Update to leveldb 1.6

Highlights
----------
Mmap at most 1000 files on Posix to improve performance for large databases.
Support for more architectures (thanks to Alexander K.)

Building and porting
--------------------
HP/UX support (issue 126)
AtomicPointer for ia64 (issue 123)
Sparc v9 support (issue 124)
Atomic ops for powerpc
Use -fno-builtin-memcmp only when using g++
Simplify IOS build rules (issue 114)
Use CXXFLAGS instead of CFLAGS when invoking C++ compiler (issue 118)
Fix snappy shared library problem (issue 94)
Fix shared library installation path regression
Endian-ness detection tweak for FreeBSD

Bug fixes
---------
Stop ignoring FLAGS_open_files in db_bench
Make bloom test behavior agnostic to endian-ness

Performance
-----------
Limit number of mmapped files to 1000 to improve perf for large dbs
Do not delay for 1 second on shutdown path (issue 125)

Misc
----
Make InMemoryEnv return a no-op logger
C binding now has a wrapper for free (issue 117)
Add thread-safety annotations
Added an in-process lock table (issue 120)
Make RandomAccessFile and SequentialFile non-copyable

Makefile

@@ -15,7 +15,8 @@ OPT ?= -O2 -DNDEBUG       # (A) Production use (optimized mode)
 #-----------------------------------------------
 
 # detect what platform we're building on
-$(shell ./build_detect_platform build_config.mk)
+$(shell CC=$(CC) CXX=$(CXX) TARGET_OS=$(TARGET_OS) \
+    ./build_detect_platform build_config.mk ./)
 # this file is generated by the previous line to set build flags and sources
 include build_config.mk
 
@@ -70,7 +71,7 @@ SHARED = $(SHARED1)
 else
 # Update db.h if you change these.
 SHARED_MAJOR = 1
-SHARED_MINOR = 5
+SHARED_MINOR = 6
 SHARED1 = libleveldb.$(PLATFORM_SHARED_EXT)
 SHARED2 = $(SHARED1).$(SHARED_MAJOR)
 SHARED3 = $(SHARED1).$(SHARED_MAJOR).$(SHARED_MINOR)
@@ -82,7 +83,7 @@ $(SHARED2): $(SHARED3)
 endif
 
 $(SHARED3):
-	$(CXX) $(LDFLAGS) $(PLATFORM_SHARED_LDFLAGS)$(SHARED2) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) $(SOURCES) -o $(SHARED3)
+	$(CXX) $(SOURCES) $(LDFLAGS) $(PLATFORM_SHARED_LDFLAGS)$(INSTALL_PATH)$(SHARED2) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) -o $(SHARED3)
 
 endif  # PLATFORM_SHARED_EXT
 
@@ -179,14 +180,14 @@ IOSVERSION=$(shell defaults read $(PLATFORMSROOT)/iPhoneOS.platform/version CFBu
 
 .cc.o:
 	mkdir -p ios-x86/$(dir $@)
-	$(SIMULATORROOT)/usr/bin/$(CXX) $(CXXFLAGS) -isysroot $(SIMULATORROOT)/SDKs/iPhoneSimulator$(IOSVERSION).sdk -arch i686 -c $< -o ios-x86/$@
+	$(CXX) $(CXXFLAGS) -isysroot $(SIMULATORROOT)/SDKs/iPhoneSimulator$(IOSVERSION).sdk -arch i686 -c $< -o ios-x86/$@
 	mkdir -p ios-arm/$(dir $@)
 	$(DEVICEROOT)/usr/bin/$(CXX) $(CXXFLAGS) -isysroot $(DEVICEROOT)/SDKs/iPhoneOS$(IOSVERSION).sdk -arch armv6 -arch armv7 -c $< -o ios-arm/$@
 	lipo ios-x86/$@ ios-arm/$@ -create -output $@
 
