Squashed 'src/leveldb/' changes from 9094c7f..7924331

42dcc7e Merge upstream LevelDB 1.17.
e353fbc Release LevelDB 1.17
269fc6c Release LevelDB 1.16
REVERT: 9094c7f Temporarily revert to writing .sst files instead of .ldb

git-subtree-dir: src/leveldb
git-subtree-split: 79243314e40ac31d79c68e5658a1d6a64800d50b

db/filename.cc

@@ -29,19 +29,14 @@ std::string LogFileName(const std::string& name, uint64_t number) {
   return MakeFileName(name, number, "log");
 }
 
-// TableFileName returns the filenames we usually write to, while
-// SSTTableFileName returns the alternative filenames we also try to read from
-// for backward compatibility. For now, swap them around.
-// TODO: when compatibility is no longer necessary, swap them back
-// (TableFileName to use "ldb" and SSTTableFileName to use "sst").
 std::string TableFileName(const std::string& name, uint64_t number) {
   assert(number > 0);
-  return MakeFileName(name, number, "sst");
+  return MakeFileName(name, number, "ldb");
 }
 
 std::string SSTTableFileName(const std::string& name, uint64_t number) {
   assert(number > 0);
-  return MakeFileName(name, number, "ldb");
+  return MakeFileName(name, number, "sst");
 }
 
 std::string DescriptorFileName(const std::string& dbname, uint64_t number) {

db/log_reader.cc

@@ -133,7 +133,9 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch) {
 
       case kEof:
         if (in_fragmented_record) {
-          ReportCorruption(scratch->size(), "partial record without end(3)");
+          // This can be caused by the writer dying immediately after
+          // writing a physical record but before completing the next; don't
+          // treat it as a corruption, just ignore the entire logical record.
           scratch->clear();
         }
         return false;
@@ -193,13 +195,12 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result) {
           eof_ = true;
         }
         continue;
-      } else if (buffer_.size() == 0) {
-        // End of file
-        return kEof;
       } else {
-        size_t drop_size = buffer_.size();
+        // Note that if buffer_ is non-empty, we have a truncated header at the
+        // end of the file, which can be caused by the writer crashing in the
+        // middle of writing the header. Instead of considering this an error,
+        // just report EOF.
         buffer_.clear();
-        ReportCorruption(drop_size, "truncated record at end of file");
         return kEof;
       }
     }
@@ -213,8 +214,14 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result) {
     if (kHeaderSize + length > buffer_.size()) {
       size_t drop_size = buffer_.size();
       buffer_.clear();
-      ReportCorruption(drop_size, "bad record length");
-      return kBadRecord;
+      if (!eof_) {
+        ReportCorruption(drop_size, "bad record length");
+        return kBadRecord;
+      }
+      // If the end of the file has been reached without reading |length| bytes
+      // of payload, assume the writer died in the middle of writing the record.
+      // Don't report a corruption.
+      return kEof;
     }
 
     if (type == kZeroType && length == 0) {

db/log_test.cc

@@ -351,20 +351,32 @@ TEST(LogTest, BadRecordType) {
   ASSERT_EQ("OK", MatchError("unknown record type"));
 }
 
-TEST(LogTest, TruncatedTrailingRecord) {
+TEST(LogTest, TruncatedTrailingRecordIsIgnored) {
   Write("foo");
   ShrinkSize(4);   // Drop all payload as well as a header byte
   ASSERT_EQ("EOF", Read());
-  ASSERT_EQ(kHeaderSize - 1, DroppedBytes());
-  ASSERT_EQ("OK", MatchError("truncated record at end of file"));
+  // Truncated last record is ignored, not treated as an error.
+  ASSERT_EQ(0, DroppedBytes());
+  ASSERT_EQ("", ReportMessage());
 }
 
 TEST(LogTest, BadLength) {
+  const int kPayloadSize = kBlockSize - kHeaderSize;
+  Write(BigString("bar", kPayloadSize));
+  Write("foo");
+  // Least significant size byte is stored in header[4].
+  IncrementByte(4, 1);
+  ASSERT_EQ("foo", Read());
+  ASSERT_EQ(kBlockSize, DroppedBytes());
+  ASSERT_EQ("OK", MatchError("bad record length"));
+}
+
+TEST(LogTest, BadLengthAtEndIsIgnored) {
   Write("foo");
   ShrinkSize(1);
   ASSERT_EQ("EOF", Read());
-  ASSERT_EQ(kHeaderSize + 2, DroppedBytes());
-  ASSERT_EQ("OK", MatchError("bad record length"));
+  ASSERT_EQ(0, DroppedBytes());
+  ASSERT_EQ("", ReportMessage());
 }
 
 TEST(LogTest, ChecksumMismatch) {
@@ -415,6 +427,24 @@ TEST(LogTest, UnexpectedFirstType) {
   ASSERT_EQ("OK", MatchError("partial record without end"));
 }
 
+TEST(LogTest, MissingLastIsIgnored) {
+  Write(BigString("bar", kBlockSize));
+  // Remove the LAST block, including header.
+  ShrinkSize(14);
+  ASSERT_EQ("EOF", Read());
+  ASSERT_EQ("", ReportMessage());
+  ASSERT_EQ(0, DroppedBytes());
+}
+
+TEST(LogTest, PartialLastIsIgnored) {
+  Write(BigString("bar", kBlockSize));
+  // Cause a bad record length in the LAST block.
+  ShrinkSize(1);
+  ASSERT_EQ("EOF", Read());
+  ASSERT_EQ("", ReportMessage());
+  ASSERT_EQ(0, DroppedBytes());
+}
+
 TEST(LogTest, ErrorJoinsRecords) {
   // Consider two fragmented records:
   //    first(R1) last(R1) first(R2) last(R2)

db/repair.cc

@@ -242,7 +242,6 @@ class Repairer {
   }
 
   void ExtractMetaData() {
-    std::vector<TableInfo> kept;
     for (size_t i = 0; i < table_numbers_.size(); i++) {
       ScanTable(table_numbers_[i]);
     }

db/version_set.cc

@@ -54,20 +54,6 @@ static int64_t TotalFileSize(const std::vector<FileMetaData*>& files) {
   return sum;
 }
 
-namespace {
-std::string IntSetToString(const std::set<uint64_t>& s) {
-  std::string result = "{";
-  for (std::set<uint64_t>::const_iterator it = s.begin();
-       it != s.end();
-       ++it) {
-    result += (result.size() > 1) ? "," : "";
-    result += NumberToString(*it);
-  }
-  result += "}";
-  return result;
-}
-}  // namespace
-
 Version::~Version() {
   assert(refs_ == 0);