[CameronBrooks11]

Adds 22 conformance cases for the spacing CommonMark 0.31.2 allows inside an HTML tag (§6.6) and after one (§4.6, start condition 7). They were written from SPEC.md at f7533e4 by an author who had not seen the implementation, and land byte for byte as written (sha256 of every file checked after copying; no name collides with an existing case). Eight of them failed against the reference. That is #66, fixed here in its own commit.

Commits

1. test(conformance): add blind cases for whitespace in an HTML tag: the cases alone.
2. fix(reference): limit HTML tag spacing to spaces, tabs and a line ending: closes reference: HTML tag whitespace accepts vertical tab and form feed #66.
3. test(conformance): raise the case floor to 542: MIN_CASES 525 → 542.
4. docs(conformance): record the provenance of the tag spacing cases: the README provenance paragraph.
5. test(conformance): pin HTML tag spacing in the mutation gate: two mutants for the reader, below.
6. test(conformance): add blind cases for VT and FF after an attribute's =: five more cases, alone (below).
7. test(conformance): pin the spacing after an attribute's = in the gate: a third mutant.
8. test(conformance): raise the case floor to 547: MIN_CASES 542 → 547.
9. docs(conformance): record the provenance of the five tag spacing cases: the README paragraph, extended.

Every case, and its outcome against the reference

"Before" is the reference at f7533e4, unchanged. "After" is with commit 2. VT is U+000B, FF is U+000C.

case	input (one line)	expected	before	after
accepts-a-form-feed-between-tag-name-and-attribute	<a FF b=c>	paragraph	FAIL: got raw-html	pass
accepts-a-vertical-tab-between-tag-name-and-attribute	<a VT b=c>	paragraph	FAIL: got raw-html	pass
refuses-a-tab-between-tag-name-and-attribute	<a tab b=c>	raw-html	pass	pass
accepts-a-form-feed-after-an-open-tag	<a b=c> FF	paragraph	FAIL: got raw-html	pass
accepts-a-vertical-tab-after-an-open-tag	<a b=c> VT	paragraph	FAIL: got raw-html	pass
refuses-a-tab-after-an-open-tag	<a b=c> tab	raw-html	pass	pass
accepts-a-form-feed-before-an-attribute-equals-sign	<a b FF =c>	paragraph	FAIL: got raw-html	pass
refuses-a-tab-before-an-attribute-equals-sign	<a b tab =c>	raw-html	pass	pass
accepts-a-form-feed-before-an-open-tag-s-bracket	<a FF >	paragraph	FAIL: got raw-html	pass
refuses-a-tab-before-an-open-tag-s-bracket	<a tab >	raw-html	pass	pass
accepts-a-form-feed-before-a-closing-tag-s-bracket	</a FF >	paragraph	FAIL: got raw-html	pass
refuses-a-tab-before-a-closing-tag-s-bracket	</a tab >	raw-html	pass	pass
refuses-a-form-feed-opening-an-unquoted-attribute-value	<a b= FF c>	raw-html	pass	pass
refuses-a-tab-before-an-unquoted-attribute-value	<a b= tab c>	raw-html	pass	pass
accepts-a-backtick-opening-an-unquoted-attribute-value	<a b=`c>	paragraph	pass	pass
refuses-a-space-and-tab-after-an-open-tag	<a b=c> space tab	raw-html	pass	pass
accepts-a-space-and-form-feed-after-an-open-tag	<a b=c> space FF	paragraph	FAIL: got raw-html	pass
accepts-a-form-feed-before-a-single-quoted-attribute-value	<a b= FF 'q'>	paragraph	FAIL: got raw-html	pass
accepts-a-vertical-tab-before-a-single-quoted-attribute-value	<a b= VT 'q'>	paragraph	FAIL: got raw-html	pass
refuses-a-tab-before-a-single-quoted-attribute-value	<a b= tab 'q'>	raw-html	pass	pass
accepts-a-vertical-tab-before-an-attribute-equals-sign	<a b VT =c>	paragraph	FAIL: got raw-html	pass
accepts-a-vertical-tab-before-an-open-tag-s-bracket	<a VT >	paragraph	FAIL: got raw-html	pass

The last five rows are the second set (commit 6). Their "before" column comes from running them against the reader as at f7533e4, because the fix had already landed on this branch when they were added. All are under refusal/raw-html/. The eight failures are exactly the accepting cases with VT or FF in a tag's spacing or after the tag, which is what #66 describes. No other case failed. The author recorded no question the text leaves undecided (AMBIGUITIES: none).

