Hi Christoph,

I have been running through compiling hamster db in VS 2010 and I think there may be a possible bug in the wrapper. The issue arises when I try to access the exception object when an exception is thrown from hamsterdb when a key is not found. For example I have been running the SampleDb1 project and tried to put a breakpoint within the catch (See below) and it seems to crash the application. Note this particular piece of code is AFTER erasing all items. If I remove the breakpoint the application completes successfully however the WriteLine within the catch is never written..

                try
                {
                    byte[] r = db.Find(key);
                    if (r == null)
                    {
                        Console.WriteLine("r is null");
                    }
                    else
                    {
                        Console.WriteLine("r is not null");
                    }
                }
                catch (DatabaseException e)
                {
                    if (e.ErrorCode != HamConst.HAM_KEY_NOT_FOUND)
                    {
                        Console.WriteLine("db.Find() returned error " + e);
                        return;
                    }
                }

Please let me know if there’s anything I can do to assist or if you require further information.

With 2.0.3 I'm trying to store a key "FooBar" and then find it by "Foo", however, the ham_find seems to leave the "Foo" in ham_key_t, failing to return the "FooBar" key.

Below is a more detailed description:

recKey << key << '\x1F' << suffix;
log_info ("cacheLocal, recKey: " << recKey << " (" << recKey.size() << ")");
ham_key_t hkey; memset (&hkey, 0, sizeof (ham_key_t));
hkey.data = (void*) recKey.data(); hkey.size = recKey.size();
ham_record_t record; memset (&record, 0, sizeof (ham_record_t));
record.data = (void*) data.data(); record.size = data.size();
ham_status_t st = ham_insert ((ham_db_t*) _db.get(), txn, &hkey, &record, HAM_OVERWRITE);
...
ham_key_t hkey; memset (&hkey, 0, sizeof (ham_key_t));
hkey.data = (void*) key.data(); hkey.size = key.size();
ham_record_t record; memset (&record, 0, sizeof (ham_record_t));
ham_status_t st = ham_find ((ham_db_t*) _db.get(), txn, &hkey, &record, HAM_FIND_GT_MATCH);
if (st == HAM_KEY_NOT_FOUND) return;
log_info ("getLocal, looking for " << key << " (" << key._buf << ", " << key.size() << "); found " << hkey.data << " (" << hkey.size << ")");

The output is:
cacheLocal, recKey: 9l1FJRJZADA6wtVfUUKC0SSvpFIF7zfK000at (38)
getLocal, looking for 9l1FJRJZADA6wtVfUUKC0SSvpFIF7zfK (0x2b68716fafe0, 32); found 0x2b68716fafe0 (32)

There is no key of size 32 in the database! Yet after ham_find (&hkey) I'm still seeing the same 32-byte prefix in hkey.

just a test

Reproduce with
./ham_bench --stop-ops=50000000 --recsize-fixed=37 --keysize-fixed --keysize=21 --distribution=ascending

Would you like to replace more defines for constant values by enumerations to stress their relationships?

• HAM_TYPE_…
• HAM_INV_…

To reproduce: run ham_bench several times
This is caused by the I/O functions in os_win32.cc: they are not atomic. Solution: use a mutex.

If openssl libraries are installed but the header files are missing there's a compilation error:

CXX env_local.lo
In file included from device_disk.h:22:0,
from device_factory.h:17,
from env_local.cc:18:
aes.h:21:25: fatal error: openssl/evp.h: No such file or directory
compilation terminated.
make[3]: *** [env_local.lo] Error 1
make[3]: Leaving directory /root/ryan/hamsterdb-2.1.2/src' make[2]: *** [all-recursive] Error 1 make[2]: Leaving directory/root/ryan/hamsterdb-2.1.2/src'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/root/ryan/hamsterdb-2.1.2'

workaround: install openssl-devel packages (Ubuntu: libssl-dev).

In a specific sequence of inserts/deletes in various databases, a segfault can happen.

I have received a test case which reproduces the issue.

On Linux 2.6.32.12 on armv5tel, using gcc version 4.2.3, I receive the following error while trying to 'make':

CXX  env_local.loenv_local.cc:102: error: integer constant is too large for 'long' type
env_local.cc:219: error: integer constant is too large for 'long' type

I can resolve this error by change instances of '0xffffffffffffffff' to '0xffffffffffffffffULL'.

See Dmitriy's mail from dec 17:

What I see is that Common.db uses bogus values for the record size:

average record size:  1086
minimum record size:  419
maximum record size:  419

If minimum and maximum is 419 then the average cannot be 1086. I think one of these values is off, and therefore maybe the max. key size is misleading. But this is just a wrong printout, it does not explain why you see unexpected results in your program.

Are there any plans to create a distributed version of upscaledb, e.g. on hdfs?

