test: close corpus coverage gaps and make fixtures self-describing

corpus/README.md

@@ -14,6 +14,18 @@ beginning-of-file position, or end-of-file position are significant remain
 isolated under descriptive filenames. Markdown input remains free of embedded
 expected HTML and parser-specific assertions.
 
+A fixture identifies its own expected state, so that a heading covering several
+cases never leaves the reader deriving which case is which. Where a construct
+has room for content, that content names the outcome, as in
+`*asterisk opens but underscore closes_`. Where the content is itself under
+test, a label outside the construct names the case, as in
+``Opening-only one: `code``. Where neither fits, such as thematic breaks,
+escape runs, and exact-byte fixtures, a section holds one case so its heading
+is unambiguous alone. Placeholder content such as `alpha`, `foo`, or `A` and
+`B` is reserved for sections where every case shares one outcome. Naming an
+outcome in prose is not the same as embedding expected HTML or a
+parser-specific assertion, which stay out of Markdown input.
+
 ## Dialect Boundaries
 
 - [`commonmark/`](./commonmark/) covers CommonMark 0.31.2 core syntax.
@@ -52,10 +64,10 @@ rules, links, images, autolinks, inline raw HTML, hard and soft breaks, and
 textual content.
 
 The [boundary fixtures](./boundaries/) preserve LF, CRLF, CR, mixed endings,
-missing final newlines, a UTF-8 BOM, representative ASCII controls, tabs, an
-empty document, and reference-label lengths. The [stress fixtures](./stress.md)
-include 32- and 33-level link-destination nesting without treating either depth
-as a CommonMark conformance cutoff.
+missing final newlines, a UTF-8 BOM, representative ASCII controls, an invalid
+UTF-8 byte, tabs, an empty document, reference-label lengths, and long lines.
+The [stress fixtures](./stress.md) include 32- and 33-level link-destination
+nesting without treating either depth as a CommonMark conformance cutoff.
 
 Formal GFM fixtures cover [tables](./gfm/tables.md),
 [task-list items](./gfm/task-lists.md), [strikethrough](./gfm/strikethrough.md),
@@ -67,8 +79,9 @@ The extension survey includes:
 - [Footnotes](./extensions/footnotes.md)
 - [YAML](./extensions/frontmatter/yaml.md),
   [TOML](./extensions/frontmatter/toml.md), and
-  [JSON](./extensions/frontmatter/json.md) frontmatter, plus a
-  [not-at-start](./extensions/frontmatter/not-at-start.md) boundary
+  [JSON](./extensions/frontmatter/json.md) frontmatter, plus
+  [not-at-start](./extensions/frontmatter/not-at-start.md) and
+  [unclosed](./extensions/frontmatter/unclosed.md) boundaries
 - [Dollar and LaTeX-style math](./extensions/math.md)
 - [Inline, leaf, and container directives](./extensions/directives.md)
 - [Definition lists](./extensions/definition-lists.md)
@@ -81,7 +94,6 @@ The extension survey includes:
 - [Block and inline attributes](./extensions/attributes.md)
 - [Highlight, subscript, superscript, insertion, and CriticMarkup-style deletion and addition](./extensions/typography.md)
 - [Emoji shortcodes](./extensions/emoji.md)
-- [Fenced diagrams](./extensions/fenced-diagrams.md)
 
 ## Practical Documents
 
@@ -90,5 +102,6 @@ The extension survey includes:
 - [Extension survey](./practical/extension-survey.md)
 - [Technical README](./practical/technical-readme.md)
 - [Multilingual field notes](./practical/multilingual-notes.md)
+- [Round-trip normalization walkthrough](./practical/round-trip.md)
 
 ![Corpus leaf](./assets/leaf.svg)

corpus/boundaries/README.md

@@ -1,6 +1,10 @@
 # Byte and Grammar Boundaries
 
 The `bytes/` fixtures preserve exact byte sequences through attributes in the
-repository's `.gitattributes`. The `limits/` fixtures exercise the normative
-CommonMark reference-label length boundary as an adjacent 999/1000-character
-pair.
+repository's `.gitattributes`. `invalid-utf8.md` is deliberately not a valid
+UTF-8 document; it must be opened through the application rather than repaired
+by an editor that rewrites it on save.
+
+The `limits/` fixtures exercise the normative CommonMark reference-label length
+boundary as an adjacent 999/1000-character pair, and line lengths that a
+viewport-wrapping prose surface has to accommodate.

