@zowe/cli 8.36.1 → 8.37.0

package/node_modules/{htmlparser2/dist/esm → sanitize-html/node_modules/htmlparser2/dist}/Tokenizer.js

@@ -1,4 +1,4 @@
-import { EntityDecoder, DecodingMode, htmlDecodeTree, xmlDecodeTree, } from "entities/decode";
+import { DecodingMode, EntityDecoder, htmlDecodeTree, xmlDecodeTree, } from "entities/decode";
 var CharCodes;
 (function (CharCodes) {
     CharCodes[CharCodes["Tab"] = 9] = "Tab";
@@ -55,13 +55,13 @@ var State;
     // Comments & CDATA
     State[State["BeforeComment"] = 18] = "BeforeComment";
     State[State["CDATASequence"] = 19] = "CDATASequence";
-    State[State["InSpecialComment"] = 20] = "InSpecialComment";
-    State[State["InCommentLike"] = 21] = "InCommentLike";
+    State[State["DeclarationSequence"] = 20] = "DeclarationSequence";
+    State[State["InSpecialComment"] = 21] = "InSpecialComment";
+    State[State["InCommentLike"] = 22] = "InCommentLike";
     // Special tags
-    State[State["BeforeSpecialS"] = 22] = "BeforeSpecialS";
-    State[State["BeforeSpecialT"] = 23] = "BeforeSpecialT";
-    State[State["SpecialStartSequence"] = 24] = "SpecialStartSequence";
-    State[State["InSpecialTag"] = 25] = "InSpecialTag";
+    State[State["SpecialStartSequence"] = 23] = "SpecialStartSequence";
+    State[State["InSpecialTag"] = 24] = "InSpecialTag";
+    State[State["InPlainText"] = 25] = "InPlainText";
     State[State["InEntity"] = 26] = "InEntity";
 })(State || (State = {}));
 function isWhitespace(c) {
@@ -78,6 +78,9 @@ function isASCIIAlpha(c) {
     return ((c >= CharCodes.LowerA && c <= CharCodes.LowerZ) ||
         (c >= CharCodes.UpperA && c <= CharCodes.UpperZ));
 }
+/**
+ * Quote style used for parsed attributes.
+ */
 export var QuoteType;
 (function (QuoteType) {
     QuoteType[QuoteType["NoValue"] = 0] = "NoValue";
@@ -92,9 +95,21 @@ export var QuoteType;
  * sequences with an increased offset.
  */
 const Sequences = {
+    Empty: new Uint8Array(0),
     Cdata: new Uint8Array([0x43, 0x44, 0x41, 0x54, 0x41, 0x5b]), // CDATA[
     CdataEnd: new Uint8Array([0x5d, 0x5d, 0x3e]), // ]]>
-    CommentEnd: new Uint8Array([0x2d, 0x2d, 0x3e]), // `-->`
+    CommentEnd: new Uint8Array([0x2d, 0x2d, 0x21, 0x3e]), // `--!>`
+    Doctype: new Uint8Array([0x64, 0x6f, 0x63, 0x74, 0x79, 0x70, 0x65]), // `doctype`
+    IframeEnd: new Uint8Array([0x3c, 0x2f, 0x69, 0x66, 0x72, 0x61, 0x6d, 0x65]), // `</iframe`
+    NoembedEnd: new Uint8Array([
+        0x3c, 0x2f, 0x6e, 0x6f, 0x65, 0x6d, 0x62, 0x65, 0x64,
+    ]), // `</noembed`
+    NoframesEnd: new Uint8Array([
+        0x3c, 0x2f, 0x6e, 0x6f, 0x66, 0x72, 0x61, 0x6d, 0x65, 0x73,
+    ]), // `</noframes`
+    Plaintext: new Uint8Array([
+        0x3c, 0x2f, 0x70, 0x6c, 0x61, 0x69, 0x6e, 0x74, 0x65, 0x78, 0x74,
+    ]), // `</plaintext`
     ScriptEnd: new Uint8Array([0x3c, 0x2f, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74]), // `</script`
     StyleEnd: new Uint8Array([0x3c, 0x2f, 0x73, 0x74, 0x79, 0x6c, 0x65]), // `</style`
     TitleEnd: new Uint8Array([0x3c, 0x2f, 0x74, 0x69, 0x74, 0x6c, 0x65]), // `</title`
@@ -103,31 +118,52 @@ const Sequences = {
     ]), // `</textarea`
     XmpEnd: new Uint8Array([0x3c, 0x2f, 0x78, 0x6d, 0x70]), // `</xmp`
 };
+/**
+ * Maps the first lowercase character of an HTML tag name to the sequence
+ * used for special-tag detection.  All sequences share a common layout
+ * where index 2 is the first tag-name character, so matching always
+ * continues from offset 3.