 .c.o:
 	mkdir -p ios-x86/$(dir $@)
-	$(SIMULATORROOT)/usr/bin/$(CC) $(CFLAGS) -isysroot $(SIMULATORROOT)/SDKs/iPhoneSimulator$(IOSVERSION).sdk -arch i686 -c $< -o ios-x86/$@
+	$(CC) $(CFLAGS) -isysroot $(SIMULATORROOT)/SDKs/iPhoneSimulator$(IOSVERSION).sdk -arch i686 -c $< -o ios-x86/$@
 	mkdir -p ios-arm/$(dir $@)
 	$(DEVICEROOT)/usr/bin/$(CC) $(CFLAGS) -isysroot $(DEVICEROOT)/SDKs/iPhoneOS$(IOSVERSION).sdk -arch armv6 -arch armv7 -c $< -o ios-arm/$@
 	lipo ios-x86/$@ ios-arm/$@ -create -output $@

TODO

@@ -7,6 +7,7 @@ db
   within [start_key..end_key]?  For Chrome, deletion of obsolete
   object stores, etc. can be done in the background anyway, so
   probably not that important.
+- There have been requests for MultiGet.
 
 After a range is completely deleted, what gets rid of the
 corresponding files if we do no future changes to that range.  Make

build_detect_platform

@@ -23,8 +23,9 @@
 #
 
 OUTPUT=$1
-if test -z "$OUTPUT"; then
-  echo "usage: $0 <output-filename>" >&2
+PREFIX=$2
+if test -z "$OUTPUT" || test -z "$PREFIX"; then
+  echo "usage: $0 <output-filename> <directory_prefix>" >&2
   exit 1
 fi
 
@@ -55,58 +56,72 @@ PLATFORM_SHARED_LDFLAGS="-shared -Wl,-soname -Wl,"
 PLATFORM_SHARED_CFLAGS="-fPIC"
 PLATFORM_SHARED_VERSIONED=true
 
-# On GCC, we pick libc's memcmp over GCC's memcmp via -fno-builtin-memcmp
+MEMCMP_FLAG=
+if [ "$CXX" = "g++" ]; then
+    # Use libc's memcmp instead of GCC's memcmp.  This results in ~40%
+    # performance improvement on readrandom under gcc 4.4.3 on Linux/x86.
+    MEMCMP_FLAG="-fno-builtin-memcmp"
+fi
+
 case "$TARGET_OS" in
     Darwin)
         PLATFORM=OS_MACOSX
-        COMMON_FLAGS="-fno-builtin-memcmp -DOS_MACOSX"
+        COMMON_FLAGS="$MEMCMP_FLAG -DOS_MACOSX"
         PLATFORM_SHARED_EXT=dylib
         PLATFORM_SHARED_LDFLAGS="-dynamiclib -install_name "
         PORT_FILE=port/port_posix.cc
         ;;
     Linux)
         PLATFORM=OS_LINUX
-        COMMON_FLAGS="-fno-builtin-memcmp -pthread -DOS_LINUX"
+        COMMON_FLAGS="$MEMCMP_FLAG -pthread -DOS_LINUX"
         PLATFORM_LDFLAGS="-pthread"
         PORT_FILE=port/port_posix.cc
         ;;
     SunOS)
         PLATFORM=OS_SOLARIS
-        COMMON_FLAGS="-fno-builtin-memcmp -D_REENTRANT -DOS_SOLARIS"
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_SOLARIS"
         PLATFORM_LDFLAGS="-lpthread -lrt"
         PORT_FILE=port/port_posix.cc
         ;;
     FreeBSD)
         PLATFORM=OS_FREEBSD
-        COMMON_FLAGS="-fno-builtin-memcmp -D_REENTRANT -DOS_FREEBSD"
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_FREEBSD"
         PLATFORM_LDFLAGS="-lpthread"
         PORT_FILE=port/port_posix.cc
         ;;
     NetBSD)
         PLATFORM=OS_NETBSD
-        COMMON_FLAGS="-fno-builtin-memcmp -D_REENTRANT -DOS_NETBSD"
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_NETBSD"
         PLATFORM_LDFLAGS="-lpthread -lgcc_s"
         PORT_FILE=port/port_posix.cc
         ;;
     OpenBSD)
         PLATFORM=OS_OPENBSD
-        COMMON_FLAGS="-fno-builtin-memcmp -D_REENTRANT -DOS_OPENBSD"
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_OPENBSD"
         PLATFORM_LDFLAGS="-pthread"
         PORT_FILE=port/port_posix.cc
         ;;
     DragonFly)
         PLATFORM=OS_DRAGONFLYBSD
-        COMMON_FLAGS="-fno-builtin-memcmp -D_REENTRANT -DOS_DRAGONFLYBSD"
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_DRAGONFLYBSD"
         PLATFORM_LDFLAGS="-lpthread"
         PORT_FILE=port/port_posix.cc
         ;;
     OS_ANDROID_CROSSCOMPILE)
         PLATFORM=OS_ANDROID
-        COMMON_FLAGS="-fno-builtin-memcmp -D_REENTRANT -DOS_ANDROID -DLEVELDB_PLATFORM_POSIX"
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_ANDROID -DLEVELDB_PLATFORM_POSIX"
         PLATFORM_LDFLAGS=""  # All pthread features are in the Android C library
         PORT_FILE=port/port_posix.cc
         CROSS_COMPILE=true
         ;;
+    HP-UX)
+        PLATFORM=OS_HPUX
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_HPUX"
+        PLATFORM_LDFLAGS="-pthread"
+        PORT_FILE=port/port_posix.cc
+        # man ld: +h internal_name
+        PLATFORM_SHARED_LDFLAGS="-shared -Wl,+h -Wl,"
+        ;;
     *)
         echo "Unknown platform!" >&2
         exit 1
@@ -116,11 +131,13 @@ esac
 # except for the test and benchmark files. By default, find will output a list
 # of all files matching either rule, so we need to append -print to make the
 # prune take effect.
-DIRS="util db table"
+DIRS="$PREFIX/db $PREFIX/util $PREFIX/table"
+
 set -f # temporarily disable globbing so that our patterns aren't expanded
 PRUNE_TEST="-name *test*.cc -prune"
 PRUNE_BENCH="-name *_bench.cc -prune"
-PORTABLE_FILES=`find $DIRS $PRUNE_TEST -o $PRUNE_BENCH -o -name '*.cc' -print | sort | tr "\n" " "`
+PORTABLE_FILES=`find $DIRS $PRUNE_TEST -o $PRUNE_BENCH -o -name '*.cc' -print | sort | sed "s,^$PREFIX/,," | tr "\n" " "`
+
 set +f # re-enable globbing
 
