closes it: `US1 creato
| 20 | JavaScript execution, 3D, multimedia, geospatial, public-key encryption | PDFsharp: no or partial. PdfPig: no. | Non-goal: actions may be copied or refused with a warning; nothing is executed, and no new encryption surface is added. | 1.1 (excluded) | Product boundary (ADR-001) | None by design | Out-of-scope statement in `README.md` and ADR-001 | Out of scope |
| 21 | Drop-in source or binary compatibility with PDFsharp or PdfPig | N/A | Non-goal: capability parity only; no reference package is referenced by the suite. | 1.1 (excluded) | Product boundary (ADR-001) | None by design | ADR-001 and the dependency check in quickstart step 8 | Out of scope |
-| 22 | Reading the fonts a document uses, and whether each program travels with it | PDFsharp: no dedicated API (font resolution is internal). PdfPig: fonts are reachable through a page's letters, not as a document-level answer. | Must: one entry per font keyed by the object its resource names, with its kind, program, encoding, `/ToUnicode` presence and the pages that select it; a resource that does not resolve and a font that states no descriptor are reported as warnings rather than dropped. | 1.6 (Unreleased) | US3 reading | `PdfFontInspectorTests.Reports_every_font_the_document_states`, `Reports_an_embedded_subset_with_its_program_and_pages`, `Reports_a_font_that_carries_no_program`, `Resolves_a_composite_font_to_its_descendant`, `Reads_a_font_a_form_paints_with`, `Reports_a_resource_that_names_no_object`, `Reports_the_fonts_of_a_real_world_document`, `Reads_the_fonts_of_a_document_it_encrypted_itself`, `Opens_an_encrypted_document_with_its_password`, `Reports_a_type3_font_as_carrying_its_own_glyphs`, `Reports_a_type3_font_that_states_no_glyphs`, `Reports_a_cid_type0_font_with_a_custom_encoding`, `Reports_where_each_program_is_stored`, `Reports_a_composite_font_that_names_no_descendant`, `Reports_a_multiple_master_font`, `Recognises_a_subset_tag_and_rejects_a_name_that_only_looks_like_one`, `Reads_the_resources_a_page_inherits_from_the_tree`, `Stops_the_walk_at_the_nesting_depth_it_is_given`, `Returns_no_font_for_a_document_that_stays_closed`, `Refuses_a_null_source` | Pending for the fixture that pins every other branch: it is built in memory by `PdfContentFixtures.Fonts()`, so a validator run needs it written to disk first. The real-world leg **was** run: `uv run --with pikepdf --with pypdf python docs/tools/validate-pdf.py tests/SharpCorePDF.Tests/Assets/parity/labelled/table.pdf` (`qpdf check: no problems`, 1.3, 1 page, 7 objects) and a pikepdf listing of the same file's `/Font` entries (`F1` Helvetica, `F2` Courier, `F3` Times-Roman, all `/Type1`, `/WinAnsiEncoding`, no `/FontDescriptor`) | Implemented (2026-11-06) — twenty tests are in the suite, the real-world leg agrees with pikepdf on the same file, and the mutation reading is recorded in `docs/mutation-testing.md`: 97.44% over `PdfFontDescriptorLookup.cs`, 94.44% over `FontAccumulator.cs`, 90.91% over `PdfFontInspector.cs`, with the single survivor in the last two explained as unobservable |
+| 22 | Reading the fonts a document uses, and whether each program travels with it | PDFsharp: no dedicated API (font resolution is internal). PdfPig: fonts are reachable through a page's letters, not as a document-level answer. | Must: one entry per font keyed by the object its resource names, with its kind, program, encoding, `/ToUnicode` presence and the pages that select it; a resource that does not resolve and a font that states no descriptor are reported as warnings rather than dropped. | 1.6 (Unreleased) | US3 reading | `PdfFontInspectorTests.Reports_every_font_the_document_states`, `Reports_an_embedded_subset_with_its_program_and_pages`, `Reports_a_font_that_carries_no_program`, `Resolves_a_composite_font_to_its_descendant`, `Reads_a_font_a_form_paints_with`, `Reports_a_resource_that_names_no_object`, `Reports_the_fonts_of_a_real_world_document`, `Reads_the_fonts_of_a_document_it_encrypted_itself`, `Opens_an_encrypted_document_with_its_password`, `Reports_a_type3_font_as_carrying_its_own_glyphs`, `Reports_a_type3_font_that_states_no_glyphs`, `Reports_a_cid_type0_font_with_a_custom_encoding`, `Reports_where_each_program_is_stored`, `Reports_a_composite_font_that_names_no_descendant`, `Reports_a_multiple_master_font`, `Recognises_a_subset_tag_and_rejects_a_name_that_only_looks_like_one`, `Reads_the_resources_a_page_inherits_from_the_tree`, `Stops_the_walk_at_the_nesting_depth_it_is_given`, `Returns_no_font_for_a_document_that_stays_closed`, `Refuses_a_null_source` | Closed on 2026-11-14 for the in-memory fixture too: `PdfFontInspectorTests.Writes_the_font_fixture_the_external_validator_reads` writes `PdfContentFixtures.Fonts()` to `%TEMP%/scpdf-fonts-fixture.pdf`, and `uv run --with pikepdf --with pypdf python docs/tools/validate-pdf.py` read it (PDF 1.7, 4 pages, 41 objects, 5 streams decoded, 0 failed). `qpdf --check` reports one repairable problem — object 9's page tree child states no `/Resources` — which is deliberate: that page is the "states no resources at all" shape the fixture exists for. A pikepdf walk of the object graph listed **17 font objects / 15 unique `/BaseFont` names**, exactly the names the inspector returns: `ABCDEF+Arial`, `Helvetica`, `Courier`, `ZapfDingbats`, `XYZABC+SomeFont`, `ZapfType3`, `MinionMM`, `Loose`, `ABCDEFHelvetica`, `abcDEF+Weird`, `EmptyType3`, `ABCDEF+X`, `FontFileOne`, `CffOne`, `BaseEncoded`. The real-world leg remains: `uv run --with pikepdf --with pypdf python docs/tools/validate-pdf.py tests/SharpCorePDF.Tests/Assets/parity/labelled/table.pdf` (`qpdf check: no problems`, 1.3, 1 page, 7 objects) and a pikepdf listing of the same file's `/Font` entries (`F1` Helvetica, `F2` Courier, `F3` Times-Roman, all `/Type1`, `/WinAnsiEncoding`, no `/FontDescriptor`) | Implemented (2026-11-06) — twenty tests are in the suite, the real-world leg agrees with pikepdf on the same file, and the mutation reading is recorded in `docs/mutation-testing.md`: 97.44% over `PdfFontDescriptorLookup.cs`, 94.44% over `FontAccumulator.cs`, 90.91% over `PdfFontInspector.cs`, with the single survivor in the last two explained as unobservable |
## 1.1 scope decision (D3, 2026-10-22)
diff --git a/tests/SharpCorePDF.Tests/PdfFontInspectorTests.cs b/tests/SharpCorePDF.Tests/PdfFontInspectorTests.cs
index f882e13..827a8fa 100644
--- a/tests/SharpCorePDF.Tests/PdfFontInspectorTests.cs
+++ b/tests/SharpCorePDF.Tests/PdfFontInspectorTests.cs
@@ -96,6 +96,42 @@ public async Task Reports_every_font_the_document_states()
Assert.Equal(expected, result.Items.Select(font => font.BaseFont));
}
+ ///
+ /// Writes the in-memory font fixture to disk, so the external pikepdf/pypdf run recorded
+ /// in docs/parity.md row 22 reads the same bytes, and pins the font tuples that run lists.
+ ///
+ [Fact]
+ public async Task Writes_the_font_fixture_the_external_validator_reads()
+ {
+ var path = Path.Combine(Path.GetTempPath(), "scpdf-fonts-fixture.pdf");
+
+ await File.WriteAllBytesAsync(path, PdfContentFixtures.Fonts(), Token);
+
+ var result = await FontsAsync(await File.ReadAllBytesAsync(path, Token));
+
+ string[] expected =
+ [
+ "ABCDEF+Arial",
+ "Helvetica",
+ "Courier",
+ "ZapfDingbats",
+ "XYZABC+SomeFont",
+ "ZapfType3",
+ "MinionMM",
+ "Loose",
+ "ABCDEFHelvetica",
+ "abcDEF+Weird",
+ "EmptyType3",
+ "ABCDEF+X",
+ "FontFileOne",
+ "CffOne",
+ "BaseEncoded",
+ ];
+
+ Assert.Equal(expected, result.Items.Select(font => font.BaseFont));
+ Assert.True(new FileInfo(path).Length > 0);
+ }
+
/// A subset that carries its program is reported with the program, the encoding and its pages.
[Fact]
public async Task Reports_an_embedded_subset_with_its_program_and_pages()