+ */
+const specialStartSequences = new Map([
+    [Sequences.IframeEnd[2], Sequences.IframeEnd],
+    [Sequences.NoembedEnd[2], Sequences.NoembedEnd],
+    [Sequences.Plaintext[2], Sequences.Plaintext],
+    [Sequences.ScriptEnd[2], Sequences.ScriptEnd],
+    [Sequences.TitleEnd[2], Sequences.TitleEnd],
+    [Sequences.XmpEnd[2], Sequences.XmpEnd],
+]);
+/**
+ * Tokenizer implementation used by `Parser`.
+ */
 export default class Tokenizer {
-    constructor({ xmlMode = false, decodeEntities = true, }, cbs) {
+    cbs;
+    /** The current state the tokenizer is in. */
+    state = State.Text;
+    /** The read buffer. */
+    buffer = "";
+    /** The beginning of the section that is currently being read. */
+    sectionStart = 0;
+    /** The index within the buffer that we are currently looking at. */
+    index = 0;
+    /** The start of the last entity. */
+    entityStart = 0;
+    /** Some behavior, eg. when decoding entities, is done while we are in another state. This keeps track of the other state type. */
+    baseState = State.Text;
+    /** For special parsing behavior inside of script and style tags. */
+    isSpecial = false;
+    /** Indicates whether the tokenizer has been paused. */
+    running = true;
+    /** The offset of the current buffer. */
+    offset = 0;
+    xmlMode;
+    decodeEntities;
+    recognizeSelfClosing;
+    entityDecoder;
+    constructor({ xmlMode = false, decodeEntities = true, recognizeSelfClosing = xmlMode, }, cbs) {
         this.cbs = cbs;
-        /** The current state the tokenizer is in. */
-        this.state = State.Text;
-        /** The read buffer. */
-        this.buffer = "";
-        /** The beginning of the section that is currently being read. */
-        this.sectionStart = 0;
-        /** The index within the buffer that we are currently looking at. */
-        this.index = 0;
-        /** The start of the last entity. */
-        this.entityStart = 0;
-        /** Some behavior, eg. when decoding entities, is done while we are in another state. This keeps track of the other state type. */
-        this.baseState = State.Text;
-        /** For special parsing behavior inside of script and style tags. */
-        this.isSpecial = false;
-        /** Indicates whether the tokenizer has been paused. */
-        this.running = true;
-        /** The offset of the current buffer. */
-        this.offset = 0;
-        this.currentSequence = undefined;
-        this.sequenceIndex = 0;
         this.xmlMode = xmlMode;
         this.decodeEntities = decodeEntities;
+        this.recognizeSelfClosing = recognizeSelfClosing;
         this.entityDecoder = new EntityDecoder(xmlMode ? xmlDecodeTree : htmlDecodeTree, (cp, consumed) => this.emitCodePoint(cp, consumed));
     }
     reset() {
@@ -136,7 +172,9 @@ export default class Tokenizer {
         this.sectionStart = 0;
         this.index = 0;
         this.baseState = State.Text;
-        this.currentSequence = undefined;
+        this.isSpecial = false;
+        this.currentSequence = Sequences.Empty;
+        this.sequenceIndex = 0;
         this.running = true;
         this.offset = 0;
     }
@@ -171,62 +209,74 @@ export default class Tokenizer {
             this.startEntity();
         }
     }
-    stateSpecialStartSequence(c) {
-        const isEnd = this.sequenceIndex === this.currentSequence.length;
-        const isMatch = isEnd
-            ? // If we are at the end of the sequence, make sure the tag name has ended
-                isEndOfTagSection(c)
-            : // Otherwise, do a case-insensitive comparison
-                (c | 0x20) === this.currentSequence[this.sequenceIndex];
-        if (!isMatch) {
-            this.isSpecial = false;
-        }
-        else if (!isEnd) {
-            this.sequenceIndex++;
-            return;
-        }
-        this.sequenceIndex = 0;
-        this.state = State.InTagName;
-        this.stateInTagName(c);
-    }
-    /** Look for an end tag. For <title> tags, also decode entities. */
-    stateInSpecialTag(c) {
-        if (this.sequenceIndex === this.currentSequence.length) {
-            if (c === CharCodes.Gt || isWhitespace(c)) {
-                const endOfText = this.index - this.currentSequence.length;
-                if (this.sectionStart < endOfText) {
-                    // Spoof the index so that reported locations match up.