corpus/boundaries/bytes/invalid-utf8.md

@@ -0,0 +1 @@
+before�after

corpus/commonmark/blocks.md

@@ -2,8 +2,8 @@
 
 ## Block structure takes precedence over inline parsing
 
-- `one
-- two`
+- `a code span does not
+- cross a list-item boundary`
 
 ## Asterisks, hyphens, and underscores form thematic breaks
 
@@ -56,6 +56,11 @@ Paragraph becomes a heading
 
 ---
 
+## An asterisk thematic break interrupts a paragraph instead
+
+Paragraph followed by asterisks
+***
+
 ## Definitions resolve full reference links
 
 [garden]: /plots/garden "Garden title"
@@ -64,6 +69,13 @@ Paragraph becomes a heading
 
 Open the [garden] and [field report].
 
+## Forward-referenced and unreferenced definitions
+
+Open the [forward reference] before its definition appears.
+
+[forward reference]: /forward "Forward title"
+[never referenced]: /unused "Unused title"
+
 ## Duplicate reference definitions
 
 [same]: /first

corpus/commonmark/code.md

@@ -2,8 +2,8 @@
 
 ## Four spaces create an indented code block
 
-    alpha
-      beta
+    four spaces open the block
+      two further spaces stay in the content
 
 ## Indented code cannot interrupt a paragraph
 
@@ -38,6 +38,12 @@ valid backtick info string
 valid tilde info string
 ~~~
 
+## A backtick in a backtick info string does not open a fence
+
+``` language`with-backtick
+the preceding line stays paragraph text rather than opening a code block
+```
+
 ## Three leading spaces open a fence while four create indented code
 
    ```

corpus/commonmark/emphasis.md

@@ -90,25 +90,25 @@ ___underscores open a combination but asterisks close***
 
 \*\*\*not combined emphasis and strong emphasis***
 
-## Rule of three
+## A three-delimiter run splits into nested strong and emphasis
 
-***alpha** beta*
+***strong closes first** and emphasis wraps both*
 
-*alpha **beta***
+*emphasis opens first **and strong closes first***
 
-**alpha *beta***
+**strong opens first *and emphasis closes first***
 
-___alpha__ beta_
+___strong closes first__ and emphasis wraps both_
 
-_alpha __beta___
+_emphasis opens first __and strong closes first___
 
-__alpha _beta___
+__strong opens first _and emphasis closes first___
 
 ## Unequal delimiter runs leave unused markers
 
-***text*
+***two markers stay literal*
 
-***text**
+***one marker stays literal**
 
 ## Matching delimiter characters can nest
 
@@ -118,14 +118,16 @@ __alpha _beta___
 
 ## Mixed asterisk and underscore runs
 
-_**text*
+_**underscore and one asterisk stay literal*
 
-_**text**
+_**underscore stays literal while strong pairs**
 
-_**text**_
+_**underscore and asterisks all pair**_
 
-*__text_
+## Mixed runs in intraword position
 
-*__text__
+foo*__nothing pairs intraword_baz
 
-*__text__*
+foo*__underscores stay literal intraword__baz
+
+foo*__asterisks pair, underscores stay literal__*baz

corpus/commonmark/html.md

@@ -6,17 +6,73 @@
 Markdown *inside* this block is raw.
 </script>
 
+## A blank line does not end a script or style block
+
+<script>
+const before = 1;
+
+const after = 2;
+</script>
+
+<style>
+.leaf { color: green; }
+
+.bed { color: brown; }
+</style>
+
 ## Comments form an HTML block
 
 <!-- a multiline
 comment -->
 
+## A blank line does not end a comment block
+
+<!-- a comment opens here
+
+and closes after a blank line -->
+
 ## Block-level tags form an HTML block
 
 <section class="garden">
 *Markdown is raw until the blank line for this block type.*
 </section>
 
+## A blank line ends a block-level tag block before its closing tag
+
+<section class="garden">
+Raw content of the block.
+
+*This paragraph follows the block rather than belonging to it.*
+
+</section>
+
+<details>
+<summary>Collapsible summary</summary>
+
+*This paragraph sits between two block-level tag blocks.*
+
+</details>
+
+## Block-level tags may interrupt a paragraph
