SYBASE Adaptive Server Anywhere
*******************************

Bug Fix Readme for Version 7.0.2
********************************

Description of available downloads
==================================

    Emergency Bug Fixes
    ===================
    A subset of the software with one or more bug fixes. The bug fixes are 
    listed in the "readme" file for the update. A Bug Fix update may only be
    applied to installed software with the same version number.

    Some testing has been performed on the software, but full testing has not
    been performed. Customers are discouraged from distributing these files
    with their application unless they verify the suitability of the software
    themselves.

    Verified Bug Fixes
    ==================
    A subset of the software with one or more bug fixes. The bug fixes are 
    listed in the "readme" file for the update. A Bug Fix update may only be
    applied to installed software with the same version number.

    Full testing has been performed on the software.

    Maintenance Release
    ===================
    A complete set of software that upgrades installed software from an older
    version with the same major version number (version number format is 
    major.minor.patch). Bug fixes and other changes are listed in the "readme"
    file for the upgrade.

    After the maintenance release upgrade has been applied, the latest
    Verified or Emergency Bug Fixes update should be applied, if available.

Bug Fixes
=============
The following is a list of bugs fixed in this collection of software,
as compared to the 7.0.2 Maintenance Release.



                      ******************************                      
                      ** Adaptive Server Anywhere **                      
                      **                          ** 
                      **   DBLIB Client Library   **                      
                      ******************************                      

=================(Build #1451  - Engineering Case #230395)=================

The client could have crashed or given a communication error, (-85).  
This was caused by a call to alloc_sqlda(n) with a parameter greater 
than 32K. The sqlda->sqln variable is defined as a short, as a result 
the number of descriptors allowed should not be greater than 32K. We 
were allowing an allocation of greater than 32K descriptors.
NOTE: The documentation is missing that a NULL pointer is returned 
when the requested number of descriptors exceeds 32K.


=================(Build #1451  - Engineering Case #230637)=================

On win32 platforms, if the asasrv.ini file did not exist, it would have 
been created in C:\Program Files\Common Files\ODBC and called "Data 
Sourcesasasrv.ini" rather than being created in the %ASANY% directory. 
This has been fixed.



=================(Build #1451  - Engineering Case #231867)=================

A local TCP/IP connection may have been terminated when doing a large 
fetch. The same fetch over shared memory or a TCP/IP connection from 
a different machine, worked fine.


=================(Build #1453  - Engineering Case #233442)=================

If xp_write_file was used to write two strings to the same file and 
the first string written was longer than the second, data from the first 
string would be left in the file beyond the end of the second string.
 For example,
call xp_write_file( 'out.txt', 'aa' );
call xp_write_file( 'out.txt', 'b' );
would leave out.txt containing the string 'ba'.


=================(Build #1458  - Engineering Case #234645)=================

If autostarting a database on an existing server failed (if, for example, 
the database file did not exist), the client could have crashed. If 
the server was also autostarted, the crash did not occur, nor did it 
occur if starting the database succeeded.



=================(Build #1460  - Engineering Case #234874)=================

When using SQLGetData or ESQL SQL GET DATA to fetch all columns in a 
result set, and at least one of the GetData's was attempting to fetch 
300 bytes or more or prefetch was disabled, then using GetData to retrieve 
the columns less than 300 bytes would be slow, especially over a network 
connection.



=================(Build #1466  - Engineering Case #235756)=================

Numeric connection parameters would accept negative values, causing 
various problems. The parameters affected were:
LivenessTimeout
CommBufferSize
PrefetchRows (6.0.3.3031+, 7.0.0.450+, 7.0.1.1101+, 7.0.2 and up)
PrefetchBuffers (6.0.3.3031+, 7.0.0.450+, 7.0.1.1101+, 7.0.2 and up)

If negative values are entered for any of these parameters, a parser 
error is now displayed.

Also, the engine switches -ti and -tl would accept negative values. 
This will now display the usage screen.