-                    const actualIndex = this.index;
-                    this.index = endOfText;
-                    this.cbs.ontext(this.sectionStart, endOfText);
-                    this.index = actualIndex;
-                }
-                this.isSpecial = false;
-                this.sectionStart = endOfText + 2; // Skip over the `</`
-                this.stateInClosingTagName(c);
-                return; // We are done; skip the rest of the function.
-            }
+    currentSequence = Sequences.Empty;
+    sequenceIndex = 0;
+    enterTagBody() {
+        if (this.currentSequence === Sequences.Plaintext) {
+            this.currentSequence = Sequences.Empty;
+            this.state = State.InPlainText;
+        }
+        else if (this.isSpecial) {
+            this.state = State.InSpecialTag;
             this.sequenceIndex = 0;
         }
-        if ((c | 0x20) === this.currentSequence[this.sequenceIndex]) {
-            this.sequenceIndex += 1;
+        else {
+            this.state = State.Text;
         }
-        else if (this.sequenceIndex === 0) {
-            if (this.currentSequence === Sequences.TitleEnd) {
-                // We have to parse entities in <title> tags.
-                if (this.decodeEntities && c === CharCodes.Amp) {
-                    this.startEntity();
+    }
+    /**
+     * Match the opening tag name against an HTML text-only tag sequence.
+     *
+     * Some tags share an initial prefix (`script`/`style`, `title`/`textarea`,
+     * `noembed`/`noframes`), so we may switch to an alternate sequence at the
+     * first distinguishing byte.  On a successful full match we fall back to
+     * the normal tag-name state; a later `>` will enter raw-text, RCDATA, or
+     * plaintext mode based on `currentSequence` / `isSpecial`.
+     * @param c Current character code point.
+     */
+    stateSpecialStartSequence(c) {
+        const lower = c | 0x20;
+        // Still matching — check for an alternate sequence at branch points.
+        if (this.sequenceIndex < this.currentSequence.length) {
+            if (lower === this.currentSequence[this.sequenceIndex]) {
+                this.sequenceIndex++;
+                return;
+            }
+            if (this.sequenceIndex === 3) {
+                if (this.currentSequence === Sequences.ScriptEnd &&
+                    lower === Sequences.StyleEnd[3]) {
+                    this.currentSequence = Sequences.StyleEnd;
+                    this.sequenceIndex = 4;
+                    return;
+                }
+                if (this.currentSequence === Sequences.TitleEnd &&
+                    lower === Sequences.TextareaEnd[3]) {
+                    this.currentSequence = Sequences.TextareaEnd;
+                    this.sequenceIndex = 4;
+                    return;
                 }
             }
-            else if (this.fastForwardTo(CharCodes.Lt)) {
-                // Outside of <title> tags, we can fast-forward.
-                this.sequenceIndex = 1;
+            else if (this.sequenceIndex === 4 &&
+                this.currentSequence === Sequences.NoembedEnd &&
+                lower === Sequences.NoframesEnd[4]) {
+                this.currentSequence = Sequences.NoframesEnd;
+                this.sequenceIndex = 5;
+                return;
             }
         }
-        else {
-            // If we see a `<`, set the sequence index to 1; useful for eg. `<</script>`.
-            this.sequenceIndex = Number(c === CharCodes.Lt);
+        else if (isEndOfTagSection(c)) {
+            // Full match on a valid tag boundary — keep the sequence.
+            this.sequenceIndex = 0;
+            this.state = State.InTagName;
+            this.stateInTagName(c);
+            return;
         }
+        // No match — abandon special-tag detection.
