[String] Invalidate breadcrumbs on mutation.

Author: milseman

stdlib/public/core/StringBreadcrumbs.swift

@@ -57,7 +57,7 @@ internal final class _StringBreadcrumbs {
     self.utf16Length = i
     _sanityCheck(self.crumbs.count == 1 + (self.utf16Length / stride))
 
-    _invariantCheck()
+    _invariantCheck(for: str)
   }
 }
 
@@ -97,10 +97,13 @@ extension _StringBreadcrumbs {
   }
 
   #if !INTERNAL_CHECKS_ENABLED
-  @nonobjc @inline(__always) internal func _invariantCheck() {}
+  @nonobjc @inline(__always) internal func _invariantCheck(for str: String) {}
   #else
   @nonobjc @inline(never) @_effects(releasenone)
-  internal func _invariantCheck() {
+  internal func _invariantCheck(for str: String) {
+    _sanityCheck(self.utf16Length == str.utf16._distance(
+      from: str.startIndex, to: str.endIndex),
+      "Stale breadcrumbs")
   }
   #endif // INTERNAL_CHECKS_ENABLED
 }

stdlib/public/core/StringObject.swift

@@ -1272,7 +1272,6 @@ extension _StringObject {
     } else {
       _sanityCheck(isLarge)
       _sanityCheck(largeCount == count)
-      _countAndFlags._invariantCheck()
       if providesFastUTF8 && largeFastIsNative {
         _sanityCheck(!isSmall)
         _sanityCheck(!largeIsCocoa)
@@ -1284,7 +1283,6 @@ extension _StringObject {
           _sanityCheck(hasNativeStorage)
           _sanityCheck(hasObjCBridgeableObject)
           _sanityCheck(nativeStorage.count == self.count)
-          nativeStorage._invariantCheck()
         }
       }
       if largeIsCocoa {

stdlib/public/core/StringStorage.swift

@@ -350,7 +350,7 @@ extension _StringStorage {
       _sanityCheck(_allASCII(self.codeUnits))
     }
     if let crumbs = _breadcrumbsAddress.pointee {
-      crumbs._invariantCheck()
+      crumbs._invariantCheck(for: self.asString)
     }
   }
   #endif // INTERNAL_CHECKS_ENABLED
@@ -369,6 +369,9 @@ extension _StringStorage {
       count: newCount, isASCII: newIsASCII)
 #endif
     self.terminator.pointee = 0
+
+    // TODO(String performance): Consider updating breadcrumbs when feasible
+    self._breadcrumbsAddress.pointee = nil
     _invariantCheck()
   }
 
@@ -552,7 +555,7 @@ extension _SharedStringStorage {
   #else
   internal func _invariantCheck() {
     if let crumbs = _breadcrumbs {
-      crumbs._invariantCheck()
+      crumbs._invariantCheck(for: self.asString)
     }
     _countAndFlags._invariantCheck()
   }

validation-test/stdlib/StringBreadcrumbs.swift

@@ -15,6 +15,8 @@ let largeUnicode = "abéÏ012345678901234567890𓀀"
 let emoji = "😀😃🤢🤮👩🏿‍🎤🧛🏻‍♂️🧛🏻‍♂️👩‍👩‍👦‍👦"
 let chinese = "Swift 是面向 Apple 平台的编程语言，功能强大且直观易用，而本次更新对其进行了全面优化。"
 
+let nonBMP = String(repeating: "𓀀", count: 1 + (64 / 2))
+
 let largeString: String = {
   var result = ""
   result += smallASCII
@@ -74,7 +76,6 @@ StringBreadcrumbsTests.test("largeString") {
 // Test various boundary conditions with surrogate pairs aligning or not
 // aligning
 StringBreadcrumbsTests.test("surrogates-heavy") {
-  let nonBMP = String(repeating: "𓀀", count: 1 + (64 / 2))
 
   // Mis-align the hieroglyphics by 1,2,3 UTF-8 and UTF-16 code units
   validateBreadcrumbs(nonBMP)
@@ -86,6 +87,18 @@ StringBreadcrumbsTests.test("surrogates-heavy") {
   validateBreadcrumbs("aé" + nonBMP)
 }
 
-// TODO(String testing): test breadcrumb validity after mutation
+// Test bread-crumb invalidation
+StringBreadcrumbsTests.test("stale breadcrumbs") {
+  var str = nonBMP + "𓀀"
+  let oldLen = str.utf16.count
+  str.removeLast()
+  expectEqual(oldLen - 2, str.utf16.count)
+  str += "a"
+  expectEqual(oldLen - 1, str.utf16.count)
+  str += "𓀀"
+  expectEqual(oldLen + 1, str.utf16.count)
+}
+
+
 
 runAllTests()