Report the correct error position in some malformed constructs

Merges #1253

Fixes #1251

Authored bt @csaboka, but I rewrote the test file to fix the diff up.

src/main/java/org/jsoup/parser/CharacterReader.java

@@ -16,7 +16,7 @@ public final class CharacterReader {
     static final char EOF = (char) -1;
     private static final int maxStringCacheLen = 12;
     static final int maxBufferLen = 1024 * 32; // visible for testing
-    private static final int readAheadLimit = (int) (maxBufferLen * 0.75);
+    static final int readAheadLimit = (int) (maxBufferLen * 0.75); // visible for testing
     private static final int minReadAheadLen = 1024; // the minimum mark length supported. No HTML entities can be larger than this.
 
     private final char[] charBuf;

src/main/java/org/jsoup/parser/TokeniserState.java

@@ -161,8 +161,8 @@ void read(Tokeniser t, CharacterReader r) {
                     t.transition(SelfClosingStartTag);
                     break;
                 case '<': // NOTE: out of spec, but clear author intent
-                    t.error(this);
                     r.unconsume();
+                    t.error(this);
                     // intended fall through to next >
                 case '>':
                     t.emitTagPending();
@@ -572,17 +572,17 @@ void read(Tokeniser t, CharacterReader r) {
                     t.transition(SelfClosingStartTag);
                     break;
                 case '<': // NOTE: out of spec, but clear (spec has this as a part of the attribute name)
-                    t.error(this);
                     r.unconsume();
+                    t.error(this);
                     // intended fall through as if >
                 case '>':
                     t.emitTagPending();
                     t.transition(Data);
                     break;
                 case nullChar:
+                    r.unconsume();
                     t.error(this);
                     t.tagPending.newAttribute();
-                    r.unconsume();
                     t.transition(AttributeName);
                     break;
                 case eof:
@@ -880,8 +880,8 @@ void read(Tokeniser t, CharacterReader r) {
                     t.transition(Data);
                     break;
                 default:
-                    t.error(this);
                     r.unconsume();
+                    t.error(this);
                     t.transition(BeforeAttributeName);
             }
 
@@ -901,8 +901,8 @@ void read(Tokeniser t, CharacterReader r) {
                     t.transition(Data);
                     break;
                 default:
-                    t.error(this);
                     r.unconsume();
+                    t.error(this);
                     t.transition(BeforeAttributeName);
             }
         }

src/test/java/org/jsoup/parser/ParserIT.java

@@ -0,0 +1,23 @@
+package org.jsoup.parser;
+
+import org.junit.Test;
+
+/**
+ * Longer running Parser tests.
+ */
+
+public class ParserIT {
+    @Test
+    public void testIssue1251() {
+        // https://github.com/jhy/jsoup/issues/1251
+        StringBuilder str = new StringBuilder("<a href=\"\"ca");
+        for (int countSpaces = 0; countSpaces < 100000; countSpaces++) {
+            try {
+                Parser.htmlParser().setTrackErrors(1).parseInput(str.toString(), "");
+            } catch (Exception e) {
+                throw new AssertionError("failed at length " + str.length(), e);
+            }
+            str.insert(countSpaces, ' ');
+        }
+    }
+}

src/test/java/org/jsoup/parser/TokeniserStateTest.java

@@ -206,4 +206,47 @@ public void testPublicAndSystemIdentifiersWithWhitespace() {
         Document doc = Jsoup.parse(html);
         assertEquals("<p></p><p></p><div id=\"one\"><span>Two</span></div>", TextUtil.stripNewlines(doc.body().html()));
     }
+
+    @Test
+    public void testUnconsumeAtBufferBoundary() {
+        String triggeringSnippet = "<a href=\"\"foo";
+        char[] padding = new char[CharacterReader.readAheadLimit - triggeringSnippet.length() + 2]; // The "foo" part must be just at the limit.
+        Arrays.fill(padding, ' ');
+        String paddedSnippet = new StringBuilder().append(padding).append(triggeringSnippet).toString();
+        ParseErrorList errorList = ParseErrorList.tracking(1);
+
+        Parser.parseFragment(paddedSnippet, null, "", errorList);
+
+        assertEquals(CharacterReader.readAheadLimit - 1, errorList.get(0).getPosition());
+    }
+
+    @Test
+    public void testOpeningAngleBracketInsteadOfAttribute() {
+        String triggeringSnippet = "<html <";
+        ParseErrorList errorList = ParseErrorList.tracking(1);
+
+        Parser.parseFragment(triggeringSnippet, null, "", errorList);
+
+        assertEquals(6, errorList.get(0).getPosition());
+    }
+
+    @Test
+    public void testMalformedSelfClosingTag() {
+        String triggeringSnippet = "<html /ouch";
+        ParseErrorList errorList = ParseErrorList.tracking(1);
+
+        Parser.parseFragment(triggeringSnippet, null, "", errorList);
+
+        assertEquals(7, errorList.get(0).getPosition());
+    }
+
+    @Test
+    public void testOpeningAngleBracketInTagName() {
+        String triggeringSnippet = "<html<";
+        ParseErrorList errorList = ParseErrorList.tracking(1);
+
+        Parser.parseFragment(triggeringSnippet, null, "", errorList);
+
+        assertEquals(5, errorList.get(0).getPosition());
+    }
 }