+        this.isSpecial = false;
+        this.currentSequence = Sequences.Empty;
+        this.sequenceIndex = 0;
+        this.state = State.InTagName;
+        this.stateInTagName(c);
     }
     stateCDATASequence(c) {
         if (c === Sequences.Cdata[this.sequenceIndex]) {
@@ -239,14 +289,20 @@ export default class Tokenizer {
         }
         else {
             this.sequenceIndex = 0;
-            this.state = State.InDeclaration;
-            this.stateInDeclaration(c); // Reconsume the character
+            if (this.xmlMode) {
+                this.state = State.InDeclaration;
+                this.stateInDeclaration(c); // Reconsume the character
+            }
+            else {
+                this.state = State.InSpecialComment;
+                this.stateInSpecialComment(c); // Reconsume the character
+            }
         }
     }
     /**
      * When we wait for one specific character, we can speed things up
      * by skipping through the buffer until we find it.
-     *
+     * @param c Current character code point.
      * @returns Whether the character was found.
      */
     fastForwardTo(c) {
@@ -264,6 +320,16 @@ export default class Tokenizer {
         this.index = this.buffer.length + this.offset - 1;
         return false;
     }
+    /**
+     * Emit a comment token and return to the text state.
+     * @param offset Number of characters in the end sequence that have already been matched.
+     */
+    emitComment(offset) {
+        this.cbs.oncomment(this.sectionStart, this.index, offset);
+        this.sequenceIndex = 0;
+        this.sectionStart = this.index + 1;
+        this.state = State.Text;
+    }
     /**
      * Comments and CDATA end with `-->` and `]]>`.
      *
@@ -271,15 +337,40 @@ export default class Tokenizer {
      * - Their end sequences have a distinct character they start with.
      * - That character is then repeated, so we have to check multiple repeats.
      * - All characters but the start character of the sequence can be skipped.
+     * @param c Current character code point.
      */
     stateInCommentLike(c) {
-        if (c === this.currentSequence[this.sequenceIndex]) {
+        if (!this.xmlMode &&
+            this.currentSequence === Sequences.CommentEnd &&
+            this.sequenceIndex <= 1 &&
+            /*
+             * We're still at the very start of the comment: the only
+             * characters consumed since `<!--` are the dashes that
+             * advanced sequenceIndex (0 for `<!-->`, 1 for `<!--->`).
+             */
+            this.index === this.sectionStart + this.sequenceIndex &&
+            c === CharCodes.Gt) {
+            // Abruptly closed empty HTML comment.
+            this.emitComment(this.sequenceIndex);
+        }
+        else if (this.currentSequence === Sequences.CommentEnd &&
+            this.sequenceIndex === 2 &&
+            c === CharCodes.Gt) {
+            // `!` is optional here, so the same sequence also accepts `-->`.
+            this.emitComment(2);
+        }
+        else if (this.currentSequence === Sequences.CommentEnd &&
+            this.sequenceIndex === this.currentSequence.length - 1 &&
+            c !== CharCodes.Gt) {
+            this.sequenceIndex = Number(c === CharCodes.Dash);
+        }
+        else if (c === this.currentSequence[this.sequenceIndex]) {
             if (++this.sequenceIndex === this.currentSequence.length) {
                 if (this.currentSequence === Sequences.CdataEnd) {
                     this.cbs.oncdata(this.sectionStart, this.index, 2);
                 }
                 else {
-                    this.cbs.oncomment(this.sectionStart, this.index, 2);
+                    this.cbs.oncomment(this.sectionStart, this.index, 3);
                 }
                 this.sequenceIndex = 0;
                 this.sectionStart = this.index + 1;
@@ -302,15 +393,56 @@ export default class Tokenizer {
      *
      * XML allows a lot more characters here (@see https://www.w3.org/TR/REC-xml/#NT-NameStartChar).
      * We allow anything that wouldn't end the tag.
+     * @param c Current character code point.
      */
     isTagStartChar(c) {
         return this.xmlMode ? !isEndOfTagSection(c) : isASCIIAlpha(c);
     }
-    startSpecial(sequence, offset) {
-        this.isSpecial = true;
-        this.currentSequence = sequence;
-        this.sequenceIndex = offset;
-        this.state = State.SpecialStartSequence;
+    /**
+     * Scan raw-text / RCDATA content for the matching end tag.
+     *
+     * For RCDATA tags (`<title>`, `<textarea>`) entities are decoded inline.
+     * For raw-text tags (`<script>`, `<style>`, etc.) we fast-forward to `<`.
+     * @param c Current character code point.
