pointrelnotes.txt

vxworks下的一个嵌入式实时数据库,可以分析源码.

    recovery. When used together with -x convert or -x autoconvert option 
    a backup version of a database can be converted to a newer version. 
    It is still recommended to remove log files before starting the 
    conversion.

SHORT DESCRIPTION: HSBG2: Support for detecting incomplete hsb netcopy 
    or copy	 
TPR#: none
LONG DESCRIPTION: If hsb netcopy or copy is not completed successfully 
    server will exit with return code 24. Client connections to a server 
    with backup server mode on will now fail with a new error code 14552. 
    Also odbc driver is changed to store native return code from the 
    server during a failing connect so that clients can detect that connect
     failed because of backup server mode.

SHORT DESCRIPTION: JDBC driver statement cachesize can be configured.
TPR#: none
LONG DESCRIPTION: If this feature is used then also latest server 
    04.10.0159be must be used. Some older server versions have some 
    re-execute problems (f.x. 04.10.0139be).

    In JDBC driver statement cache size is now configurable using
    connection properties. E.g.

    Properties props = new Properties();
    props.put("StatementCache","32");
    java.sql.Connection conn = 
            java.sql.DriverManager.getConnection(urlstr, props);

    Default statementcache for 4.1 is not changed (it is 1).
    If value for this property is zero then default is used.
    Minimun value is 1 and maximum is 512.


PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.157
BUGS FIXED..: 5
PLATFORMS...: w32, l2x, s9x, h1x, h1x64, hia64, lsh4bx
RELEASE DATE: May 24, 2004

SHORT DESCRIPTION: Serious connect error in Solid 4.10.153 
TPR#: 11654
LONG DESCRIPTION: The fix to TPR 870025 introduced two new bugs: 
    - with zero timeout, the ses_gethostbyname caused a bad memory
      free (and other problems), 
    - with any timeout the ses_gethostbyname would overwrite the port 
      number of the sockaddr, causing the connection to fail
      practically always.

    These problems are now fixed.

SHORT DESCRIPTION: When PreparedStatement method is used to store data
    into VARBINARY column and the size of the data is not correctly set. 
TPR#: 11641
LONG DESCRIPTION: The cause of the problem was an erroneous comparation
    used to detect if the data must be stored as a blob.

SHORT DESCRIPTION: Solid accelerator library causes segmentation fault 
    on Linux.
TPR#: 380044
LONG DESCRIPTION:  

SHORT DESCRIPTION:  Delete trigger caused "incorrect values for insert" 
    to be returned when the replica table had a delete trigger and it 
    received subscriptions from the master.
TPR#: 451879
LONG DESCRIPTION:   

SHORT DESCRIPTION:  Master stores versions of some old replicas that 
    are not registered anymore.
TPR#: 750077
LONG DESCRIPTION: A check is added to history cleanup functionality to make 
    sure that the version info stored belongs to a replica that is still 
    being registered in the master.

PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.153
BUGS FIXED..: 11
PLATFORMS...: w32, l2x, s9x, h1x, h1x64, hia64
RELEASE DATE: April 30, 2004

SHORT DESCRIPTION: PreparedStatement method setNull works incorrectly.
TPR#: 11617
LONG DESCRIPTION: Once a parameter value is set using setNull method, 
    setting the value again has no effect. Now the value is cleared
    correctly when it is set.

SHORT DESCRIPTION: Re-executed SQL statement returns wrong number of rows.
TPR#: 50153, 860056, 452159
LONG DESCRIPTION: Partial reset of a sql cursor loses information when used
    with queries using DISTICT, ORDER BY or DESC. Now the reset works 
    correctly.

SHORT DESCRIPTION: Unregistering a replica with the procedure 
    sync_unregister_replica causes Error 25093.
TPR#: 451765
LONG DESCRIPTION: Now sync_unregister_replica code returns error if the master 
    node is not available: 
    SOLID Communication Error 21306: Server ''TCP 1315'' not found, 
    connection failed

SHORT DESCRIPTION: SOLID accelerator on VXWORKS does not find solid ini file.
TPR#: 452008
LONG DESCRIPTION: SOLID accelerator on VXWORKS does not use the ini file present 
    in the directory specified by the -xpathprefix option but instead uses the 
    default values for the parameters, using -c along with -xpathprefix makes 
    SOLID use the correct ini file but vxworks this changes the directory of the shell
    to the one specified in the -c option. Now SOLID accelerator behaves the same way
    as other versions.
  
SHORT DESCRIPTION: SOLID BoostEngine library defines the symbol "s".
TPR#: 452021
LONG DESCRIPTION: SOLID BoostEngine library which defines the symbol 
    "s" causes problem on VXWORKS by overwriting the vxWorks symbol for debugger 
     step command. Now all defines have been checked and renamed if necessary.