I've build and installed libuv from source, but I get this error:

make[3]: Entering directory `/home/wouter/upscaledb/src/5server'
  CXX      upsserver.lo
upsserver.cc: In function 'void upscaledb::on_new_connection(uv_stream_t*, int)':
upsserver.cc:1994:71: error: invalid conversion from 'uv_buf_t (*)(uv_handle_t*, size_t) {aka uv_buf_t (*)(uv_handle_s*, long unsigned int)}' to 'uv_alloc_cb {aka void (*)(uv_handle_s*, long unsigned int, uv_buf_t*)}' [-fpermissive]
     uv_read_start((uv_stream_t *)client, on_alloc_buffer, on_read_data);
                                                                       ^
In file included from upsserver.cc:26:0:
/usr/local/include/uv.h:465:15: error:   initializing argument 2 of 'int uv_read_start(uv_stream_t*, uv_alloc_cb, uv_read_cb)' [-fpermissive]
 UV_EXTERN int uv_read_start(uv_stream_t*,
               ^
upsserver.cc:1994:71: error: invalid conversion from 'void (*)(uv_stream_t*, ssize_t, uv_buf_t) {aka void (*)(uv_stream_s*, long int, uv_buf_t)}' to 'uv_read_cb {aka void (*)(uv_stream_s*, long int, const uv_buf_t*)}' [-fpermissive]
     uv_read_start((uv_stream_t *)client, on_alloc_buffer, on_read_data);
                                                                       ^
In file included from upsserver.cc:26:0:
/usr/local/include/uv.h:465:15: error:   initializing argument 3 of 'int uv_read_start(uv_stream_t*, uv_alloc_cb, uv_read_cb)' [-fpermissive]
 UV_EXTERN int uv_read_start(uv_stream_t*,
               ^
upsserver.cc: In function 'ups_status_t ups_srv_init(ups_srv_config_t*, ups_srv_t**)':
upsserver.cc:2041:50: error: too few arguments to function 'int uv_ip4_addr(const char*, int, sockaddr_in*)'
   bind_addr = uv_ip4_addr("0.0.0.0", config->port);
                                                  ^
In file included from upsserver.cc:26:0:
/usr/local/include/uv.h:1358:15: note: declared here
 UV_EXTERN int uv_ip4_addr(const char* ip, int port, struct sockaddr_in* addr);
               ^
upsserver.cc:2042:38: error: cannot convert 'sockaddr_in' to 'const sockaddr*' for argument '2' to 'int uv_tcp_bind(uv_tcp_t*, const sockaddr*, unsigned int)'
   uv_tcp_bind(&srv->server, bind_addr);
                                      ^
upsserver.cc:2058:52: error: invalid conversion from 'void (*)(uv_async_t*, int) {aka void (*)(uv_async_s*, int)}' to 'uv_async_cb {aka void (*)(uv_async_s*)}' [-fpermissive]
   uv_async_init(srv->loop, &srv->async, on_async_cb);
                                                    ^
In file included from upsserver.cc:26:0:
/usr/local/include/uv.h:763:15: error:   initializing argument 3 of 'int uv_async_init(uv_loop_t*, uv_async_t*, uv_async_cb)' [-fpermissive]
 UV_EXTERN int uv_async_init(uv_loop_t*,

When compiling hamsterdb 2.1.11 on ubuntu 15.10, the configure fails with
checking whether the compiler supports GCC C++ ABI name demangling... yes
checking for Boost headers version >= 1.53.0... yes
checking for Boost's header version...

configure: error: invalid value: boost_major_version=

Boost library is installed (1.58).

I'm not sure where this is going wrong, but $boost_cv_lib_version contains an empty string after the test in configure (around line 17338).

Quick 'n dirty workaround is to manually assign "158" to boost_cv_lib_version after the test.

This allows hamsterdb to compile successfully.

/usr/include/boost/version.hpp:
// Boost version.hpp configuration header file ------------------------------//

// (C) Copyright John maddock 1999. Distributed under the Boost
// Software License, Version 1.0. (See accompanying file
// LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)

// See http://www.boost.org/libs/config for documentation

ifndef BOOST_VERSION_HPP

define BOOST_VERSION_HPP

//
// Caution: this is the only Boost header that is guaranteed
// to change with every Boost release. Including this header
// will cause a recompile every time a new Boost version is
// used.
//
// BOOST_VERSION % 100 is the patch level
// BOOST_VERSION / 100 % 1000 is the minor version
// BOOST_VERSION / 100000 is the major version

define BOOST_VERSION 105800

//
// BOOST_LIB_VERSION must be defined to be the same as BOOST_VERSION
// but as a string in the form "x_y[_z]" where x is the major version
// number, y is the minor version number, and z is the patch level if not 0.
// This is used by <config/auto_link.hpp> to select which library version to link to.