+     */
+    stateInSpecialTag(c) {
+        if (this.sequenceIndex === this.currentSequence.length) {
+            if (isEndOfTagSection(c)) {
+                const endOfText = this.index - this.currentSequence.length;
+                if (this.sectionStart < endOfText) {
+                    // Spoof the index so that reported locations match up.
+                    const actualIndex = this.index;
+                    this.index = endOfText;
+                    this.cbs.ontext(this.sectionStart, endOfText);
+                    this.index = actualIndex;
+                }
+                this.isSpecial = false;
+                this.sectionStart = endOfText + 2; // Skip over the `</`
+                this.stateInClosingTagName(c);
+                return; // We are done; skip the rest of the function.
+            }
+            this.sequenceIndex = 0;
+        }
+        if ((c | 0x20) === this.currentSequence[this.sequenceIndex]) {
+            this.sequenceIndex += 1;
+        }
+        else if (this.sequenceIndex === 0) {
+            if (this.currentSequence === Sequences.TitleEnd ||
+                this.currentSequence === Sequences.TextareaEnd) {
+                // RCDATA tags have to parse entities while still looking for their end tag.
+                if (this.decodeEntities && c === CharCodes.Amp) {
+                    this.startEntity();
+                }
+            }
+            else if (this.fastForwardTo(CharCodes.Lt)) {
+                // Outside of RCDATA tags, we can fast-forward.
+                this.sequenceIndex = 1;
+            }
+        }
+        else {
+            // If we see a `<`, set the sequence index to 1; useful for eg. `<</script>`.
+            this.sequenceIndex = Number(c === CharCodes.Lt);
+        }
     }
     stateBeforeTagName(c) {
         if (c === CharCodes.ExclamationMark) {
@@ -318,24 +450,29 @@ export default class Tokenizer {
             this.sectionStart = this.index + 1;
         }
         else if (c === CharCodes.Questionmark) {
-            this.state = State.InProcessingInstruction;
-            this.sectionStart = this.index + 1;
+            if (this.xmlMode) {
+                this.state = State.InProcessingInstruction;
+                this.sequenceIndex = 0;
+                this.sectionStart = this.index + 1;
+            }
+            else {
+                this.state = State.InSpecialComment;
+                this.sectionStart = this.index;
+            }
         }
         else if (this.isTagStartChar(c)) {
-            const lower = c | 0x20;
             this.sectionStart = this.index;
-            if (this.xmlMode) {
+            const special = this.xmlMode || this.cbs.isInForeignContext?.()
+                ? undefined
+                : specialStartSequences.get(c | 0x20);
+            if (special === undefined) {
                 this.state = State.InTagName;
             }
-            else if (lower === Sequences.ScriptEnd[2]) {
-                this.state = State.BeforeSpecialS;
-            }
-            else if (lower === Sequences.TitleEnd[2] ||
-                lower === Sequences.XmpEnd[2]) {
-                this.state = State.BeforeSpecialT;
-            }
             else {
-                this.state = State.InTagName;
+                this.isSpecial = true;
+                this.currentSequence = special;
+                this.sequenceIndex = 3;
+                this.state = State.SpecialStartSequence;
             }
         }
         else if (c === CharCodes.Slash) {
@@ -356,10 +493,19 @@ export default class Tokenizer {
     }
     stateBeforeClosingTagName(c) {
         if (isWhitespace(c)) {
-            // Ignore
+            if (this.xmlMode) {
+                // Ignore
+            }
+            else {
+                this.state = State.InSpecialComment;
+                this.sectionStart = this.index;
+            }
         }
         else if (c === CharCodes.Gt) {
             this.state = State.Text;
+            if (!this.xmlMode) {
+                this.sectionStart = this.index + 1;
+            }
         }
         else {
             this.state = this.isTagStartChar(c)
@@ -369,7 +515,7 @@ export default class Tokenizer {
         }
     }
     stateInClosingTagName(c) {
-        if (c === CharCodes.Gt || isWhitespace(c)) {
+        if (isEndOfTagSection(c)) {
             this.cbs.onclosetag(this.sectionStart, this.index);
             this.sectionStart = -1;
             this.state = State.AfterClosingTagName;
@@ -386,13 +532,7 @@ export default class Tokenizer {
     stateBeforeAttributeName(c) {
         if (c === CharCodes.Gt) {
             this.cbs.onopentagend(this.index);
-            if (this.isSpecial) {
-                this.state = State.InSpecialTag;
-                this.sequenceIndex = 0;
-            }
-            else {
-                this.state = State.Text;
-            }
+            this.enterTagBody();
             this.sectionStart = this.index + 1;
         }
         else if (c === CharCodes.Slash) {
@@ -403,12 +543,26 @@ export default class Tokenizer {
             this.sectionStart = this.index;
         }
     }
+    /**
+     * Handle `/` before `>` in an opening tag.