+
+Paragraph text before a known block tag.
+<section class="garden">
+Raw content after the interruption.
+</section>
+
+## A closing tag alone begins a block
+
+</section>
+Raw content following a closing tag that opened the block.
+
+## Attributes may span lines, omit quotation, and use single quotes
+
+<section
+  class=garden
+  data-bed='north'>
+*Raw content of a start tag spread across several lines.*
+</section>
+
 ## CDATA declarations form an HTML block
 
 <![CDATA[
@@ -41,6 +97,11 @@ north bed declaration
 *Raw until the blank line ends this type-seven block.*
 </garden-card>
 
+## Complete custom tags may not interrupt a paragraph
+
+Paragraph text before a custom tag.
+<garden-card data-bed="north">
+
 ## Tags may appear inside paragraph text
 
 Text with <span class="leaf">raw *inline* HTML</span> after it.
@@ -51,6 +112,16 @@ An empty tag <br /> and a custom tag <garden-plot data-bed="north" /> remain raw
 
 Text <!-- comment --> then <?garden inspect?> and <!GARDEN note> and <![CDATA[raw <leaf>]]>.
 
+## Mismatched inline tags remain raw
+
+Text with <em>an opening tag and a foreign </strong> closing tag.
+
+## Event-handler attributes and script content remain literal text
+
+<div onclick="alert(1)">A handler attribute stays part of the raw source.</div>
+
+An inline <img src=x onerror=alert(1) /> and a <script>alert(1)</script> span.
+
 ## Malformed tag-like text remains literal
 
 < a> <33> <__> <tag bad*name=value> </tag attribute=value>

corpus/commonmark/links-and-images.md

@@ -32,8 +32,24 @@
 
 <foo.bar.baz>
 
+<data:text/html,<script>alert(1)</script>>
+
 <https://example.com
 
+## Script-like and data schemes are ordinary destinations
+
+<javascript:alert(1)>
+
+<data:text/plain,alert>
+
+[Inline script scheme](javascript:alert(1))
+
+[Data destination](data:text/html;base64,PHNjcmlwdD5hbGVydCgxKTwvc2NyaXB0Pg==)
+
+[Entity-obfuscated scheme](&#106;avascript&#58;alert&lpar;1&rpar;)
+
+![Data image](data:image/svg+xml;base64,PHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciLz4=)
+
 ## Inline and reference image forms
 
 [leaf]: ../assets/leaf.svg "Leaf"

corpus/commonmark/lists-and-blockquotes.md

@@ -7,6 +7,14 @@ lazy continuation without another marker
 >
 > Final quoted paragraph.
 
+## Lazy continuation cannot begin a new block
+
+> Quoted paragraph before a heading
+# heading outside the quote
+
+> Quoted paragraph before a list
+- list item outside the quote
+
 ## Nested blockquotes and lists
 
 > Outer quote
@@ -76,6 +84,17 @@ An ordered list beginning with two does not interrupt:
 *
 * third
 
+## A list item can begin with a blank line
+
+-
+  content on the line after the marker
+
+## An empty list item cannot interrupt a paragraph
+
+A paragraph that
+-
+continues instead of becoming a list.
+
 ## Marker-padding boundaries
 
 - one space

corpus/commonmark/text-and-breaks.md

@@ -54,6 +54,13 @@ This line follows the break.
 Three spaces follow this line.   
 This line follows the second break.
 
+## Trailing whitespace that does not create a hard break
+
+One space follows this line. 
+This line continues the same paragraph.
+
+Two spaces follow the end of this paragraph.  
+
 ## Hard breaks using a backslash
 
 A backslash follows this line.\

corpus/environment/asset-handoff.md

@@ -24,3 +24,15 @@
 - ![Protocol-relative image](//example.com/image.png)
 - ![Unsupported bitmap](../assets/unsupported.bmp)
 - ![Outside icon](../../src-tauri/icons/128x128.png)
+
+## Raw HTML embeds
+
+<img src="../assets/leaf.svg" alt="Raw HTML image" />
+
+<video src="./missing-clip.mp4" controls></video>
+
+<audio src="./missing-tone.mp3" controls></audio>
+
+<iframe src="./article-navigator/01-overview.md"></iframe>
+
+<embed src="../../docs/specification.md" type="application/pdf" />