=================(Build #1487  - Engineering Case #238285)=================

A Unix client, connecting through either ESQL or ODBC, would experience 
a memory leak if it executed multiple calls to db_init()...db_fini(), 
or SQLAllocEnv()...SQLFreeEnv().  The client could also have hung.	





                      ******************************                      
                      ** Adaptive Server Anywhere **                      
                      **                          ** 
                      **   ODBC Client Library    **                      
                      ******************************                      

=================(Build #1472  - Engineering Case #236630)=================

When attempting to translate or count the length of a UTF8 character 
whose leading byte indicated it was a 4,5, or 6-byte multi-byte character, 
the ODBC driver could go into an infinite loop.


=================(Build #1483  - Engineering Case #238243)=================

If a UNIX application loaded the ODBC driver library, created an environment, 
connected, (or at least attempted to connect) at least once, and then 
unloaded the odbc library without destroying the environment, the application 
could seg fault.

A work-around is to ensure that all ODBC environments are destroyed 
prior to unloading the ODBC driver.




                      ******************************                      
                      ** Adaptive Server Anywhere **                      
                      **                          ** 
                      **   OLEDB Client Library   **                      
                      ******************************                      

=================(Build #1126  - Engineering Case #224155)=================

Executing commands with single byte string input arguments from Delphi 
using the ASA OLEDB provider (ASAProv) would fail to pass any of the 
string arguments through to the engine.  BSTR and WSTR values would 
be passed through without problems.


=================(Build #1448  - Engineering Case #229624)=================

When using the OLEDB driver and calling a stored procedure with output 
parameters, the application  may have crashed.


=================(Build #1448  - Engineering Case #229625)=================

When reading a number of chunks out of a blob or long varchar column 
with a stream object, it was not advancing to read all data.


=================(Build #1450  - Engineering Case #231291)=================

The OLE DB provider, (ASAProv), could have asserted in IDBProperties::GetPropertyInfo() 
when called with a property set containing a single, unsupported, property 
set GUID.


=================(Build #1450  - Engineering Case #231302)=================

This change fixes several  problems with the OLE DB driver on Windows 
CE:
1. Sometimes the OLE DB driver would hang on a database connection.
2. Columns of type DOUBLE, would sometimes produce incorrect results
3. ADO 3.1 recordset.MoveFirst would fail


=================(Build #1450  - Engineering Case #231849)=================

The OLEDB provider, (ASAProv), failed to return string data on Microsoft 
ADO 3.1 when run on WIN CE. Also, this fix solves a problem where, in 
some cases, the provider would hang when opening a database connection.



=================(Build #1460  - Engineering Case #234877)=================

In ADO, updating a recordset row value sometimes failed to show current 
values -- the new values were in the database, but ADO did not see them. 
Also, scanning the same recordset twice caused a provider error after 
the last row (the second time). These problems are now fixed.


=================(Build #1466  - Engineering Case #235951)=================

If a connection fails for a valid reason when using the ASAProv provider, 
no specific error information was being set up so all the user got back 
was a generic error message.



=================(Build #1466  - Engineering Case #235958)=================

When asking for table schema information from the ASAProv provider, 
an error is returned.  This has been fixed.


=================(Build #1468  - Engineering Case #236343)=================

Various problems, which caused crashes when using  Microsoft's RowsetViewer 
with the SQL Anywhere OLE DB provider, have been fixed.


=================(Build #1472  - Engineering Case #236914)=================

Pocket PC applicationa would hang or crash when retrieving string column 
data as a BSTR, when the column contained NULL data and the database 
used the UTF8 character set. This has now ben fixed.



=================(Build #1472  - Engineering Case #236965)=================

Pocket PC applications could hang or crash when retrieving wide column 
data that was internally truncated. Bug was only exhibited when truncations 
occurred in more then one row.This is now fixed


=================(Build #1475  - Engineering Case #237521)=================

An application using ADO, attempting to update a row which had been 
fetched from a dynamic scroll cursor and then had some column values 
changed, would fail.



