When case conversion changes the length of the text, ensure only

valid memory accessed and that the converted text is selected.

2 files changed, 51 insertions, 6 deletions

diff --git a/src/Editor.cxx b/src/Editor.cxx
index f778ea933..f925beab7 100644
--- a/src/Editor.cxx
+++ b/src/Editor.cxx
@@ -5000,18 +5000,28 @@ void Editor::ChangeCaseOfSelection(int caseMapping) {
 				size_t firstDifference = 0;
 				while (sMapped[firstDifference] == sText[firstDifference])
 					firstDifference++;
-				size_t lastDifference = sMapped.size() - 1;
-				while (sMapped[lastDifference] == sText[lastDifference])
-					lastDifference--;
-				size_t endSame = sMapped.size() - 1 - lastDifference;
+				size_t lastDifferenceText = sText.size() - 1;
+				size_t lastDifferenceMapped = sMapped.size() - 1;
+				while (sMapped[lastDifferenceMapped] == sText[lastDifferenceText]) {
+					lastDifferenceText--;
+					lastDifferenceMapped--;
+				}
+				size_t endDifferenceText = sText.size() - 1 - lastDifferenceText;
 				pdoc->DeleteChars(
 					static_cast<int>(currentNoVS.Start().Position() + firstDifference),
-					static_cast<int>(rangeBytes - firstDifference - endSame));
+					static_cast<int>(rangeBytes - firstDifference - endDifferenceText));
 				pdoc->InsertString(
 					static_cast<int>(currentNoVS.Start().Position() + firstDifference),
 					sMapped.c_str() + firstDifference,
-					static_cast<int>(lastDifference - firstDifference + 1));
+					static_cast<int>(lastDifferenceMapped - firstDifference + 1));
 				// Automatic movement changes selection so reset to exactly the same as it was.
+				int diffSizes = sMapped.size() - sText.size();
+				if (diffSizes != 0) {
+					if (current.anchor > current.caret)
+						current.anchor.Add(diffSizes);
+					else
+						current.caret.Add(diffSizes);
+				}
 				sel.Range(r) = current;
 			}
 		}
diff --git a/test/simpleTests.py b/test/simpleTests.py
index a101857f8..e9aef81e2 100644
--- a/test/simpleTests.py
+++ b/test/simpleTests.py
@@ -1413,6 +1413,41 @@ class TestCaseMapping(unittest.TestCase):
 		self.assertEquals(self.ed.Length, 1)
 		self.assertEquals(self.ed.Contents(), r)
 
+	def testUTFGrows(self):
+		# FIXME: change to using Unicode compliant case conversion on Windows
+		if sys.platform != "win32":
+			# This crashed at one point in debug builds due to looking past end of shorter string
+			self.ed.SetCodePage(65001)
+			# ﬖ is a single character ligature taking 3 bytes in UTF8: EF AC 96 
+			t = 'ﬖﬖ'.encode("UTF-8")
+			self.ed.SetText(len(t), t)
+			self.assertEquals(self.ed.Length, 6)
+			self.ed.SetSel(0,self.ed.Length)
+			self.ed.UpperCase()
+			# To convert to upper case the ligature is separated into վ and ն then uppercased to Վ and Ն
+			# each of which takes 2 bytes in UTF-8: D5 8E D5 86
+			r = 'ՎՆՎՆ'.encode("UTF-8")
+			self.assertEquals(self.ed.Length, 8)
+			self.assertEquals(self.ed.Contents(), r)
+			self.assertEquals(self.ed.SelectionEnd, self.ed.Length)
+
+	def testUTFShrinks(self):
+		# FIXME: change to using Unicode compliant case conversion on Windows
+		if sys.platform != "win32":
+			self.ed.SetCodePage(65001)
+			# fi is a single character ligature taking 3 bytes in UTF8: EF AC 81
+			t = 'fifi'.encode("UTF-8")
+			self.ed.SetText(len(t), t)
+			self.assertEquals(self.ed.Length, 6)
+			self.ed.SetSel(0,self.ed.Length)
+			self.ed.UpperCase()
+			# To convert to upper case the ligature is separated into f and i then uppercased to F and I
+			# each of which takes 1 byte in UTF-8: 46 49
+			r = 'FIFI'.encode("UTF-8")
+			self.assertEquals(self.ed.Length, 4)
+			self.assertEquals(self.ed.Contents(), r)
+			self.assertEquals(self.ed.SelectionEnd, self.ed.Length)
+
 class TestCaseInsensitiveSearch(unittest.TestCase):
 	def setUp(self):
 		self.xite = Xite.xiteFrame