Bug 1746374 - Use the number of UTF-16 code units as column number. r=iain

Differential Revision: https://phabricator.services.mozilla.com/D185372

dom/media/webaudio/AudioWorkletNode.cpp

@@ -50,7 +50,7 @@ struct ProcessorErrorDetails {
   ProcessorErrorDetails() : mLineno(0), mColno(0) {}
   // Line number (1-origin).
   unsigned mLineno;
-  // Column number (1-origin).
+  // Column number in UTF-16 code units (1-origin).
   unsigned mColno;
   nsString mFilename;
   nsString mMessage;

dom/workers/WorkerError.h

@@ -22,8 +22,10 @@ class WorkerErrorBase {
  public:
   nsString mMessage;
   nsString mFilename;
-  uint32_t mLineNumber;    // 1-origin.
-  uint32_t mColumnNumber;  // 1-origin
+  // Line number (1-origin).
+  uint32_t mLineNumber;
+  // Column number in UTF-16 code units (1-origin).
+  uint32_t mColumnNumber;
   uint32_t mErrorNumber;
 
   WorkerErrorBase() : mLineNumber(0), mColumnNumber(0), mErrorNumber(0) {}

js/public/CompileOptions.h

@@ -455,7 +455,7 @@ class JS_PUBLIC_API ReadOnlyCompileOptions : public TransitiveCompileOptions {
 
   // Line number of the first character (1-origin).
   unsigned lineno = 1;
-  // Column number of the first character (0-origin).
+  // Column number of the first character in UTF-16 code units (0-origin).
   unsigned column = 0;
 
   // The offset within the ScriptSource's full uncompressed text of the first

js/public/ErrorReport.h

@@ -120,7 +120,7 @@ class JSErrorBase {
   // Source line number (1-origin).
   unsigned lineno;
 
-  // Column number in line (1-origin).
+  // Column number in line in UTF-16 code units (1-origin).
   unsigned column;
 
   // the error number, e.g. see js/public/friend/ErrorNumbers.msg.

js/public/JitCodeAPI.h

@@ -39,7 +39,7 @@ struct JitCodeSourceInfo {
 
   // Line number (1-origin).
   uint32_t lineno = 0;
-  // Column number (0-origin).
+  // Column number in UTF-16 code units (0-origin).
   uint32_t colno = 0;
 };
 

js/public/UbiNode.h

@@ -248,7 +248,7 @@ class BaseStackFrame {
   // Get this frame's line number (1-origin).
   virtual uint32_t line() const = 0;
 
-  // Get this frame's column number (1-origin).
+  // Get this frame's column number in UTF-16 code units (1-origin).
   virtual uint32_t column() const = 0;
 
   // Get this frame's source name. Never null.

js/src/builtin/ModuleObject.h

@@ -69,8 +69,12 @@ class ImportEntry {
   const HeapPtr<ModuleRequestObject*> moduleRequest_;
   const HeapPtr<JSAtom*> importName_;
   const HeapPtr<JSAtom*> localName_;
-  const uint32_t lineNumber_;    // 1-origin
-  const uint32_t columnNumber_;  // 0-origin
+
+  // Line number (1-origin).
+  const uint32_t lineNumber_;
+
+  // Column number in UTF-16 code units (0-origin).
+  const uint32_t columnNumber_;
 
  public:
   ImportEntry(Handle<ModuleRequestObject*> moduleRequest,
@@ -93,8 +97,12 @@ class ExportEntry {
   const HeapPtr<ModuleRequestObject*> moduleRequest_;
   const HeapPtr<JSAtom*> importName_;
   const HeapPtr<JSAtom*> localName_;
-  const uint32_t lineNumber_;    // 1-origin.
-  const uint32_t columnNumber_;  // 0-origin
+
+  // Line number (1-origin).
+  const uint32_t lineNumber_;
+
+  // Column number in UTF-16 code units (0-origin).
+  const uint32_t columnNumber_;
 
  public:
   ExportEntry(Handle<JSAtom*> maybeExportName,
@@ -115,8 +123,12 @@ using ExportEntryVector = GCVector<ExportEntry, 0, SystemAllocPolicy>;
 