define BOOST_LIB_VERSION "1_58"

endif

Hello,

Even if I set the -fpermissive flag, it's still fails to compile.

make[3]: Entering directory `/home/randy/prg/hamsterdb-2.1.7/src/server'
CXX hamserver.lo
hamserver.cc: In function 'void hamsterdb::on_new_connection(uv_stream_t_, int)':
hamserver.cc:1316:71: error: invalid conversion from 'uv_buf_t ()(uv_handle_t, size_t) {aka uv_buf_t ()(uv_handle_s, unsigned int)}' to 'uv_alloc_cb {aka void ()(uv_handle_s, unsigned int, uv_buf_t_)}' [-fpermissive]
In file included from hamserver.h:25:0,
from hamserver.cc:25:
/usr/local/include/uv.h:645:15: error: initializing argument 2 of 'int uv_read_start(uv_stream_t_, uv_alloc_cb, uv_read_cb)' [-fpermissive]
hamserver.cc:1316:71: error: invalid conversion from 'void ()(uv_stream_t, ssize_t, uv_buf_t) {aka void ()(uv_stream_s, int, uv_buf_t)}' to 'uv_read_cb {aka void ()(uv_stream_s, int, const uv_buf_t_)}' [-fpermissive]
In file included from hamserver.h:25:0,
from hamserver.cc:25:
/usr/local/include/uv.h:645:15: error: initializing argument 3 of 'int uv_read_start(uv_stream_t_, uv_alloc_cb, uv_read_cb)' [-fpermissive]
hamserver.cc: In function 'ham_status_t ham_srv_init(ham_srv_config_t_, ham_srv_t*)':
hamserver.cc:1354:69: error: too few arguments to function 'int uv_ip4_addr(const char, int, sockaddr_in_)'
In file included from hamserver.h:25:0,
from hamserver.cc:25:
/usr/local/include/uv.h:2052:15: note: declared here
hamserver.cc:1355:38: error: cannot convert 'sockaddr_in' to 'const sockaddr_' for argument '2' to 'int uv_tcp_bind(uv_tcp_t_, const sockaddr_, unsigned int)'
hamserver.cc:1365:52: error: invalid conversion from 'void ()(uv_async_t, int) {aka void ()(uv_async_s, int)}' to 'uv_async_cb {aka void ()(uv_async_s)}' [-fpermissive]
In file included from hamserver.h:25:0,
from hamserver.cc:25:
/usr/local/include/uv.h:1352:15: error: initializing argument 3 of 'int uv_async_init(uv_loop_t_, uv_async_t_, uv_async_cb)' [-fpermissive]
make[3]: *** [hamserver.lo] Error 1

build fails with google perftools (2.0) is installed

Undefined symbols for architecture x86_64:
"MallocExtension::instance()", referenced from:
hamsterdb::Memory::get_global_metrics(ham_env_metrics_t_) in libhamsterdb.a(mem.o)
hamsterdb::Memory::release_to_system() in libhamsterdb.a(mem.o)
"tc_calloc", referenced from:
HamsterdbFixture::callocTest() in hamsterdb.o
hamsterdb::BtreeCursor::uncouple(unsigned int) in libhamsterdb.a(btree_cursor.o)
hamsterdb::BtreeCursor::clone(hamsterdb::BtreeCursor) in libhamsterdb.a(btree_cursor.o)
hamsterdb::BtreeCursor::get_duplicate_table(hamsterdb::PDupeTable*, bool) in libhamsterdb.a(btree_cursor.o)
"tc_free", referenced from:
hamsterdb::ByteArray::clear() in db.o
hamsterdb::ExtKeyCache::ExtKeyHelper::remove_if(hamsterdb::ExtKeyCache::ExtKey) in extkeys.o
hamsterdb::ExtKeyCache::remove(unsigned long) in extkeys.o
HamsterdbFixture::callocTest() in hamsterdb.o
hamsterdb::JournalFixture::compareJournal(hamsterdb::Journal_, hamsterdb::LogEntry_, unsigned int) in journal.o
hamsterdb::JournalFixture::appendEraseTest() in journal.o
hamsterdb::JournalFixture::appendPartialInsertTest() in journal.o
...
"tc_malloc", referenced from:
hamsterdb::ExtKeyCache::insert(unsigned long, unsigned int, unsigned char const) in extkeys.o
hamsterdb::Database::copy_key(ham_key_t const_, ham_key_t_) in misc.o
hamsterdb::BtreeIndex::copy_key(hamsterdb::PBtreeKey const_, ham_key_t_) in libhamsterdb.a(btree.o)
hamsterdb::ExtKeyCache::insert(unsigned long, unsigned int, unsigned char const_) in libhamsterdb.a(btree_insert.o)
hamsterdb::Database::copy_key(ham_key_t const_, ham_key_t_) in libhamsterdb.a(btree_insert.o)
hamsterdb::Database::copy_key(ham_key_t const_, ham_key_t_) in libhamsterdb.a(btree_cursor.o)
hamsterdb::DiskDevice::read_page(hamsterdb::Page_) in libhamsterdb.a(env.o)
...
"tc_realloc", referenced from:
void* hamsterdb::Memory::reallocate(void, unsigned long) in libhamsterdb.a(btree.o)
void_ hamsterdb::Memory::reallocate(void_, unsigned long) in libhamsterdb.a(txn_cursor.o)
void_ hamsterdb::Memory::reallocate(void_, unsigned long) in libhamsterdb.a(db.o)
void_ hamsterdb::Memory::reallocate(void_, unsigned long) in libhamsterdb.a(blob_manager_disk.o)
void_ hamsterdb::Memory::reallocate(void*, unsigned long) in libhamsterdb.a(blob_manager_inmem.o)
ld: symbol(s) not found for architecture x86_64
clang: error: linker command failed with exit code 1 (use -v to see invocation)

