pointrelnotes.txt

vxworks下的一个嵌入式实时数据库,可以分析源码.

SHORT DESCRIPTION: SQL_ATTR_QUERY_TIMEOUT does not work as expected: 1) multiple 
    iterations of the same query aborted by querytimeout may cause the server 
    to go into a busy loop and 2) the timeout does not always work with pessimistic tables.
TPR #: 452225
LONG DESCRIPTION: Pessimistic locks have a default 30 second timeout, which caused some extra 
    delay to querytimeout. Also the busyloop inside SQL interpreter is fixed.

SHORT DESCRIPTION: Unregistering to a publication through call SYNC_UNREGISTER_PUBLICATION 
    system procedure with the delete data flag on fails with syntax error if the publication
    accepts parameters.
TPR #: 452357
LONG DESCRIPTION: Statement "DROP SUBSCRIPTION ..." lacked the parameter list 
    of the subscription meant to be dropped. 
    Added the keyword "ALL" to the statement in question. 
    Now the procedure SYNC_UNREGISTER_PUBLICATION should works as specified.

SHORT DESCRIPTION: Database failure due to transaction logs not being deleted.
TPR #: 452379
LONG DESCRIPTION: Now netcopy (remote backup) now tries to delete all local 
    log files.

SHORT DESCRIPTION: Inconsistent log messages in solmsg.out.
TPR #: 940020
LONG DESCRIPTION: Now the emergency exit and failed startup generate clear
    messages:
        <SRV30149> Server emergency shutdown
    and
        <SRV30150> Server failed to startup

    The new message "Database started" have been added to denote that the database 
    is ready to server clients.


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

SHORT DESCRIPTION: Enhancement request for SOLID to post an event when the 
    ImdbLowPercentage limit set in the ini file is reached
TPR #: 452276
LONG DESCRIPTION: There are now two new system events: "SYS_EVENT_MEMORY_LOW" 
    and "SYS_EVENT_MEMORY_BACKTONORMAL". The former indicates that the amount 
    of available memory has sunk below the ImdbMemoryLowPercentage limit. 
    Correspondingly, event "SYS_EVENT_MEMORY_BACKTONORMAL" informs that the 
    amount of available memory has returned back to 'normal' level.

SHORT DESCRIPTION: Enhancement request for a command to simulate a server 
    assertion.
TPR #: 940014
LONG DESCRIPTION: A new command ADMIN COMMAND 'assertexit' causes assertion 
    failure in the server. Before assert the following message is printed to 
    the solmsg.out file:
        "Exiting server with ADMIN COMMAND 'assertexit'..."
    Message number is <SRV30124>.


PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.194
BUGS FIXED..: 1 (#452410)
PLATFORMS...: l2x
RELEASE DATE: September 17, 2004

SHORT DESCRIPTION: Failing connection may cause SOLID Communication Error 21309.
TPR #: 452410
LONG DESCRIPTION: A fix was done to free recources correctly if a connection
     fails. Earlier they were left as garbage until server shutdown.

PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.191
BUGS FIXED..: 3 (#11720, 50176, 452224)
PLATFORMS...: w32, l2x, jdbc
RELEASE DATE: August 31, 2004

SHORT DESCRIPTION: SQLConnect loops in the client library.
TPR #: 11720
LONG DESCRIPTION: Now a 100 milliseconds wait has been added between 
    consecutive connect attempts.

SHORT DESCRIPTION: Bug in Solid JDBC driver - getTimestamp() cannot access
    a DATE column.
TPR #: 50176
LONG DESCRIPTION: This particular feature was now implemented.

SHORT DESCRIPTION: Dense sequence object does not work as expected. Sometimes
    a sequence number may be lost if the result set used to fetch a sequence
    is not closed before executing the next sequence operation. 
TPR #: 452224
LONG DESCRIPTION: Sequence operations and rows per message feature caused this 
    problem. Now any sequence operation disables rows per message.

PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.187
BUGS FIXED..: 2 (#451691, 452318)
PLATFORMS...: l2x
RELEASE DATE: August 9, 2004

SHORT DESCRIPTION: SOLID 4.x asserts with 2244@tab0conn.c 
    Code: 10020 (DBE_ERR_CPACT) when machine is powered down.
TPR #: 451691
LONG DESCRIPTION: SIGTERM signal handling is fixed. Now SIGTERM causes 
    (indirectly) a call to sqlsrv_shutdown_force which waits for a 
    checkpoint to complete before shutdown.

SHORT DESCRIPTION: In a HSB scenario with 2safedurable, rows are missing 
    from both the primary and the secondary on a reboot (double point of
    failure).
TPR #: 452318
LONG DESCRIPTION: Now double point of failure is HSB recoverable within 
    parameter 'RelaxedMaxDelay'. Eg. system generates durable logrec if 
    there are any non-recoverable commit.

PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.186
BUGS FIXED..: 2 (#11733, 11731)
PLATFORMS...: w32, l2x
RELEASE DATE: August 4, 2004

SHORT DESCRIPTION: SQL estimator chooses a wrong plan.
TPR #: 11715, 11731
LONG DESCRIPTION: External sorter cost estimate is improved. Now it should 
    give better estimate that includes also disk access time. Also a new 
    configuration parameter 
        [Sorter]
        SorterEnabled=y/n
    is added. Its default is yes.

PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.185
BUGS FIXED..: 1 (#11733)
PLATFORMS...: w32, l2x
RELEASE DATE: July 28, 2004

SHORT DESCRIPTION: Using the function convert_char in the query where INNER JOIN 
    is used may cause the server to crash
TPR #: 11733
LONG DESCRIPTION: Here is an example of the problematic query:
    select * from players inner join penalties on (players.playerno=penalties.playerno) 
    and convert_char((select teamno from teams where teamno = '1')) = players.playerno;

    Now the correct error message is returned:
    SOLID SQL Error 59: comparison between incompatible types

PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.183
BUGS FIXED..: 5 (#11694,11695,11705,20179,50174)
PLATFORMS...: w32, l2x
RELEASE DATE: July 28, 2004

SHORT DESCRIPTION: TPR 11641 not completely fixed
TPR #: 11694
LONG DESCRIPTION: Internal storage of SolidTB (class handling binary column data) 
    was not properly reset after binding new parameter value, hence the problem. 
    Now this is fixed.

SHORT DESCRIPTION: Message: "Warning: Unregistered parameter Srv.ReportInterval is used." 
displayed in solmsg.out at startup, although no such parameter is defined in solid.ini.
TPR #: 11695

SHORT DESCRIPTION: Solid JDBC driver handles WVARCHAR buffers faulty.
TPR #: 11705
LONG DESCRIPTION: There was a bug in ssaj/s_AttrValUcode.j file: byte count 
    was used instead of (wide)character count when determining how many 
    (wide)characters were to be written into the network.
    Also a bug in ssac/ssa0java.c was fixed, where unicode characters 
    (read as bytes) in nativecall_Execute function were converted to SSACHAR 
    incorrectly.

SHORT DESCRIPTION: Solid halts for a few seconds (2s-15s) during heavy update 
load but otherwise normal operation.
TPR #: 20179
LONG DESCRIPTION: Nowindex (db) file uses synchronized IO 
    (SS_BF_SYNCWRITE, operating system flag O_SYNC) by default.  This evens 
    out the IO significantly during checkpoints.  
    A new solid.ini parameter 
        [IndexFile]
        SynchronizedWrite 
    controls the behavior, 
    and is Yes by default.

SHORT DESCRIPTION: Solid ODBC driver is unable to store a timestamp into the 
    database in the 'yyyy-mm-dd hh:mm:ss.f' format.
TPR #: 50174
LONG DESCRIPTION: ODBC driver prints second fraction when converting a 
    timestamp to a string. For the conversion from a string to a timestamp
    there was no support for second fractions.
    Now Parsing of second fraction has been implemented.

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

SHORT DESCRIPTION: The format of the message log (solmsg.out) is improved to 
    allow effective automatic processing.
TPR #: 940018
LONG DESCRIPTION: The following changes are implemeted:

    1. Tagging of the message code with escape characters "<" and ">", e.g.:
    25.06 14:39:24 <SRV10081> The database file is incomplete or corrupt. 

    2. The following three messages will denote successful or unsuccesful startup, 
    or server run:

    A. The message after a successful startup:
    26.06 17:09:14 <SRV30119> Solid BoostEngine started at Sat Jun 26 17:09:14 2004

    B. The message after a successful shutdown:
    26.06 17:20:14  <SRV30105> SOLID BoostEngine is shut down.

    C. The message after a controlled shudown (including general assert) or unsuccessful startup
    26.06 17:20:14 <SRV30149> Server emergency shutdown.

    3. Specific failure message. Before the emergency shutdown message, a message explaining 
    the cause, like in:
    26.06 17:20:14 <DBE30254> Database is a broken HSB copy or netcopy database.
    26.06 17:20:14 <SRV30149> Server emergency shutdown.

PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.179
BUGS FIXED..: 1 (#11724)
PLATFORMS...: w32, l2x
RELEASE DATE: July 15, 2004

SHORT DESCRIPTION: INNER JOIN may return an erroneous resultset
TPR #: 11724
LONG DESCRIPTION:  INNER JOIN of two tables messes up the result set 
    when columns when the SELECT list contains subqueries to a third table


PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.175
BUGS FIXED..: 1 (#452282)
PLATFORMS...: l2x
RELEASE DATE: July 9, 2004

SHORT DESCRIPTION: Version 04.10.0170 dumps core when a netcopy is done 
    from from version 04.10.0145 during hot upgrade
TPR #: 452282
LONG DESCRIPTION: The problem is that version 04.10.0170 erroneously
    detects the database to be a broken netcopy database. Now the 
    detection works correctly.

PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.170
BUGS FIXED..: 4 (#452229, #50174, #11604, #452256/#50171)
PLATFORMS...: w32, l2x, jdbc
RELEASE DATE: June 22, 2004

SHORT DESCRIPTION: SYNC_MASTER_MESSAGE_ERROR_OCCURRED event not posted.
TPR #: 452229
LONG DESCRIPTION: The event SYNC_MASTER_MESSAGE_ERROR_OCCURRED was not
    posted in the master when a message failed.

SHORT DESCRIPTION: JDBC addBatch did not work.
TPR #: 452256, 50171
LONG DESCRIPTION: JDBC PreparedStatement.addBatch function was
    keeping the last set parameters and using it for all set
    of batch updates. Previously set parameters were lost.

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

SHORT DESCRIPTION: Compatibility option for timestamps to act like
    timestamps in version 4.0 or earlier (ODBC).
TPR #: 50174
LONG DESCRIPTION: A compatibility option was added to provide similar
    interface to timestamps as was with Solid versions 4.0 and earlier.

    [SQL]
    TimestampDisplaySize19=no     (default; set to Yes to enable)

    Forces the server to return the value "19" as a result of the
    display size query (in ODBC), for any timestamp column. This
    corresponds to the precision of 1 second. This behavior was
    present in v. 4.0 and earlier. The value of the parameter does not
    affect the way timestamps are stored in the database, only the
    output. Also, it is possible to retrieve the timestamps with full
    precision by providing a buffer of a sufficient length (29 in
    versions after 4.0).

SHORT DESCRIPTION: Support for blanks in filenames (or strings in quotes).
TPR #: 11604
LONG DESCRIPTION: It is now possible to have blanks in filenames by
    surrounding the filename with quotes.  For example: 

    [Indexfile]
    Filespec_1 = "C:\This here is separated\And\So is this.db" 1000000

PRODUCT.....: BoostEngine 4.1
VERSION.....: 04.10.159
BUGS FIXED..: 4
PLATFORMS...: w32, l2x, s9x, h1x
RELEASE DATE: June 5, 2004

BUG FIXES
---------

SHORT DESCRIPTION: SQL_ATTR_QUERY_TIMEOUT set by SQLSetStmtAttr function 
    affects randomly on fetches
TPR#: 11679
LONG DESCRIPTION: The query timeout implementation is changed so that 
    only SQLExecute can timeout. Also there was a bug that caused 
    the random timeouts during fetch.

SHORT DESCRIPTION: Cached statement returning occasionally wrong data
TPR#: 50169
LONG DESCRIPTION: Database may return wrong data in the following scenario
    1. select statement with parameter marker is prepared
    2. parameter marker is assigned a value that is supposed to return no rows
    3. database returns 0 rows correctly
    4. statement is closed, but it remains in the statement cache
    5. statement is programmatically prepared again (but in the server
       end, due to statement caching)
    6. parameter is assigned a new value, which is supposed to return one row
    7. however, database still returns 0 rows.

    Now the statement cache works correctly.
 

SHORT DESCRIPTION: Long exclusive table lock does not work as expected 
    when DDL is followed by DML transaction in the same session
TPR#: 710054
LONG DESCRIPTION: The problem happens when a DDL command adds a column 
    to a table, and a DML command adds default values to new column.
    This results in error message when update is executed: 
        SOLID Database Error 10014: Resource is locked
    The lock conflict has been fixed.

SHORT DESCRIPTION: Secondary asserts at 325@hsb0secexec.c during
    catchup 
TPR#: 300012

LONG DESCRIPTION: If secondary has started with corrupted logfiles
    roll-forward recovery may have automatically skipped corrupted part of
    the logfiles. This causes an assertion 325@hsb0secexec.c when primary
    tries to connect because secondary does not have all the transactions
    it should. Now secondary can switch to out of sync state if it
    recognizes a logfile corruption. Primary must perform a netcopy before
    it can connect again.
    

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

SHORT DESCRIPTION: Possibility to skip recovery in order to convert a 
    backup database 
TPR#: 11672
LONG DESCRIPTION: Now option -x nologrecovery ignores log files during