 class RequestedModule {
   const HeapPtr<ModuleRequestObject*> moduleRequest_;
-  const uint32_t lineNumber_;    // 1-origin
-  const uint32_t columnNumber_;  // 0-origin
+
+  // Line number (1-origin).
+  const uint32_t lineNumber_;
+
+  // Column number in UTF-16 code units (0-origin).
+  const uint32_t columnNumber_;
 
  public:
   RequestedModule(Handle<ModuleRequestObject*> moduleRequest,

js/src/debugger/Script.cpp

@@ -1136,7 +1136,7 @@ class FlowGraphSummary {
     // Line number (1-origin).
     size_t lineno_;
 
-    // Column number (0-origin).
+    // Column number in UTF-16 code units (0-origin).
     size_t column_;
   };
 

js/src/doc/Debugger/Debugger.Object.md

@@ -166,8 +166,8 @@ If the referent is an Error object, this is the 1-origin line number at which
 the referent was created; `undefined`  otherwise.
 
 ### `errorColumnNumber`
-If the referent is an Error object, this is the 1-origin column number at which
-the referent was created; `undefined`  otherwise.
+If the referent is an Error object, this is the 1-origin column number in
+UTF-16 code units at which the referent was created; `undefined`  otherwise.
 
 ### `isBoundFunction`
 If the referent is a debuggee function, returns `true` if the referent is a

js/src/doc/Debugger/Debugger.Source.md

@@ -111,9 +111,10 @@ source within the file or URL it was loaded from. This is normally `1`, but
 may have another value if the source is part of an HTML document.
 
 ### `startColumn`
-**If the instance refers to JavaScript source**, the 0-origin start column of
-the source within the file or URL it was loaded from. This is normally `0`, but
-may have another value if the source is part of an HTML document.
+**If the instance refers to JavaScript source**, the 0-origin start column in
+UTF-16 code units of the source within the file or URL it was loaded from. This
+is normally `0`, but may have another value if the source is part of an HTML
+document.
 
 ### `id`
 **If the instance refers to JavaScript source**, an int32 counter that identifies

js/src/frontend/BytecodeSection.h

@@ -340,7 +340,7 @@ class BytecodeSection {
   // we can get undefined behavior.
   uint32_t currentLine_;
 
-  // 0-origin column index on currentLine_ of last
+  // 0-origin column index in UTF-16 code units on currentLine_ of last
   // SrcNoteType::ColSpan-annotated opcode.
   //
   // WARNING: If this becomes out of sync with already-emitted srcnotes,

js/src/frontend/Stencil.h

@@ -611,7 +611,7 @@ class StencilModuleEntry {
   // Line number (1-origin).
   uint32_t lineno = 0;
 
-  // Column number (0-origin).
+  // Column number in UTF-16 code units (0-origin).
   uint32_t column = 0;
 
  private:

js/src/frontend/TokenStream.cpp

@@ -597,8 +597,23 @@ uint32_t TokenStreamAnyChars::computePartialColumn(
     const SourceUnits<Unit>& sourceUnits) const {
   lineToken.assertConsistentOffset(offset);
 
-  const uint32_t line = lineNumber(lineToken);
   const uint32_t start = srcCoords.lineStart(lineToken);
+  const uint32_t offsetInLine = offset - start;
+
+  if constexpr (std::is_same_v<Unit, char16_t>) {
+    // Column number is in UTF-16 code units.
+    return offsetInLine;
+  }
+
+  return computePartialColumnForUTF8(lineToken, offset, start, offsetInLine,
+                                     sourceUnits);
+}
+
+template <typename Unit>
+uint32_t TokenStreamAnyChars::computePartialColumnForUTF8(
+    const LineToken lineToken, const uint32_t offset, const uint32_t start,
+    const uint32_t offsetInLine, const SourceUnits<Unit>& sourceUnits) const {
+  const uint32_t line = lineNumber(lineToken);
 