+     *
+     * In HTML mode, text-only tags ignore the self-closing flag and still enter
+     * their raw-text/RCDATA/plaintext state unless self-closing tags are being
+     * recognized. In XML mode, or for ordinary tags, the tokenizer returns to
+     * regular text parsing after emitting the self-closing callback.
+     * @param c Current character code point.
+     */
     stateInSelfClosingTag(c) {
         if (c === CharCodes.Gt) {
             this.cbs.onselfclosingtag(this.index);
-            this.state = State.Text;
             this.sectionStart = this.index + 1;
+            if (!this.recognizeSelfClosing) {
+                this.enterTagBody();
+                return;
+            }
+            this.state = State.Text;
             this.isSpecial = false; // Reset special state, in case of self-closing special tags
+            this.currentSequence = Sequences.Empty;
         }
         else if (!isWhitespace(c)) {
             this.state = State.BeforeAttributeName;
@@ -486,17 +640,66 @@ export default class Tokenizer {
             this.startEntity();
         }
     }
+    /**
+     * Distinguish between CDATA, declarations, HTML comments, and HTML bogus
+     * comments after `<!`.
+     *
+     * In HTML mode, only real comments and doctypes stay on declaration paths;
+     * everything else becomes a bogus comment terminated by the next `>`.
+     * @param c Current character code point.
+     */
     stateBeforeDeclaration(c) {
         if (c === CharCodes.OpeningSquareBracket) {
             this.state = State.CDATASequence;
             this.sequenceIndex = 0;
         }
-        else {
+        else if (this.xmlMode) {
             this.state =
                 c === CharCodes.Dash
                     ? State.BeforeComment
                     : State.InDeclaration;
         }
+        else if ((c | 0x20) === Sequences.Doctype[0]) {
+            this.state = State.DeclarationSequence;
+            this.currentSequence = Sequences.Doctype;
+            this.sequenceIndex = 1;
+        }
+        else if (c === CharCodes.Gt) {
+            this.cbs.oncomment(this.sectionStart, this.index, 0);
+            this.state = State.Text;
+            this.sectionStart = this.index + 1;
+        }
+        else if (c === CharCodes.Dash) {
+            this.state = State.BeforeComment;
+        }
+        else {
+            this.state = State.InSpecialComment;
+        }
+    }
+    /**
+     * Continue matching `doctype` after `<!d`.
+     *
+     * A full `doctype` match stays on the declaration path; any other name falls
+     * back to an HTML bogus comment, which matches browser behavior for
+     * non-doctype `<!...>` constructs.
+     * @param c Current character code point.
+     */
+    stateDeclarationSequence(c) {
+        if (this.sequenceIndex === this.currentSequence.length) {
+            this.state = State.InDeclaration;
+            this.stateInDeclaration(c);
+        }
+        else if ((c | 0x20) === this.currentSequence[this.sequenceIndex]) {
+            this.sequenceIndex += 1;
+        }
+        else if (c === CharCodes.Gt) {
+            this.cbs.oncomment(this.sectionStart, this.index, 0);
+            this.state = State.Text;
+            this.sectionStart = this.index + 1;
+        }
+        else {
+            this.state = State.InSpecialComment;
+        }
     }
     stateInDeclaration(c) {
         if (c === CharCodes.Gt || this.fastForwardTo(CharCodes.Gt)) {
@@ -505,64 +708,53 @@ export default class Tokenizer {
             this.sectionStart = this.index + 1;
         }
     }
+    /**
+     * XML processing instructions (`<?...?>`).
+     *
+     * In HTML mode `<?` is routed to `InSpecialComment` instead, so this
+     * state is only reachable in XML mode.
+     * @param c Current character code point.
+     */
     stateInProcessingInstruction(c) {
-        if (c === CharCodes.Gt || this.fastForwardTo(CharCodes.Gt)) {
-            this.cbs.onprocessinginstruction(this.sectionStart, this.index);
+        if (c === CharCodes.Questionmark) {
+            // Remember that we just consumed `?`, so the next `>` closes the PI.