It seems the compilation fails if the protobuf-compiler and/or the libprotocol-libraries are missing.

make[3]: Entering directory `/home/joe/hamsterdb-2.1.0/src'
/bin/bash ../libtool --tag=CXX --mode=compile g++ -DHAVE_CONFIG_H -I. -I.. -I../include -I../include -g -O2 -Wall -DHAM_LITTLE_ENDIAN -fno-tree-vectorize -D_FILE_OFFSET_BITS=64 -MT blob.lo -MD -MP -MF .deps/blob.Tpo -c -o blob.lo blob.cc
libtool: compile: g++ -DHAVE_CONFIG_H -I. -I.. -I../include -I../include -g -O2 -Wall -DHAM_LITTLE_ENDIAN -fno-tree-vectorize -D_FILE_OFFSET_BITS=64 -MT blob.lo -MD -MP -MF .deps/blob.Tpo -c blob.cc -fPIC -DPIC -o .libs/blob.o
In file included from env.h:34:0,
from db.h:28,
from blob.cc:18:
protocol/protocol.h:23:25: fatal error: messages.pb.h: No such file or directory
compilation terminated.
make[3]: *** [blob.lo] Error 1

there are 2 locations in configure that do:
LIBS="-ltcmalloc_minimal $LIBS"

these appear to run even when --without-tcmalloc is specified

i'm not sure why, but consequently this is used when checking for protobuf and causes the test to segfault, causing configure to not detect the library

configure:18216: g++ -o conftest -g -O2 conftest.cpp -lcrypto -ltcmalloc_minimal >&5
conftest.cpp:43:1: warning: "HAVE_GOOGLE_TCMALLOC_H" redefined
conftest.cpp:33:1: warning: this is the location of the previous definition
configure:18216: $? = 0
configure:18216: ./conftest
./configure: line 2107: 39775 Segmentation fault: 11 ./conftest$ac_exeext
configure:18216: $? = 139
configure: program exited with status 139

this was observed on osx 10.8.4

mem.cc:61:5: error: no type named 'malloc_trim' in the global namespace
::malloc_trim(os_get_granularity());

This function is missing on some platforms:
http://www.gnu.org/software/gnulib/manual/html_node/malloc_005ftrim.html

As reported by Michael Möllney; i'm able to reproduce this

Example: The time series is created sequentially by inserting (1,1) (2,2) (3,3) ...

Key, Value
1,1
2,2
3,3
4,4
5,5
6,6

The read cursor is put at the end via cursor.move_last() to be at the end of the series shortly after a new entry is inserted. So the read cursor is at the end while a new entry is put "behind" the end. This seems to be a critical action, because if this is not this constellation the following does not happen!

Read cursor is at position 5.
Write cursor puts in (6,6).
Read cursor is positioned via cursor.move_last().
Now the read cursor is put again at position 6 (ok)
Now the read cursor is asked to find(key,record,HAM_FIND_LT_MATCH) with key = 6. The result is key=5 and record=5 (ok)
Now repeat the step backward in time: find(key,record,HAM_FIND_LT_MATCH) with key=5. The result is key=4 and record = 4 (ok)
Now ask for the step forward in time: find(key,record,HAM_FIND_GT_MATCH) with key=4. The result is key=4 and record=6 (?????)

All this is NOT happening, if the last entry is put at the end and the read cursor is not at the last position.

If cache size is very small then file size grows a LOT:

./ham_bench --cache=1024 --stop-ops=10000
hamsterdb filesize 154468352

./ham_bench --stop-ops=10000
hamsterdb filesize 10534912 (10 times less!)

I have looked at a few source files for your current software. I have noticed that some checks for return codes are missing.

Would you like to add more error handling for return values from functions like the following?

• ftell ⇒ read_config
• fwrite ⇒ BinaryExporter

The variable "running" should use the data type "sig_atomic_t", shouldn't it?