=================(Build #1482  - Engineering Case #236625)=================

When using the OLEDB driver, the default cursor type has been changed 
from Dynamic to Forward-only. Some ADO applications were expecting Forward-only 
and would thus would fail.



=================(Build #1487  - Engineering Case #213815)=================

When opening a table with BLOB fields, ADODB.Recordset.MoveFirst would 
not be able to retrieve those fields. This has now been fixed.




                      ******************************                      
                      ** Adaptive Server Anywhere **                      
                      **                          ** 
                      **          Other           **                      
                      ******************************                      

=================(Build #1451  - Engineering Case #232421)=================

When using dbsvc -g to get information on a service, the account name 
displayed would be incorrect if a network account was used.


=================(Build #1451  - Engineering Case #232422)=================

Listing available services and getting information about a service should 
not require administrator access, but it did. Now, only administrator 
access is required to add or delete services.



=================(Build #1456  - Engineering Case #233902)=================

An install of the GA, Eval or Authenticated versions of ASA 7.x would 
fail on PA-RISC1.x-based HP machines. The error messages are:  "./bin/gdf:  
Execute permission denied" and "You don't have enough disk space [...] 
Available=0K".  This is due to the gdf executable that is shipped with 
the install, having been accidentally built for PA-RISC2.0 architecture, 
which will not run on PA-RISC1.x machines.  A new, PA-RISC1.1 version 
of gdf was created to solve this problem. Please  contact Tech Support 
to get a correctedcopy of gdf. The work around then is to copy the contents 
of the hp directory from the CD to disk, copy the new gdf executable 
on top of the old one, and run the install from the hard disk.  If the 
user wishes to install the on-line (html) docs, then they will need 
to manually copy them from the CD.
Note that the EBF installs do not have this problem, since they do 
not check available disk space.


=================(Build #1466  - Engineering Case #235757)=================

When using a SELECT statement to fetch data from a proxy table, which 
refers to a Visual Foxpro stand-alone free table, the message: 'Error 
- Command contains unrecognized phrase/keyword.' would be thrown by 
the engine. Now the engine can process these kinds of proxy tables correctly.





                      ******************************                      
                      ** Adaptive Server Anywhere **                      
                      **                          ** 
                      **    Procedure Debugger    **                      
                      ******************************                      

=================(Build #1472  - Engineering Case #236915)=================

Double-clicking in the "Value" or "Data Type" columns of the "Inspect" 
window, may have caused the Procedure Debugger to throw a NullPointerException. 
This has been fixed.





                      ******************************                      
                      ** Adaptive Server Anywhere **                      
                      **                          ** 
                      **          Server          **                      
                      ******************************                      

=================(Build #1144  - Engineering Case #227258)=================

Results from DatabaseMetadata would vanish after commit. This problem 
was common for both server-side jdbc or jConnect clients.



=================(Build #1390  - Engineering Case #230316)=================

In files written by Java ISQL, by clicking the File/Save or File/Save 
As menu items, lines were separated by only a line feed character (0x0a) 
which some text editors (e.g. Windows Notepad) did not interpret as 
a newline. Now lines are separated by the operating system's newline 
character.



=================(Build #1448  - Engineering Case #229432)=================

Using the -gs or -gss switches on the server would set the same value 
for the -gm switch. For example, using -gs 4096 would also implicitly 
add -gm 4096.



=================(Build #1448  - Engineering Case #229627)=================

DBISQL did not accept "YES" as a synonym for "ON" in the AUTOSTOP clause 
of the START DATABASE statement. Now it does; this is the documented 
behaviour.



=================(Build #1449  - Engineering Case #229164)=================

The server could crash while executing a create view statement.