 # The sources consist of the portable files, plus the platform-specific port
@@ -133,7 +150,7 @@ if [ "$CROSS_COMPILE" = "true" ]; then
     true
 else
     # If -std=c++0x works, use <cstdatomic>.  Otherwise use port_posix.h.
-    $CXX $CFLAGS -std=c++0x -x c++ - -o /dev/null 2>/dev/null  <<EOF
+    $CXX $CXXFLAGS -std=c++0x -x c++ - -o /dev/null 2>/dev/null  <<EOF
       #include <cstdatomic>
       int main() {}
 EOF
@@ -146,7 +163,7 @@ EOF
 
     # Test whether Snappy library is installed
     # http://code.google.com/p/snappy/
-    $CXX $CFLAGS -x c++ - -o /dev/null 2>/dev/null  <<EOF
+    $CXX $CXXFLAGS -x c++ - -o /dev/null 2>/dev/null  <<EOF
       #include <snappy.h>
       int main() {}
 EOF
@@ -156,7 +173,7 @@ EOF
     fi
 
     # Test whether tcmalloc is available
-    $CXX $CFLAGS -x c++ - -o /dev/null -ltcmalloc 2>/dev/null  <<EOF
+    $CXX $CXXFLAGS -x c++ - -o /dev/null -ltcmalloc 2>/dev/null  <<EOF
       int main() {}
 EOF
     if [ "$?" = 0 ]; then

db/c.cc

@@ -578,4 +578,8 @@ void leveldb_env_destroy(leveldb_env_t* env) {
   delete env;
 }
 
+void leveldb_free(void* ptr) {
+  free(ptr);
+}
+
 }  // end extern "C"

db/c_test.c

@@ -204,6 +204,12 @@ int main(int argc, char** argv) {
   CheckCondition(err != NULL);
   Free(&err);
 