The following loop writes many "insert" entries to the journal .jrn0:

ham_env_create(... HAM_ENABLE_TRANSACTIONS);

while (1) {
ham_insert(db, NULL, key, record, 0);
}

The problem is that the ham_txn_begin/ham_txn_commit markers are not written to the journal, and therefore .jrn1 is never used.

This could be implemented in Journal::append_insert; if the txn pointer is a TEMPORARY one then the txn counters in the journal will be increased/decreased.

I would like to point out that identifiers like "HAM_HAMSTERDB_HPP__" and "HAM_TYPES_H__" do not fit to the expected naming convention of the C++ language standard.
Would you like to adjust your selection for unique names?

see this thread:
https://groups.google.com/forum/#!topic/hamsterdb-user/klHj8j1bh5c

When opening a file, the mmap view is limited to 2^32 bytes.
What if the file is larger? Will this lead to problems?

Journal::recover -> __abort_uncommitted_txns -> ham_txn_abort
-> env->get_txn_manager()->abort
crashes because env->m_txn_manager == 0

... and HAM_PARAM_KEY_SIZE > 32
and HAM_KEY_TYPE = HAM_TYPE_BINARY or HAM_TYPE_CUSTOM

I have received a test case which reproduces the issue.

I'm getting a "Data blob not found" error from the ham_find. Makes no sense to me.
Database: http://glim.ru/personal/bugs/FUC.7z

btree: insert "aa"
txn: begin
txn: erase "aa"
txn: find "aa" with GEQ match
-> segfault

log file switching is basically disabled when creating temporary transactions. Same when using recovery (without transactions) - the changesets are always written to the same log.

Reason: a native function returning char * cannot be directly marshalled in a string!

See http://stackoverflow.com/questions/6300093/why-cant-i-return-a-char-string-from-c-to-c-sharp-in-a-release-build

I find that include guards like "GRAPH_H" and "COMMON_H" are too short for the safe reuse of your header files (when they belong to an application programming interface).

The code below can be pasted into dotnet\unittests\CursorTest.cs and run with either ReproduceBug(true) or ReproduceBug(false). The point in the test at which each fails has been annotated within the code. I've tried to provide a minimal example of the error, however the following things seems to be required to obtain the errors: 1) Use trasactions, 2) Inserting unequal length arrays for the record values, and 3) Using Cursor.Find followed by Cursor.Move to iterate over the contents of the DB.

By the way, the .NET project is missing the file Properties\AssemblyInfo.cs in the GitHub repo.

        private void WithTransaction(Action<Transaction> f)
        {
            var txn = env.Begin();
            try
            {
                f(txn);
                txn.Commit();
            }
            catch
            {
                txn.Abort();
                throw;
            }
        }

        private void ReproduceBug(bool bugVariation)
        {
            env = new Hamster.Environment();
            env.Create("ntest.db", HamConst.HAM_ENABLE_TRANSACTIONS);   // Note: not using transactions works fine

            var prm = new Parameter();
            prm.name = HamConst.HAM_PARAM_KEY_TYPE;
            prm.value = HamConst.HAM_TYPE_UINT64;
            var prms = new Parameter[] { prm };

            db = new Database();
            db = env.CreateDatabase(1, HamConst.HAM_ENABLE_DUPLICATE_KEYS, prms);

            var k1 = new byte[] { 128, 93, 150, 237, 49, 178, 92, 8 };
            var k2 = new byte[] { 0, 250, 234, 1, 199, 250, 128, 8 };
            var k3 = new byte[] { 128, 17, 181, 113, 1, 220, 132, 8 };

            // print keys (note they are in ascending order as UInt64)
            Console.WriteLine("{0}", BitConverter.ToUInt64(k1, 0));
            Console.WriteLine("{0}", BitConverter.ToUInt64(k2, 0));
            Console.WriteLine("{0}", BitConverter.ToUInt64(k3, 0));
            Console.WriteLine();

            var v1  = new byte[46228];   // Note: using equal size value byte arrays works fine!
            var v11 = new byte[446380];
            var v12 = new byte[525933];
            var v21 = new byte[334157];
            var v22 = new byte[120392];
            WithTransaction(txn => db.Insert(txn, k1, v1, Hamster.HamConst.HAM_DUPLICATE));
            WithTransaction(txn => db.Insert(txn, k2, v11, Hamster.HamConst.HAM_DUPLICATE));
            WithTransaction(txn => db.Insert(txn, k2, v12, Hamster.HamConst.HAM_DUPLICATE));
            WithTransaction(txn => db.Insert(txn, k3, v21, Hamster.HamConst.HAM_DUPLICATE));
            WithTransaction(txn => db.Insert(txn, k3, v22, Hamster.HamConst.HAM_DUPLICATE));