+            this.sequenceIndex = 1;
+        }
+        else if (c === CharCodes.Gt && this.sequenceIndex === 1) {
+            this.cbs.onprocessinginstruction(this.sectionStart, this.index - 1);
+            this.sequenceIndex = 0;
             this.state = State.Text;
             this.sectionStart = this.index + 1;
         }
+        else {
+            // Keep scanning for the next `?`, which can start a closing `?>`.
+            this.sequenceIndex = Number(this.fastForwardTo(CharCodes.Questionmark));
+        }
     }
     stateBeforeComment(c) {
         if (c === CharCodes.Dash) {
             this.state = State.InCommentLike;
             this.currentSequence = Sequences.CommentEnd;
-            // Allow short comments (eg. <!-->)
-            this.sequenceIndex = 2;
+            this.sequenceIndex = 0;
             this.sectionStart = this.index + 1;
         }
-        else {
+        else if (this.xmlMode) {
             this.state = State.InDeclaration;
         }
-    }
-    stateInSpecialComment(c) {
-        if (c === CharCodes.Gt || this.fastForwardTo(CharCodes.Gt)) {
+        else if (c === CharCodes.Gt) {
             this.cbs.oncomment(this.sectionStart, this.index, 0);
             this.state = State.Text;
             this.sectionStart = this.index + 1;
         }
-    }
-    stateBeforeSpecialS(c) {
-        const lower = c | 0x20;
-        if (lower === Sequences.ScriptEnd[3]) {
-            this.startSpecial(Sequences.ScriptEnd, 4);
-        }
-        else if (lower === Sequences.StyleEnd[3]) {
-            this.startSpecial(Sequences.StyleEnd, 4);
-        }
         else {
-            this.state = State.InTagName;
-            this.stateInTagName(c); // Consume the token again
+            this.state = State.InSpecialComment;
         }
     }
-    stateBeforeSpecialT(c) {
-        const lower = c | 0x20;
-        switch (lower) {
-            case Sequences.TitleEnd[3]: {
-                this.startSpecial(Sequences.TitleEnd, 4);
-                break;
-            }
-            case Sequences.TextareaEnd[3]: {
-                this.startSpecial(Sequences.TextareaEnd, 4);
-                break;
-            }
-            case Sequences.XmpEnd[3]: {
-                this.startSpecial(Sequences.XmpEnd, 4);
-                break;
-            }
-            default: {
-                this.state = State.InTagName;
-                this.stateInTagName(c); // Consume the token again
-            }
+    stateInSpecialComment(c) {
+        if (c === CharCodes.Gt || this.fastForwardTo(CharCodes.Gt)) {
+            this.cbs.oncomment(this.sectionStart, this.index, 0);
+            this.state = State.Text;
+            this.sectionStart = this.index + 1;
         }
     }
     startEntity() {
@@ -604,6 +796,7 @@ export default class Tokenizer {
         // If we are inside of text or attributes, emit what we already have.
         if (this.running && this.sectionStart !== this.index) {
             if (this.state === State.Text ||
+                this.state === State.InPlainText ||
                 (this.state === State.InSpecialTag && this.sequenceIndex === 0)) {
                 this.cbs.ontext(this.sectionStart, this.index);
                 this.sectionStart = this.index;
@@ -632,6 +825,11 @@ export default class Tokenizer {
                     this.stateText(c);
                     break;
                 }
+                case State.InPlainText: {
+                    // Skip to end of buffer; cleanup() emits the text.