=================(Build #1449  - Engineering Case #229475)=================

When stopping a database, (ie stop database <db> unconditionally), running 
on a Netware server from a client connected via SPX, the database would 
shutdown, but hitting 'q' on the NetWare console to shut down the server, 
would either cause it to hang  or crash after a 10-second delay.


=================(Build #1449  - Engineering Case #229629)=================

A CREATE VIEW statement over a Transact-SQL outer join query would fail 
if either of the tables involved in the join had a CHECK constraint.



=================(Build #1449  - Engineering Case #229876)=================

This change fixes the following two problems:

1 - If the JDBCStatement.setCursorName() was used, the second execution 
of the statement would give a WJ004 exception. Explicitly closing the 
statement, as well removing all references to the statement, did not 
guarantee the cursor name would become available again. The name was 
not freed until the next garbage collection.

2 - The automatically generated cursor name is a hex string. If it 
began with a numeric character, then referring to the cursor name in 
another statement, (say a positioned delete statement), would have resulted 
in a syntax error.



=================(Build #1449  - Engineering Case #230024)=================

The TCP Broadcast parameter was being ignored. This was also a client 
problem, so a new dblib[6/7].dll or dbodbc[6/7].dll would be required.


=================(Build #1449  - Engineering Case #230025)=================

The return value for xp_sendmail was 5 even when the call completed 
successfully.


=================(Build #1450  - Engineering Case #229182)=================

Engine might crash if a stored procedure had been in use by multiple 
connections and separate connection attempted to drop or alter the procedure.


=================(Build #1450  - Engineering Case #230204)=================

Shutting down the ASA Netware server may have caused it to hang, if 
there are active SPX connections.


=================(Build #1450  - Engineering Case #230508)=================

Given a view V defined as:

Create view V as
Select * from T
UNION ALL
Select * from T

the a query:

Select * from V where V.x = abs(10)

The query rewrite rule to push the query predicate into each query 
specification making up the view would fail to fire. Hence, even if 
column X in table T was indexed, T would only have been accessed through 
a sequential scan.

Although, for the query:

Select * from V where V.x = 10

the rewrite rule would fire, and indexed retreival would be considered.



=================(Build #1450  - Engineering Case #231578)=================

An invalid Trasact-SQL query of the form:

select * from product a, product b
where a.id *= b.id
and a.description *= a.description

would put the server into an infinite loop.


=================(Build #1451  - Engineering Case #227868)=================

User tables created by the DBO user were not being populated with data 
when
being unloaded or extracted using the -an or -ac switch.



=================(Build #1451  - Engineering Case #228832)=================

A server executable, processing concurrent requests on Windows NT/2000, 
could have GPFed when a tcp/ip connection disconnected or when was disconnect 
due to an idle timeout.  The first problem was rare on multiprocessor 
machines, and even less likely on single processor machines, but the 
idle timeout problem occured on both single and multiprocessor machines.


=================(Build #1451  - Engineering Case #228974)=================

Grouped queries over view containing subselects in the select could 
have generated an erroneous syntax error.
For example:
    Create view V2 (vempid) as select emp_id from employee;
    Create view V1 (orderdate) as select (select first order_date from 
sales_order where sales_rep = vempid) from V2;

Query (A):
    Select orderdate from V1 group by orderdate
would execute correctly, but:
    Select V1.orderdate from V1 group by V1.orderdate 
would not. Utilizing a select list alias, as in

Query (B):
    Select V1.orderdate as foo from V1 group by foo
was a useful workaround.


=================(Build #1451  - Engineering Case #228982)=================

A non-shared memory client-application could hang during a request if 
a liveness packet was sent at the same time.  Liveness packets are sent 
automatically every (liveness timeout time)/4 seconds when the connection 
is idle.  Liveness packets are not necessary on shared memory or named 
pipes links.