            WithTransaction(txn =>
            {
                using (var c = new Cursor(db, txn))
                {
                    // Note: calling c.Move(HamConst.HAM_CURSOR_NEXT) instead works fine!
                    if (bugVariation)                   
                        c.Find(k1, HamConst.HAM_FIND_GEQ_MATCH);    
                    else
                        c.Find(k1);

                    var s1 = c.GetKey();
                    Console.WriteLine("{0}", BitConverter.ToUInt64(s1, 0));

                    c.Move(HamConst.HAM_CURSOR_NEXT);
                    var s2 = c.GetKey();
                    Console.WriteLine("{0}", BitConverter.ToUInt64(s2, 0));

                    c.Move(HamConst.HAM_CURSOR_NEXT);
                    var s3 = c.GetKey();
                    Console.WriteLine("{0}", BitConverter.ToUInt64(s3, 0));

                    c.Move(HamConst.HAM_CURSOR_NEXT);
                    var s4 = c.GetKey();
                    Console.WriteLine("{0}", BitConverter.ToUInt64(s4, 0));

                    c.Move(HamConst.HAM_CURSOR_NEXT);   // fails here when bugVariation == false
                    var s5 = c.GetKey();
                    Console.WriteLine("{0}", BitConverter.ToUInt64(s5, 0));

                    checkEqual(k1, s1);
                    checkEqual(k2, s2); // fails here when bugVariation == true
                    checkEqual(k2, s3);
                    checkEqual(k3, s4);
                    checkEqual(k3, s5);
                }
            });

            env.Close();

            return;
        }

In attempting to build for osx and redhat/centos 6.4, I was unable to build without specifying ./configure LIBS="-lgnutls"

Error:
hamsterdb/unittests/main.cpp:230: undefined reference to `gnutls_global_deinit'

During recovery, the compare callback is required - but obviously it's not yet installed.

Hamsterdb 2.0.2

/bin/sh ../libtool --tag=CXX   --mode=compile g++ -DHAVE_CONFIG_H -I. -I..  -I../include -I../3rdparty/aes    -g -O2 -Wall -DHAM_LITTLE_ENDIAN  -MT log.lo -MD -MP -MF .deps/log.Tpo -c -o log.lo log.cc
libtool: compile:  g++ -DHAVE_CONFIG_H -I. -I.. -I../include -I../3rdparty/aes -g -O2 -Wall -DHAM_LITTLE_ENDIAN -MT log.lo -MD -MP -MF .deps/log.Tpo -c log.cc  -fno-common -DPIC -o .libs/log.o
log.cc: In member function 'std::string Log::get_path()':
log.cc:410: error: '::basename' has not been declared
make[3]: *** [log.lo] Error 1
make[2]: *** [all-recursive] Error 1
make[1]: *** [all-recursive] Error 1
make: *** [all] Error 2

Darwin wsantos 11.4.0 Darwin Kernel Version 11.4.0: Mon Apr  9 19:32:15 PDT 2012; root:xnu-1699.26.8~1/RELEASE_X86_64 x86_64

hamsterdb-2.0.2 ❯ libtool -V       
Apple Inc. version cctools-822

Using built-in specs.
Target: i686-apple-darwin11
Configured with: /private/var/tmp/llvmgcc42/llvmgcc42-2336.9~22/src/configure --disable-checking --enable-werror --prefix=/Applications/Xcode.app/Contents/Developer/usr/llvm-gcc-4.2 --mandir=/share/man --enable-languages=c,objc,c++,obj-c++ --program-prefix=llvm- --program-transform-name=/^[cg][^.-]*$/s/$/-4.2/ --with-slibdir=/usr/lib --build=i686-apple-darwin11 --enable-llvm=/private/var/tmp/llvmgcc42/llvmgcc42-2336.9~22/dst-llvmCore/Developer/usr/local --program-prefix=i686-apple-darwin11- --host=x86_64-apple-darwin11 --target=i686-apple-darwin11 --with-gxx-include-dir=/usr/include/c++/4.2.1
Thread model: posix
gcc version 4.2.1 (Based on Apple Inc. build 5658) (LLVM build 2336.9.00)

• CentOS 5.9 x86_64
• boost: 1.33.1 is installed by yum.