+                    this.index = this.buffer.length + this.offset - 1;
+                    break;
+                }
                 case State.SpecialStartSequence: {
                     this.stateSpecialStartSequence(c);
                     break;
@@ -644,6 +842,10 @@ export default class Tokenizer {
                     this.stateCDATASequence(c);
                     break;
                 }
+                case State.DeclarationSequence: {
+                    this.stateDeclarationSequence(c);
+                    break;
+                }
                 case State.InAttributeValueDq: {
                     this.stateInAttributeValueDoubleQuotes(c);
                     break;
@@ -696,14 +898,6 @@ export default class Tokenizer {
                     this.stateAfterClosingTagName(c);
                     break;
                 }
-                case State.BeforeSpecialS: {
-                    this.stateBeforeSpecialS(c);
-                    break;
-                }
-                case State.BeforeSpecialT: {
-                    this.stateBeforeSpecialT(c);
-                    break;
-                }
                 case State.InAttributeValueNq: {
                     this.stateInAttributeValueNoQuotes(c);
                     break;
@@ -745,38 +939,99 @@ export default class Tokenizer {
         this.handleTrailingData();
         this.cbs.onend();
     }
+    handleTrailingCommentLikeData(endIndex) {
+        if (this.state !== State.InCommentLike) {
+            return false;
+        }
+        if (this.currentSequence === Sequences.CdataEnd) {
+            if (this.xmlMode) {
+                if (this.sectionStart < endIndex) {
+                    this.cbs.oncdata(this.sectionStart, endIndex, 0);
+                }
+            }
+            else {
+                /* In HTML mode, unclosed CDATA is a bogus comment. */
+                const cdataStart = this.sectionStart - Sequences.Cdata.length - 1;
+                this.cbs.oncomment(cdataStart, endIndex, 0);
+            }
+        }
+        else {
+            const offset = this.xmlMode
+                ? 0
+                : Math.min(this.sequenceIndex, Sequences.CommentEnd.length - 1);
+            this.cbs.oncomment(this.sectionStart, endIndex, offset);
+        }
+        return true;
+    }
+    handleTrailingMarkupDeclaration(endIndex) {
+        if (this.xmlMode) {
+            switch (this.state) {
+                case State.InSpecialComment:
+                case State.BeforeComment:
+                case State.CDATASequence:
+                case State.DeclarationSequence:
+                case State.InDeclaration: {
+                    this.cbs.ontext(this.sectionStart, endIndex);
+                    return true;
+                }
+                default: {
+                    return false;
+                }
+            }
+        }
+        switch (this.state) {
+            case State.BeforeDeclaration:
+            case State.InSpecialComment:
+            case State.BeforeComment:
+            case State.CDATASequence: {
+                this.cbs.oncomment(this.sectionStart, endIndex, 0);
+                return true;
+            }
+            case State.DeclarationSequence: {
+                if (this.sequenceIndex !== Sequences.Doctype.length) {
+                    this.cbs.oncomment(this.sectionStart, endIndex, 0);
+                }
+                return true;
+            }
+            case State.InDeclaration: {
+                return true;
+            }
+            default: {
+                return false;
+            }
+        }
+    }
     /** Handle any trailing data. */
     handleTrailingData() {
         const endIndex = this.buffer.length + this.offset;
+        if (this.handleTrailingCommentLikeData(endIndex) ||
+            this.handleTrailingMarkupDeclaration(endIndex)) {
+            return;
+        }
         // If there is no remaining data, we are done.
         if (this.sectionStart >= endIndex) {
             return;
         }
-        if (this.state === State.InCommentLike) {
-            if (this.currentSequence === Sequences.CdataEnd) {
-                this.cbs.oncdata(this.sectionStart, endIndex, 0);
+        switch (this.state) {
+            case State.InTagName:
+            case State.BeforeAttributeName:
+            case State.BeforeAttributeValue:
+            case State.AfterAttributeName:
+            case State.InAttributeName:
+            case State.InAttributeValueSq:
+            case State.InAttributeValueDq:
+            case State.InAttributeValueNq:
+            case State.InClosingTagName: {
+                /*
+                 * If we are currently in an opening or closing tag, us not calling the
+                 * respective callback signals that the tag should be ignored.
+                 */
+                break;
             }
-            else {
-                this.cbs.oncomment(this.sectionStart, endIndex, 0);
+            default: {
+                this.cbs.ontext(this.sectionStart, endIndex);
             }
         }
-        else if (this.state === State.InTagName ||
-            this.state === State.BeforeAttributeName ||
-            this.state === State.BeforeAttributeValue ||
-            this.state === State.AfterAttributeName ||
-            this.state === State.InAttributeName ||
-            this.state === State.InAttributeValueSq ||
-            this.state === State.InAttributeValueDq ||
-            this.state === State.InAttributeValueNq ||
-            this.state === State.InClosingTagName) {
-            /*
-             * If we are currently in an opening or closing tag, us not calling the
-             * respective callback signals that the tag should be ignored.
-             */
-        }
-        else {
-            this.cbs.ontext(this.sectionStart, endIndex);
-        }
     }
     emitCodePoint(cp, consumed) {
         if (this.baseState !== State.Text &&