+  StartPhase("leveldb_free");
+  db = leveldb_open(options, dbname, &err);
+  CheckCondition(err != NULL);
+  leveldb_free(err);
+  err = NULL;
+
   StartPhase("open");
   leveldb_options_set_create_if_missing(options, 1);
   db = leveldb_open(options, dbname, &err);

db/db_bench.cc

@@ -693,6 +693,7 @@ class Benchmark {
     options.create_if_missing = !FLAGS_use_existing_db;
     options.block_cache = cache_;
     options.write_buffer_size = FLAGS_write_buffer_size;
+    options.max_open_files = FLAGS_open_files;
     options.filter_policy = filter_policy_;
     Status s = DB::Open(options, FLAGS_db, &db_);
     if (!s.ok()) {

db/db_impl.cc

@@ -609,7 +609,11 @@ void DBImpl::BackgroundCall() {
   assert(bg_compaction_scheduled_);
   if (!shutting_down_.Acquire_Load()) {
     Status s = BackgroundCompaction();
-    if (!s.ok()) {
+    if (s.ok()) {
+      // Success
+    } else if (shutting_down_.Acquire_Load()) {
+      // Error most likely due to shutdown; do not wait
+    } else {
       // Wait a little bit before retrying background compaction in
       // case this is an environmental problem and we do not want to
       // chew up resources for failed compactions for the duration of

db/db_impl.h

@@ -13,6 +13,7 @@
 #include "leveldb/db.h"
 #include "leveldb/env.h"
 #include "port/port.h"
+#include "port/thread_annotations.h"
 
 namespace leveldb {
 
@@ -71,7 +72,7 @@ class DBImpl : public DB {
   // Recover the descriptor from persistent storage.  May do a significant
   // amount of work to recover recently logged updates.  Any changes to
   // be made to the descriptor are added to *edit.
-  Status Recover(VersionEdit* edit);
+  Status Recover(VersionEdit* edit) EXCLUSIVE_LOCKS_REQUIRED(mutex_);
 
   void MaybeIgnoreError(Status* s) const;
 
@@ -80,27 +81,34 @@ class DBImpl : public DB {
 
   // Compact the in-memory write buffer to disk.  Switches to a new
   // log-file/memtable and writes a new descriptor iff successful.
-  Status CompactMemTable();
+  Status CompactMemTable()
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
 
   Status RecoverLogFile(uint64_t log_number,
                         VersionEdit* edit,
-                        SequenceNumber* max_sequence);
+                        SequenceNumber* max_sequence)
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
 
-  Status WriteLevel0Table(MemTable* mem, VersionEdit* edit, Version* base);
+  Status WriteLevel0Table(MemTable* mem, VersionEdit* edit, Version* base)
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
 
-  Status MakeRoomForWrite(bool force /* compact even if there is room? */);
+  Status MakeRoomForWrite(bool force /* compact even if there is room? */)
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
   WriteBatch* BuildBatchGroup(Writer** last_writer);
 
-  void MaybeScheduleCompaction();
+  void MaybeScheduleCompaction() EXCLUSIVE_LOCKS_REQUIRED(mutex_);
   static void BGWork(void* db);
   void BackgroundCall();
-  Status BackgroundCompaction();
-  void CleanupCompaction(CompactionState* compact);
-  Status DoCompactionWork(CompactionState* compact);
+  Status BackgroundCompaction() EXCLUSIVE_LOCKS_REQUIRED(mutex_);
+  void CleanupCompaction(CompactionState* compact)
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
+  Status DoCompactionWork(CompactionState* compact)
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
 
   Status OpenCompactionOutputFile(CompactionState* compact);
   Status FinishCompactionOutputFile(CompactionState* compact, Iterator* input);
-  Status InstallCompactionResults(CompactionState* compact);
+  Status InstallCompactionResults(CompactionState* compact)
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
 