SHORT DESCRIPTION: SOLID Accelerator timer errors during shutdown on VXWORKS.
TPR#: 452043
LONG DESCRIPTION: The shutdown logic has been modified to avoid dropping
    statements during shutdown.

SHORT DESCRIPTION: Parameter checkpointdeletelog is missing.
TPR#: 452160
LONG DESCRIPTION: admin command 'par general.checkpointdeletelog' 
    returns 'no such parameter' and 
    admin command 'par' does not list the this parameter.
    Now the parameter is available to users.

SHORT DESCRIPTION: Warning: Unregistered parameter HotStandby.DurableLogRowslimit
TPR#: 452164
LONG DESCRIPTION: Warning: Unregistered parameter HotStandby.DurableLogRowslimit is
    displayed each time SOLID is started.
    Now the parameter is properly registered.

SHORT DESCRIPTION: Curtime and now funtion millisecond resolution does not work 
    with build 4.10.145
TPR#: 452176
LONG DESCRIPTION: Now all platforms support this feature.

SHORT DESCRIPTION: SOLID 4.10.145 asserts at 1156@snc1blob.c
TPR#: 452187
LONG DESCRIPTION: Fixed. Blob handling was not able to handle READ-ONLY case 
correctly (e.g. secondary).

SHORT DESCRIPTION: MESSAGE..FORWARD TO...TIMEOUT x doesn't work
TPR#: 870025
LONG DESCRIPTION: The message timeout now affects both gethostbyname and connect, although 
the timeout is given separately to each call, so that the total timeout is not fully accurate.  
The fix is only available for UNIX platforms at the moment.

PRODUCT ENHANCEMENTS
--------------------

SHORT DESCRIPTION: Sequencer object does not support BIGINT. 
TPR#: 50150
LONG DESCRIPTION: Now sequencer objects returns BIGINT values.

SHORT DESCRIPTION: Enchancement request to increase the granularity of 
    timestamp in soltrace.out
TPR#: 452169
LONG DESCRIPTION: Added a new ini parameter to control the granularity:
    [Srv]
    TraceSecDecimals=0

    Allowed values= {0,1,2,3}
    Default=0
   
    Description: the number of second decimals displayed in the
    timestamps, in the trace (soltrace.out) files.
    The accuracy of timestamping may be thus set between 1 s
    (default) and 1 ms.

    Access mode: RW (i.e. dynamic)

PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.0147
PLATFORMS...: h1x64, hia64
RELEASE DATE: March 22, 2004

PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.0145
PLATFORMS...: s8x, hia
RELEASE DATE: April 29, 2004

PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.0145
PLATFORMS...: w32, l2x, h1x, s9x, s8x, hia
RELEASE DATE: March 3, 2004

SHORT DESCRIPTION: ADO access to Solid fails with error
    "[HY010] Function Sequence Error (-1)"
    when SQLExtendedFetch is called.
TPR#: 20169
LONG DESCRIPTION: Ado seems to use row-wise binding in a way that
    is conflicting with ODBC specs:
    it passes the pointer to the buffer as ROW_BIND_OFFSET_PTR
    attribute, which is integer per ODBC specs, and passes offsets 
    as BindCol parameters, where pointers are expected.

SHORT DESCRIPTION: When a zero length varbinary value is retrieved 
    from database, fetch returns a -1 length, signifying NULL value 
    in the database.
TPR#: 20170
LONG DESCRIPTION: Now ODBC drivers handles zero length binary strings
    correctly.

SHORT DESCRIPTION: Query from LONG WVARCHAR field returns the full 
    string length, instead of a shorted defined length in SQLBindCol. 
TPR#: 20173
LONG DESCRIPTION: Now WVARCHAR data is null terminated correctly when
    the data needs to be truncated.

SHORT DESCRIPTION: Conversion SQL_TIMESTAMP -> SQL_[W]VARCHAR does not
     work for fractions as expected.
TPR#: 20175
LONG DESCRIPTION: Now SQLBindParameter checks the precision correctly and
     data is stored as requested.

SHORT DESCRIPTION: JDBC Driver: getObject returns Integer for SMALLINTs.
TPR#: 50140
LONG DESCRIPTION: Now driver casts the data correctly before returning
     the data the client application.

SHORT DESCRIPTION: Subquery optimization
TPR#: 451977
LONG DESCRIPTION: Now subqueries using pseudocolumn ROWNUM work faster. 
     
SHORT DESCRIPTION: Unhandled expecption when executing a query.
TPR#: 452019
LONG DESCRIPTION: Queries such as 
     "SELECT id from bar where name = ? or name = 'unknown'"
      work now correctly.

SHORT DESCRIPTION: SQL performance problem
TPR#: 452056
LONG DESCRIPTION: Nested queries such as 
     "select * from ALIAS_T where EP_ID in (select EP_ID from ENDPOINT_T where
         EP_NAME like '5618600029') order by ALIAS_ID"
     work now faster.