  hamsterdb.cc:1191: error: no matching function for call to 'boost::detail::thread::scoped_lockboost::mutex::scoped_lock()'
  /usr/include/boost/thread/detail/lock.hpp:68: note: candidates are: boost::detail::thread::scoped_lock::scoped_lock(Mutex&, bool) [with Mutex = boost::mutex]
  /usr/include/boost/thread/detail/lock.hpp:64: note:                 boost::detail::thread::scoped_lockboost::mutex::scoped_lock(const boost::detail::thread::scoped_lockboost::mutex&)
  /usr/include/boost/noncopyable.hpp: In member function 'boost::detail::thread::scoped_lockboost::mutex& boost::detail::thread::scoped_lockboost::mutex::operator=(const boost::detail::thread::scoped_lockboost::mutex&)':
  /usr/include/boost/noncopyable.hpp:28: error: 'const boost::noncopyable_::noncopyable& boost::noncopyable_::noncopyable::operator=(const boost::noncopyable_::noncopyable&)' is private
  /usr/include/boost/thread/detail/lock.hpp:64: error: within this context
  /usr/include/boost/thread/detail/lock.hpp:64: error: non-static reference member 'boost::mutex& boost::detail::thread::scoped_lockboost::mutex::m_mutex', can't use default assignment operator
  

Hamsterdb 2.1.10 falls into the endless recursion if I declare an environment with the HAM_ENABLE_TRANSACTIONS flag, but the cursor itself has no associated transaction (also there're no active transactions around):

ham_cursor_create(&cursor, _dbi, NULL, 0);
ham_cursor_find(cursor, &key, &rec, HAM_FIND_GEQ_MATCH);

Dbx_kv.dll!hamsterdb::LocalDatabase::find_txn(hamsterdb::Context * context=0x000000000012f128, hamsterdb::Cursor * cursor=0x0000000003510060, ham_key_t * key=0x000000000012f468, ham_record_t * record=0x000000000012f428, unsigned int flags=24576) Line 438 C++
Dbx_kv.dll!hamsterdb::LocalDatabase::find_txn(hamsterdb::Context * context=0x000000000012f128, hamsterdb::Cursor * cursor=0x0000000003510060, ham_key_t * key=0x000000000012f468, ham_record_t * record=0x000000000012f428, unsigned int flags=24576) Line 438 C++
Dbx_kv.dll!hamsterdb::LocalDatabase::find_txn(hamsterdb::Context * context=0x000000000012f128, hamsterdb::Cursor * cursor=0x0000000003510060, ham_key_t * key=0x000000000012f468, ham_record_t * record=0x000000000012f428, unsigned int flags=24576) Line 438 C++
Dbx_kv.dll!hamsterdb::LocalDatabase::find_txn(hamsterdb::Context * context=0x000000000012f128, hamsterdb::Cursor * cursor=0x0000000003510060, ham_key_t * key=0x000000000012f468, ham_record_t * record=0x000000000012f428, unsigned int flags=24576) Line 438 C++
Dbx_kv.dll!hamsterdb::LocalDatabase::find_impl(hamsterdb::Context * context=0x000000000012f128, hamsterdb::Cursor * cursor=0x0000000003510060, ham_key_t * key=0x000000000012f468, ham_record_t * record=0x000000000012f428, unsigned int flags=24576) Line 1670 C++
Dbx_kv.dll!hamsterdb::LocalDatabase::find(hamsterdb::Cursor * cursor=0x0000000003510060, hamsterdb::Transaction * txn=0x0000000000000000, ham_key_t * key=0x000000000012f468, ham_record_t * record=0x000000000012f428, unsigned int flags=24576) Line 1135 C++
Dbx_kv.dll!ham_cursor_find(ham_cursor_t * hcursor=0x0000000003510060, ham_key_t * key=0x000000000012f468, ham_record_t * record=0x000000000012f428, unsigned int flags=24576) Line 1278 C++
Dbx_kv.dll!CDbxKV::EnumContactSettings(unsigned int contactID=0, DBCONTACTENUMSETTINGS * dbces=0x000000000012f6f8) Line 585 C++

./ham_bench --key=uint64 --recsize=10241024 --distribution=ascending --use-transactions=1

then press ctrl-c after a few seconds

../ham_recover test-ham.db3journal/journal.cc[137]: Changeset magic is invalid, skipping
3journal/journal.cc[137]: Changeset magic is invalid, skipping
3journal/journal.cc[651]: Changeset magic is invalid, skipping
1os/os_posix.cc[213]: File::pread() failed with short read (No such file or directory)
ham_env_open() returned error -18: System I/O error

seems that the recovery routine cannot cope with the incomplete changeset entry.

add a unittest for this!

configure fails early on with:
conftest.cpp:49:16: error: no type named 'thread_clock' in namespace 'boost::chrono'; did you mean 'steady_clock'?
boost::chrono::thread_clock d;

Looking elsewhere, seems thread_clock is not available on osx
Doesn't seem like this is used elsewhere, commenting out thread_clock test produces a working build.

hamsterdb.h defines the parameter as
/** Parameter name for @ref ham_open_ex, @ref ham_create_ex; sets the page

• size */
  #define HAM_PARAM_PAGESIZE 0x00000101

this should be @ref ham_env_create_ex, @ref ham_create_ex;

A call to Cursor.Find with one of the search flags (e.g., HAM_FIND_GT_MATCH) should return both the matching key and value of the result. This works fine when accessing the database file directly. However, when accessing the database via the hamzilla server, the original search key is returned instead of the key corresponding to the result.

