[parser] Fix chunked utf8 stream handling.

BUG=v8:6377

Change-Id: I5bdd41bdda83d7efe4b37d24d44e2e8c2339a30a
Reviewed-on: https://chromium-review.googlesource.com/500168
Commit-Queue: Wiktor Garbacz <wiktorg@google.com>
Reviewed-by: Daniel Vogelheim <vogelheim@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45204}

src/parsing/scanner-character-streams.cc

@@ -387,8 +387,10 @@ void Utf8ExternalStreamingStream::SearchPosition(size_t position) {
     //  checking whether the # bytes in a chunk are equal to the # chars, and if
     //  so avoid the expensive SkipToPosition.)
     bool ascii_only_chunk =
+        chunks_[chunk_no].start.incomplete_char ==
+            unibrow::Utf8::Utf8IncrementalBuffer(0) &&
         (chunks_[chunk_no + 1].start.bytes - chunks_[chunk_no].start.bytes) ==
-        (chunks_[chunk_no + 1].start.chars - chunks_[chunk_no].start.chars);
+            (chunks_[chunk_no + 1].start.chars - chunks_[chunk_no].start.chars);
     if (ascii_only_chunk) {
       size_t skip = position - chunks_[chunk_no].start.chars;
       current_ = {chunk_no,

test/cctest/parsing/test-scanner-streams.cc

@@ -22,6 +22,13 @@ class ChunkSource : public v8::ScriptCompiler::ExternalSourceStream {
       chunks++;
     } while (chunks_.back().len > 0);
   }
+  explicit ChunkSource(const char* chunks) : current_(0) {
+    do {
+      chunks_.push_back(
+          {reinterpret_cast<const uint8_t*>(chunks), strlen(chunks)});
+      chunks += strlen(chunks) + 1;
+    } while (chunks_.back().len > 0);
+  }
   ChunkSource(const uint8_t* data, size_t len, bool extra_chunky)
       : current_(0) {
     // If extra_chunky, we'll use increasingly large chunk sizes.
@@ -450,3 +457,42 @@ TEST(Regress651333) {
     CHECK_EQ(i::Utf16CharacterStream::kEndOfInput, stream->Advance());
   }
 }
+
+TEST(Regress6377) {
+  const char* cases[] = {
+      "\xf0\x90\0"  // first chunk - start of 4-byte seq
+      "\x80\x80"    // second chunk - end of 4-byte seq
+      "a\0",        // and an 'a'
+
+      "\xe0\xbf\0"  // first chunk - start of 3-byte seq
+      "\xbf"        // second chunk - one-byte end of 3-byte seq
+      "a\0",        // and an 'a'
+
+      "\xc3\0"  // first chunk - start of 2-byte seq
+      "\xbf"    // second chunk - end of 2-byte seq
+      "a\0",    // and an 'a'
+
+      "\xf0\x90\x80\0"  // first chunk - start of 4-byte seq
+      "\x80"            // second chunk - one-byte end of 4-byte seq
+      "a\xc3\0"         // and an 'a' + start of 2-byte seq
+      "\xbf\0",         // third chunk - end of 2-byte seq
+  };
+  const std::vector<std::vector<uint16_t>> unicode = {
+      {0xd800, 0xdc00, 97}, {0xfff, 97}, {0xff, 97}, {0xd800, 0xdc00, 97, 0xff},
+  };
+  CHECK_EQ(unicode.size(), sizeof(cases) / sizeof(cases[0]));
+  for (size_t c = 0; c < unicode.size(); ++c) {
+    ChunkSource chunk_source(cases[c]);
+    std::unique_ptr<i::Utf16CharacterStream> stream(i::ScannerStream::For(
+        &chunk_source, v8::ScriptCompiler::StreamedSource::UTF8, nullptr));
+    for (size_t i = 0; i < unicode[c].size(); i++) {
+      CHECK_EQ(unicode[c][i], stream->Advance());
+    }
+    CHECK_EQ(i::Utf16CharacterStream::kEndOfInput, stream->Advance());
+    stream->Seek(0);
+    for (size_t i = 0; i < unicode[c].size(); i++) {
+      CHECK_EQ(unicode[c][i], stream->Advance());
+    }
+    CHECK_EQ(i::Utf16CharacterStream::kEndOfInput, stream->Advance());
+  }
+}