   // Reset the previous offset/column cache for this line, if the previous
   // lookup wasn't on this line.
@@ -630,13 +645,13 @@ uint32_t TokenStreamAnyChars::computePartialColumn(
     partialOffset += offsetDelta;
 
     if (unitsType == UnitsType::GuaranteedSingleUnit) {
-      MOZ_ASSERT(unicode::CountCodePoints(begin, end) == offsetDelta,
+      MOZ_ASSERT(unicode::CountUTF16CodeUnits(begin, end) == offsetDelta,
                  "guaranteed-single-units also guarantee pointer distance "
-                 "equals code point count");
+                 "equals UTF-16 code unit count");
       partialCols += offsetDelta;
     } else {
       partialCols +=
-          AssertedCast<uint32_t>(unicode::CountCodePoints(begin, end));
+          AssertedCast<uint32_t>(unicode::CountUTF16CodeUnits(begin, end));
     }
 
     this->lastOffsetOfComputedColumn_ = partialOffset;
@@ -644,8 +659,6 @@ uint32_t TokenStreamAnyChars::computePartialColumn(
     return partialCols;
   };
 
-  const uint32_t offsetInLine = offset - start;
-
   // We won't add an entry to |longLineColumnInfo_| for lines where the maximum
   // column has offset less than this value.  The most common (non-minified)
   // long line length is likely 80ch, maybe 100ch, so we use that, rounded up to
@@ -786,16 +799,17 @@ uint32_t TokenStreamAnyChars::computePartialColumn(
       MOZ_ASSERT(chunkLimit <= limit);
 
       size_t numUnits = PointerRangeSize(begin, chunkLimit);
-      size_t numCodePoints = unicode::CountCodePoints(begin, chunkLimit);
+      size_t numUTF16CodeUnits =
+          unicode::CountUTF16CodeUnits(begin, chunkLimit);
 
       // If this chunk (which will become non-final at the end of the loop) is
       // all single-unit code points, annotate the chunk accordingly.
-      if (numUnits == numCodePoints) {
+      if (numUnits == numUTF16CodeUnits) {
         lastChunkVectorForLine_->back().guaranteeSingleUnits();
       }
 
       partialOffset += numUnits;
-      partialColumn += numCodePoints;
+      partialColumn += numUTF16CodeUnits;
 
       lastChunkVectorForLine_->infallibleEmplaceBack(
           partialColumn, UnitsType::PossiblyMultiUnit);

js/src/frontend/TokenStream.h

@@ -363,12 +363,12 @@ class SourceUnits;
  *
  * for either |Unit = Utf8Unit| or |Unit = char16_t|.
  *
- * Note that the latter quantity is *not* the same as a column number, which is
- * a count of code *points*.  Computing a column number requires the offset
- * within the line and the source units of that line (including what type |Unit|
- * is, to know how to decode them).  If you need a column number, functions in
- * |GeneralTokenStreamChars<Unit>| will consult this and source units to compute
- * it.
+ * Note that, if |Unit = Utf8Unit|, the latter quantity is *not* the same as a
+ * column number, which is a count of UTF-16 code units.  Computing a column
+ * number requires the offset within the line and the source units of that line
+ * (including what type |Unit| is, to know how to decode them).  If you need a
+ * column number, functions in |GeneralTokenStreamChars<Unit>| will consult
+ * this and source units to compute it.
  */
 class SourceCoords {
   // For a given buffer holding source code, |lineStartOffsets_| has one
@@ -515,7 +515,7 @@ enum class UnitsType : unsigned char {
 
 class ChunkInfo {
  private:
-  // Column number (0-origin).
+  // Column number in UTF-16 code units (0-origin).
   // Store everything in |unsigned char|s so everything packs.
   unsigned char column_[sizeof(uint32_t)];
   unsigned char unitsType_;
@@ -580,6 +580,7 @@ class TokenStreamAnyChars : public TokenStreamShared {
   JS::ConstUTF8CharsZ filename_;
 
   // Column number computation fields.
+  // Used only for UTF-8 case.
 