SHORT DESCRIPTION: SOLID asserts with error 1537@rs0atype.c when attempting to 
     create a procedure that posts an event with parmeters but the event does 
     not accept parameters.
TPR#: 452086
LONG DESCRIPTION: Now the checks can handle this situation correctly.

SHORT DESCRIPTION: Solid accelerator library dumps core
TPR#: 452119
LONG DESCRIPTION: This happened when start after commit statement is executed 
     after the commit although it failed with readonly error. Now this is
     handled correctly ie. start after commit is not executed if there has been
     an error in commit.

SHORT DESCRIPTION: "Lost Data" in HSB environment occures when the OS is rebooted.
TPR#: 870023
LONG DESCRIPTION: The scenario that can lead to "lost data" is:
     1) execute a transaction on PRIMARY_ACTIVE -node
     2) crash PRIMARY
     3) switch SECONDARY to new PRIMARY
     4) crash that new PRIMARY or shutdown the new PRIMARY
     5) re-start servers and make new PRIMARY to PRIMARY
     6) try to do hsb connect
     Step 6) fails:databases are out-of-sync.
     Now this is fixed.

OTHER FIXES
-----------

1. JDBC driver handles now handles correctly commits while fetching a resultset.

PRODUCT ENHANCEMENTS
--------------------

SHORT DESCRIPTION: Solid Light Client for Linux
TPR#: 50144
LONG DESCRIPTION: Now this product is part the default product package.



PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.0142
PLATFORMS...: w32, l2x, h1x, s9x, h1x64, hia64
RELEASE DATE: March 3, 2004

SHORT DESCRIPTION: Segmentation fault in SQLConnect during OS startup.
TPR#: 11497
LONG DESCRIPTION: It is possible that when two SQLConnect calls are being processed 
    simultanously the mutexing does not work as it should. Now this is fixed.

SHORT DESCRIPTION: Parameters is not passed correctly.
TPR#: 11517
LONG DESCRIPTION: Parameter passing machanism in the ODBC driver fixed.

SHORT DESCRIPTION: SQLPutData() results in a endless loop.
TPR#: 20161, 50146
LONG DESCRIPTION: Fixed.

SHORT DESCRIPTION: SQLGetInfo maxLen value
TPR#: 20162
LONG DESCRIPTION: SQLGetInfo(hdbc, i, &strVal, maxLen, &recLen) does not work 
if maxLen is too large, but still within ODBC specs. Fixed.

SHORT DESCRIPTION: Problem with ODBC SQLGetData
TPR#: 20164
LONG DESCRIPTION: Internal error in SQLGetData, SQLstate=HY000. Unbinding removed a row descriptor and SQLFetch didn't recreate it thus causing internal error in SQLGetData. Fixed.

SHORT DESCRIPTION: Deadlock with MME tables
TPR 20165
LONG DESCRIPTION: The deadlock when updating a single row in an M-table through 
two different indexes in two different transactions is now fixed.

SHORT DESCRIPTION: Unicode string lengths wrong with v4.1 ODBC
TPR#: 20167
LONG DESCRIPTION: Unicode string retrieval brings wrong bytelength with 
v4.1 ODBC driver. Fixed.

SHORT DESCRIPTION: ODBC function SQLDescribeParam incorrectly returns -1 as length.
TPR#: 50142
LONG DESCRIPTION: When SQLDescribeParam returns -1 SQLBindParameter fails due to 
    wrong buffer length.

SHORT DESCRIPTION: Problem with fetching decimal values with ODBC 
TPR#: 451937
LONG DESCRIPTION: Truncated decimal value is now also assigned to binded variable.

SHORT DESCRIPTION: 4.1 Solid asserts with 5155@tab1dd.c when starting up the engine.
TPR#: 451970
LONG DESCRIPTION: Some diskless system tables were erroneously not be given system 
    table priviledges but full access. So it is possible to drop these tables using
    DROP TABLE statement.

SHORT DESCRIPTION: HSB connect fails with error 
        "SOLID Server Error 14706: Invalid read thread mode for HotStandby, 
            only mode 2 is supported"
    after "Failed to create thread 'dnet_clientthread'" warning is received.
TPR#: 452033
LONG DESCRIPTION: Now HSB is enabled without thread/client. If incoming hsb 
    connection does not have thread/client flag set it uses one tasking system 
    thread for hsb connection.

SHORT DESCRIPTION: SQL subquery that returns a NULL row value is not handled properly
    may cause a server crash. 
TPR#: 452062
LONG DESCRIPTION: Now SQL subqueries work correctly.

SHORT DESCRIPTION: Command 'hsb switch primary' returns: "14537 HotStandby connection is broken".