The following code gives an example of the error, and can be copy-pasted into 'CursorTest.cs' in the 'Unittests' project of the hamsterdb-dotnet project.

        // Note: this is just the 'checkEqual' method from CursorTest.cs made into a static method
        static void checkEqual(byte[] lhs, byte[] rhs)
        {
            Assert.AreEqual(lhs.Length, rhs.Length);
            for (int i = 0; i < lhs.Length; i++)
                Assert.AreEqual(lhs[i], rhs[i]);
        }

        private static void Search()
        {
            byte[] k1 = new byte[5]; k1[0] = 5;
            byte[] k2 = new byte[5]; k2[0] = 6;
            byte[] r1 = new byte[5]; r1[0] = 1;
            byte[] r2 = new byte[5]; r2[0] = 2;

            // direct file access
            var env = new Hamster.Environment();
            var db = new Database();
            env.Create("ntest.db");
            db = env.CreateDatabase(1, 0);
            Cursor c = new Cursor(db);
            db.Insert(k1, r1, Hamster.HamConst.HAM_OVERWRITE);
            db.Insert(k2, r2, Hamster.HamConst.HAM_OVERWRITE);
            byte[] krslt = k1;
            var f = c.Find(ref krslt, HamConst.HAM_FIND_GT_MATCH);
            checkEqual(k2, krslt); // WORKS FINE
            checkEqual(r2, f);
            db.Close();
            env.Close();

            // remote access
            env = new Hamster.Environment();
            db = new Database();
            env.Open("ham://localhost:8080/ntest.db");
            db = env.OpenDatabase(1, 0);
            c = new Cursor(db);
            db.Insert(k1, r1, Hamster.HamConst.HAM_OVERWRITE);
            db.Insert(k2, r2, Hamster.HamConst.HAM_OVERWRITE);
            krslt = k1;
            f = c.Find(ref krslt, HamConst.HAM_FIND_GT_MATCH);
            checkEqual(k2, krslt);  // FAILS HERE (krslt is unchanged)
            checkEqual(r2, f);
            db.Close();
            env.Close();
        }

The corresponding hamzilla.config file is as follows:

{
    /* global configuration settings */
    "global": {
        "port": 8080
    },

    /* list of hamsterdb Environments that are served */
    "environments": [
        {
            "url": "/ntest.db",
            "path": "./ntest.db",
            "flags": "",
            "databases": [
                {
                    "name": 1,
                    "flags": ""
                }
            ]
        }
    ]
}

This is an internal flag, and ups_cursor_find can fail if the flag is specified. Remove it from the header file, also from the wrappers!

Hello,

I'm trying to compile version 2.1.1 on centos 6 + EPEL (for gperftools).

configure doesn't seems very happy about tcmalloc libraries, because one of the check fails :

checking google/tcmalloc.h usability... yes
checking google/tcmalloc.h presence... yes
checking for google/tcmalloc.h... yes
checking for tc_malloc in -ltcmalloc_minimal... no

But in some way, tc_malloc support is still enabled, because the build fails in sample :

Making all in samples
make[2]: Entering directory `/root/hamsterdb-2.1.1/samples'
  CC     db1.o
  CCLD   db1
../src/.libs/libhamsterdb.so: undefined reference to `tc_free'
../src/.libs/libhamsterdb.so: undefined reference to `tc_malloc'
../src/.libs/libhamsterdb.so: undefined reference to `tc_calloc'
../src/.libs/libhamsterdb.so: undefined reference to `tc_realloc'
../src/.libs/libhamsterdb.so: undefined reference to `MallocExtension::instance()'
collect2: ld returned 1 exit status
make[2]: *** [db1] Error 1
make[2]: Leaving directory `/root/hamsterdb-2.1.1/samples'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/root/hamsterdb-2.1.1'
make: *** [all] Error 2

gperftools are from EPEL. Synced from upstream svn r219.

If the email text includes a

<--

then the mail that is sent is truncated.

Hello Christoph.
I use hamsterdb with some tests and have question.
I dont shrink DB file after create db, records inserts,delete db and close Env.
If i do loop with records insert-delete file growth continuously.
Any comments?

Thanks, Vladimir.

when trying to build for osx (using homebrew libs), build fails with the following. works fine on linux... thoughts?

Undefined symbols for architecture x86_64:
"boost::system::system_category()", referenced from:
global constructors keyed to _ZN12_GLOBAL__N_12_1Ein ham_info.o
"boost::system::generic_category()", referenced from:
global constructors keyed to _ZN12_GLOBAL__N_12_1Ein ham_info.o
ld: symbol(s) not found for architecture x86_64

Was able to work around the issue with the following:
./configure LIBS="-lboost_system-mt -lgnutls"