   /**
    * A map of (line number => sequence of the column numbers at
@@ -911,7 +912,7 @@ class TokenStreamAnyChars : public TokenStreamShared {
 
  private:
   /**
-   * Compute the "partial" column number in Unicode code points of the absolute
+   * Compute the "partial" column number in UTF-16 code units of the absolute
    * |offset| within source text on the line of |lineToken| (which must have
    * been computed from |offset|).
    *
@@ -944,13 +945,13 @@ class TokenStreamAnyChars : public TokenStreamShared {
    * the browser before SpiderMonkey would see it.  So the partial column of the
    * "4" in the inequality would be 16, not 19.
    *
-   * Code points are not all equal length, so counting requires *some* kind of
-   * linear-time counting from the start of the line.  This function attempts
-   * various tricks to reduce this cost.  If these optimizations succeed,
-   * repeated calls to this function on a line will pay a one-time cost linear
-   * in the length of the line, then each call pays a separate constant-time
-   * cost.  If the optimizations do not succeed, this function works in time
-   * linear in the length of the line.
+   * UTF-16 code units are not all equal length in UTF-8 source, so counting
+   * requires *some* kind of linear-time counting from the start of the line.
+   * This function attempts various tricks to reduce this cost.  If these
+   * optimizations succeed, repeated calls to this function on a line will pay
+   * a one-time cost linear in the length of the line, then each call pays a
+   * separate constant-time cost.  If the optimizations do not succeed, this
+   * function works in time linear in the length of the line.
    *
    * It's unusual for a function in *this* class to be |Unit|-templated, but
    * while this operation manages |Unit|-agnostic fields in this class and in
@@ -962,6 +963,11 @@ class TokenStreamAnyChars : public TokenStreamShared {
                                 const uint32_t offset,
                                 const SourceUnits<Unit>& sourceUnits) const;
 
+  template <typename Unit>
+  uint32_t computePartialColumnForUTF8(
+      const LineToken lineToken, const uint32_t offset, const uint32_t start,
+      const uint32_t offsetInLine, const SourceUnits<Unit>& sourceUnits) const;
+
   /**
    * Update line/column information for the start of a new line at
    * |lineStartOffset|.

js/src/irregexp/RegExpAPI.cpp

@@ -179,7 +179,7 @@ static size_t ComputeColumn(const Latin1Char* begin, const Latin1Char* end) {
 }
 
 static size_t ComputeColumn(const char16_t* begin, const char16_t* end) {
-  return unicode::CountCodePoints(begin, end);
+  return unicode::CountUTF16CodeUnits(begin, end);
 }
 
 // This function is varargs purely so it can call ReportCompileErrorLatin1.

js/src/jit/MIRGraph.h

@@ -642,7 +642,7 @@ class MBasicBlock : public TempObject, public InlineListNode<MBasicBlock> {
   // Line number (1-origin).
   unsigned lineno_;
 
-  // Column number (0-origin).
+  // Column number in UTF-16 code units (0-origin).
   unsigned columnIndex_;
 
  public:

js/src/jsapi-tests/testPrintError.cpp

@@ -96,13 +96,13 @@ bool cls_testPrintError_PrintWarning::warningSuccess = false;
 
 #define BURRITO "\xF0\x9F\x8C\xAF"
 
-BEGIN_TEST(testPrintError_UTF16CodePoints) {
+BEGIN_TEST(testPrintError_UTF16CodeUnits) {
   AutoStreamBuffer buf;
 
   static const char utf8code[] =
       "function f() {\n  var x = `\n" BURRITO "`; " BURRITO "; } f();";
 
-  CHECK(!execDontReport(utf8code, "testPrintError_UTF16CodePoints.js", 1));
+  CHECK(!execDontReport(utf8code, "testPrintError_UTF16CodeUnits.js", 1));
 
   JS::ExceptionStack exnStack(cx);
   CHECK(JS::StealPendingExceptionStack(cx, &exnStack));
@@ -112,14 +112,14 @@ BEGIN_TEST(testPrintError_UTF16CodePoints) {
   JS::PrintError(buf.stream(), builder, false);
 