   // Constant after construction
   Env* const env_;

db/db_test.cc

@@ -1442,6 +1442,12 @@ TEST(DBTest, DBOpen_Options) {
   db = NULL;
 }
 
+TEST(DBTest, Locking) {
+  DB* db2 = NULL;
+  Status s = DB::Open(CurrentOptions(), dbname_, &db2);
+  ASSERT_TRUE(!s.ok()) << "Locking did not prevent re-opening db";
+}
+
 // Check that number of files does not grow when we are out of space
 TEST(DBTest, NoSpace) {
   Options options = CurrentOptions();

db/version_set.cc

@@ -865,7 +865,7 @@ Status VersionSet::Recover() {
         if (edit.has_comparator_ &&
             edit.comparator_ != icmp_.user_comparator()->Name()) {
           s = Status::InvalidArgument(
-              edit.comparator_ + "does not match existing comparator ",
+              edit.comparator_ + " does not match existing comparator ",
               icmp_.user_comparator()->Name());
         }
       }

db/version_set.h

@@ -21,6 +21,7 @@
 #include "db/dbformat.h"
 #include "db/version_edit.h"
 #include "port/port.h"
+#include "port/thread_annotations.h"
 
 namespace leveldb {
 
@@ -159,7 +160,8 @@ class VersionSet {
   // current version.  Will release *mu while actually writing to the file.
   // REQUIRES: *mu is held on entry.
   // REQUIRES: no other thread concurrently calls LogAndApply()
-  Status LogAndApply(VersionEdit* edit, port::Mutex* mu);
+  Status LogAndApply(VersionEdit* edit, port::Mutex* mu)
+      EXCLUSIVE_LOCKS_REQUIRED(mu);
 
   // Recover the last saved descriptor from persistent storage.
   Status Recover();

doc/bench/db_bench_sqlite3.cc

@@ -618,7 +618,7 @@ class Benchmark {
         ErrorCheck(status);
 
         // Execute read statement
-        while ((status = sqlite3_step(read_stmt)) == SQLITE_ROW);
+        while ((status = sqlite3_step(read_stmt)) == SQLITE_ROW) {}
         StepErrorCheck(status);
 
         // Reset SQLite statement for another use

doc/index.html

@@ -408,7 +408,7 @@ <h2>Filters</h2>
 the number of disk reads substantially.
 <pre>
    leveldb::Options options;
-   options.filter_policy = NewBloomFilter(10);
+   options.filter_policy = NewBloomFilterPolicy(10);
    leveldb::DB* db;
    leveldb::DB::Open(options, "/tmp/testdb", &amp;db);
    ... use the database ...

helpers/memenv/memenv.cc

@@ -221,6 +221,11 @@ class WritableFileImpl : public WritableFile {
   FileState* file_;
 };
 
+class NoOpLogger : public Logger {
+ public:
+  virtual void Logv(const char* format, va_list ap) { }
+};
+
 class InMemoryEnv : public EnvWrapper {
  public:
   explicit InMemoryEnv(Env* base_env) : EnvWrapper(base_env) { }
@@ -358,6 +363,11 @@ class InMemoryEnv : public EnvWrapper {
     return Status::OK();
   }
 
+  virtual Status NewLogger(const std::string& fname, Logger** result) {
+    *result = new NoOpLogger;
+    return Status::OK();
+  }
+
  private:
   // Map from filenames to FileState objects, representing a simple file system.
   typedef std::map<std::string, FileState*> FileSystem;

include/leveldb/c.h

@@ -28,6 +28,7 @@
   be true on entry:
      *errptr == NULL
      *errptr points to a malloc()ed null-terminated error message
+       (On Windows, *errptr must have been malloc()-ed by this library.)
   On success, a leveldb routine leaves *errptr unchanged.
   On failure, leveldb frees the old value of *errptr and
   set *errptr to a malloc()ed error message.
@@ -268,6 +269,15 @@ extern void leveldb_cache_destroy(leveldb_cache_t* cache);
 extern leveldb_env_t* leveldb_create_default_env();
 extern void leveldb_env_destroy(leveldb_env_t*);
 
+/* Utility */
+
+/* Calls free(ptr).
+   REQUIRES: ptr was malloc()-ed and returned by one of the routines
+   in this file.  Note that in certain cases (typically on Windows), you
+   may need to call this routine instead of free(ptr) to dispose of
+   malloc()-ed memory returned by this library. */
+extern void leveldb_free(void* ptr);
+
 #ifdef __cplusplus
 }  /* end extern "C" */
 #endif