The second set: the spacing after =

Review found that no case pinned the spacing after an attribute's =. With VT and FF put back into that one place, every case passed (0 failures across 542; reproduced). <a b= FF 'q'> tells the two readings apart. Under §6.6 the FF is not spacing, so it starts an unquoted value. That value has to stop before ', so the line holds no tag. If FF were spacing, 'q' would be a quoted value, and the line would be a type 7 HTML block.

The same author added five cases, from the same files and with the same constraints, and landed them byte for byte (hash-checked; no name collides). The author recorded no undecided question. Two of the five are the VT twins of refuses-a-tab-before-an-attribute-equals-sign and refuses-a-tab-before-an-open-tag-s-bracket. Those two refusals are left unchanged, so they do not list the twins as controls. The new tab refusal lists the FF and VT single-quoted accepts as its controls.

Commit d3dd75c's message says that a form feed or vertical tab "between the tag name and an attribute, before an attribute's =, before an open or closing tag's >, or after the tag" makes no HTML block. Its cases covered VT in only two of those positions: between the tag name and an attribute, and after an open tag. The new cases add VT before = and before an open tag's >. One position the message names still has no VT case: VT before a closing tag's >. The FF case there (accepts-a-form-feed-before-a-closing-tag-s-bracket) goes through the same spacing class, and the in-tag mutant puts VT and FF back together, so no mutant in the table survives on it.

The reference fix (#66)

The reader used one class, space, tab, LF, VT, FF and CR, both for spacing inside a tag and for the run after it. Now:

• Spacing inside a tag (§6.6: "spaces, tabs, and up to one line ending") is spaces and tabs with at most one line ending among them. Before an attribute it must be non-empty.
• The run after the tag (§4.6, condition 7: "zero or more spaces and tabs") is spaces and tabs only.
• The unquoted attribute value class from reference: unquoted attribute value refuses control characters CommonMark allows #64 is unchanged, so <a b= FF c> is still refused: the FF is part of the value.

Each half is pinned by its own cases, and just mutants now checks both (commits 5 and 7). The reader's table has three new mutants:

• tag-spacing-takes-vertical-tab-and-form-feed puts VT and FF back into the spacing inside a tag (_WS and _WS1). It is killed by 9 cases: the 5 from the first set and the 4 accepting cases of the second.
• spacing-after-an-attribute-equals-takes-vertical-tab-and-form-feed puts VT and FF back into the spacing after = only. It survived the first set and is killed by 2 cases: the FF and VT single-quoted accepts.
• a-tag-may-be-followed-by-vertical-tab-and-form-feed puts VT and FF back into condition 7's trailing run. It is killed by 3 cases: FF, VT, and space plus FF after an open tag.

The reader splits lines before it matches a start condition, so no line ending reaches these patterns and no case can observe the "up to one line ending" limit. Matched directly, the patterns take <a, LF, b=c> and </a, LF, > as tags, and reject the same with two LFs. They also reject <a href='bar'title=title>, §6.6's Missing whitespace example.

Gates

Each run on its own, exit code read directly:

just check    exit 0   All checks passed!
just test     exit 0   181 passed
just conform  exit 0   0 failure(s) across 547 case(s) in the fixture tree
just tree     exit 0   547 cases: 178 accept, 91 refuse, 278 anchor / 0 problem(s)
just mutants  exit 0   reader:   96 killed, 0 survived, 0 equivalent, 0 stale, 0 broken
                       resolver: 123 killed, 0 survived, 1 equivalent, 0 stale, 0 broken

On main the reader's table has 93 mutants, none on tag spacing. The three new ones bring it to 96, and all are killed. The resolver's totals are unchanged.

Refs #66