   CHECK(
-      buf.contains("testPrintError_UTF16CodePoints.js:3:5 SyntaxError: illegal "
+      buf.contains("testPrintError_UTF16CodeUnits.js:3:6 SyntaxError: illegal "
                    "character U+1F32F:\n"
-                   "testPrintError_UTF16CodePoints.js:3:5 " BURRITO
-                   "`; " BURRITO "; } f();\n"
-                   "testPrintError_UTF16CodePoints.js:3:5 .....^\n"));
+                   "testPrintError_UTF16CodeUnits.js:3:6 " BURRITO "`; " BURRITO
+                   "; } f();\n"
+                   "testPrintError_UTF16CodeUnits.js:3:6 .....^\n"));
 
   return true;
 }
-END_TEST(testPrintError_UTF16CodePoints)
+END_TEST(testPrintError_UTF16CodeUnits)
 
 #undef BURRITO

js/src/util/Text.cpp

@@ -416,8 +416,8 @@ template size_t js::PutEscapedString(char* buffer, size_t bufferSize,
                                      const char16_t* chars, size_t length,
                                      uint32_t quote);
 
-size_t js::unicode::CountCodePoints(const Utf8Unit* begin,
-                                    const Utf8Unit* end) {
+size_t js::unicode::CountUTF16CodeUnits(const Utf8Unit* begin,
+                                        const Utf8Unit* end) {
   MOZ_ASSERT(begin <= end);
 
   size_t count = 0;
@@ -430,36 +430,14 @@ size_t js::unicode::CountCodePoints(const Utf8Unit* begin,
       continue;
     }
 
-#ifdef DEBUG
-    Maybe<char32_t> cp =
-#endif
-        DecodeOneUtf8CodePoint(lead, &ptr, end);
+    Maybe<char32_t> cp = DecodeOneUtf8CodePoint(lead, &ptr, end);
     MOZ_ASSERT(cp.isSome());
+    if (*cp > unicode::UTF16Max) {
+      // This uses surrogate pair.
+      count++;
+    }
   }
   MOZ_ASSERT(ptr == end, "bad code unit count in line?");
 
   return count;
 }
-
-size_t js::unicode::CountCodePoints(const char16_t* begin,
-                                    const char16_t* end) {
-  MOZ_ASSERT(begin <= end);
-
-  size_t count = 0;
-
-  const char16_t* ptr = begin;
-  while (ptr < end) {
-    count++;
-
-    if (!IsLeadSurrogate(*ptr++)) {
-      continue;
-    }
-
-    if (ptr < end && IsTrailSurrogate(*ptr)) {
-      ptr++;
-    }
-  }
-  MOZ_ASSERT(ptr == end, "should have consumed the full range");
-
-  return count;
-}

js/src/util/Text.h

@@ -355,18 +355,20 @@ bool ContainsFlag(const char* str, const char* flag);
 
 namespace unicode {
 
-/** Compute the number of code points in the valid UTF-8 range [begin, end). */
-extern size_t CountCodePoints(const mozilla::Utf8Unit* begin,
-                              const mozilla::Utf8Unit* end);
+/**
+ * Compute the number of UTF-16 code units in the valid UTF-8 range
+ * [begin, end).
+ */
+extern size_t CountUTF16CodeUnits(const mozilla::Utf8Unit* begin,
+                                  const mozilla::Utf8Unit* end);
 
 /**
- * Count the number of code points in [begin, end).
- *
- * Unlike the UTF-8 case above, consistent with legacy ECMAScript practice,
- * every sequence of 16-bit units is considered valid.  Lone surrogates are
- * treated as if they represented a code point of the same value.
+ * Count the number of UTF-16 code units in [begin, end).
  */
-extern size_t CountCodePoints(const char16_t* begin, const char16_t* end);
+inline size_t CountUTF16CodeUnits(const char16_t* begin, const char16_t* end) {
+  MOZ_ASSERT(begin <= end);
+  return end - begin;
+}
 
 }  // namespace unicode
 

js/src/vm/BytecodeUtil-inl.h

@@ -231,7 +231,7 @@ class BytecodeRangeWithPosition : private BytecodeRange {
   // Line number (1-origin).
   size_t lineno;
 
-  // Column number (0-origin).
+  // Column number in UTF-16 code units (0-origin).
   size_t column;
 
   const SrcNote* sn;

js/src/vm/ErrorObject.h

@@ -111,7 +111,7 @@ class ErrorObject : public NativeObject {
   // Line number (1-origin).
   inline uint32_t lineNumber() const;
 
-  // Column number (1-origin).
+  // Column number in UTF-16 code units (1-origin).
   inline uint32_t columnNumber() const;
 
   inline JSObject* stack() const;

js/src/vm/ErrorReporting.h

@@ -41,7 +41,7 @@ struct ErrorMetadata {
   // Line number (1-origin).
   uint32_t lineNumber;
 
-  // Column number (0-origin).
+  // Column number in UTF-16 code units (0-origin).
   uint32_t columnNumber;
 
   // If the error occurs at a particular location, context surrounding the

js/src/vm/JSScript.h

@@ -620,7 +620,8 @@ class ScriptSource {
 
   // Line number within the file where this source starts (1-origin).
   uint32_t startLine_ = 0;
-  // Column number within the file where this source starts (0-origin).
+  // Column number within the file where this source starts,
+  // in UTF-16 code units (0-origin).
   uint32_t startColumn_ = 0;
 
   // See: CompileOptions::mutedErrors.
@@ -1542,7 +1543,7 @@ class BaseScript : public gc::TenuredCellWithNonGCPointer<uint8_t> {
 
   // Line number (1-origin)
   uint32_t lineno() const { return extent_.lineno; }
-  // Column number in Unicode Code Points (0-origin)
+  // Column number in UTF-16 code units (0-origin)
   uint32_t column() const { return extent_.column; }
 
   JS::DelazificationOption delazificationMode() const {

js/src/vm/SavedFrame.h

@@ -50,7 +50,7 @@ class SavedFrame : public NativeObject {
   uint32_t getSourceId();
   // Line number (1-origin).
   uint32_t getLine();
-  // Column number (1-origin).
+  // Column number in UTF-16 code units (1-origin).
   uint32_t getColumn();
   JSAtom* getFunctionDisplayName();
   JSAtom* getAsyncCause();

js/src/vm/SavedStacks.cpp

@@ -218,7 +218,7 @@ struct MOZ_STACK_CLASS SavedFrame::Lookup {
   // Line number (1-origin).
   uint32_t line;
 
-  // Columm number (1-origin).
+  // Columm number in UTF-16 code units (1-origin).
   uint32_t column;
 
   JSAtom* functionDisplayName;

js/src/vm/SavedStacks.h

@@ -269,7 +269,7 @@ class SavedStacks {
     // Line number (1-origin).
     size_t line;
 
-    // Column number (1-origin).
+    // Column number in UTF-16 code units (1-origin).
     uint32_t column;
   };
 

js/src/vm/SharedStencil.h

@@ -219,8 +219,10 @@ struct SourceExtent {
   uint32_t toStringEnd = 0;
 
   // Line and column of |sourceStart_| position.
-  uint32_t lineno = 1;  // Line number (1-origin)
-  uint32_t column = 0;  // Column number in Unicode Code Points (0-origin)
+  // Line number (1-origin).
+  uint32_t lineno = 1;
+  // Column number in UTF-16 code units (0-origin).
+  uint32_t column = 0;
 
   FunctionKey toFunctionKey() const {
     // In eval("x=>1"), the arrow function will have a sourceStart of 0 which

js/xpconnect/src/xpcpublic.h

@@ -602,8 +602,10 @@ class ErrorBase {
   nsString mErrorMsg;
   nsString mFileName;
   uint32_t mSourceId;
-  uint32_t mLineNumber;  // 1-origin.
-  uint32_t mColumn;      // 1-origin.
+  // Line number (1-origin).
+  uint32_t mLineNumber;
+  // Column number in UTF-16 code units (1-origin).
+  uint32_t mColumn;
 
   ErrorBase() : mSourceId(0), mLineNumber(0), mColumn(0) {}