=================(Build #1451  - Engineering Case #229868)=================

Concatenation in a subquery could cause a crash. For example, the following 
query on asademo.db would crash:
	SELECT  emp_lname,
 	       (select max( distinct dept_name || ' ')
		 from Department
		 where  Department.dept_id = Employee.dept_id
		 and	Department.dept_id=100 )
	FROM Employee



=================(Build #1451  - Engineering Case #230027)=================

An ASA Server on Windows NT/2000 could have GPFed when a SPX connection 
disconnected.  This problem was rare on multiprocessor machines, and 
even less likely on single processor machines.



=================(Build #1451  - Engineering Case #230348)=================

Assertion failure 50102, in 5.x, or 101504, in 6.x and 7.x, "Memory 
allocation size too large", could be reported when the engine was dealing 
with value sets of certain particular sizes. Wide inserts to a 5.x server 
could be affected by this bug but not wide inserts to a 6.x or 7.x server.


=================(Build #1451  - Engineering Case #230919)=================

A database server would hang on startup if it was started as a service 
that was not allowed to interact with the desktop, with a command line 
containing no parameters. Now, the server will fail to start.


=================(Build #1451  - Engineering Case #231031)=================

The database server could hang and stop responding to requests while 
a backup was in progress


=================(Build #1451  - Engineering Case #231751)=================

Attempting to execute a query like:
	select * from employee where (emp_id=1234,'1234')
where employee is a local table results in an "Illegal user selectivity 
estimate specified" warning; however,
executing the same query when employee is a remote table would end 
up crashing the engine. You
now get the same warning but the query will NOT be executed.


=================(Build #1451  - Engineering Case #231868)=================

The statement:
	comment on java jar 'jar-name' is 'comment-string'
was being logged as:
	comment on java class 'jar-name' is 'comment-string'
This would cause recovery to fail.


=================(Build #1451  - Engineering Case #232420)=================

Running a server on Windows 9x, with a working directory different than 
the directory containing the executable, could have produced the following 
message multiple times:

  "Warning: Unable to load dbmem.vxd.  Dynamic cache sizing is disabled."

even though dbmem.vxd appeared in the same directory as the executable. 
VXDs are loaded using the CreateFile API, which only checks the path 
and the current working directory for the VXD, not the directory containing 
the executable.  Thus, if the VXD was not in the path and the engine 
was launched outside of its directory, the VXD would not be loaded.  
Code was added to to check the directory containing the executable for
the VXD instead of using the default behaviour of CreateFile.


=================(Build #1451  - Engineering Case #232975)=================

Attempting to execute a query like:

	SELECT  YEARS('1998-07-03') FROM DUMMY

on a big endian machine would yield a reult like 130940928 rather than 
the expected result of 1998.


=================(Build #1451  - Engineering Case #232983)=================

The query:
	select * from SYS.DUMMY where 'a' = 'A'
would return no rows in a case-insensitive database; however, one row 
should be returned.


=================(Build #1452  - Engineering Case #233135)=================

The following use of the replace() function would have caused a crash 
or a loop accessing the temporary file:
	select replace( repeat('x',300) || '...' || 'yyyy' || '...' || 'zzzz', 
'.', '----')
The string being replaced must have been a single character, and two 
or more adjacent occurrences of the string must have existed in the 
original string.


=================(Build #1452  - Engineering Case #233145)=================

If a database file accessed via a network share was marked as a read-only 
file (the normal state of a database file after a clean shut-down), 
a subsequent attempt to start that database would result in a "permission 
denied" error. A simple workaround is to mark the file as writable prior 
to starting the database (eg. "attrib -r database.db").

NOTE: Sybase does not support the use of database files that are accessed 
via a network share and cannot guarantee the recoverability of databases 
accessed in this way.


=================(Build #1453  - Engineering Case #223631)=================

At the end of a successful execution, the 'ALTER DATABASE UPGRADE' statement 
will now perform an automatic checkpoint. Users should still continue 
to follow the earlier advice of restarting the database after each upgrade.


=================(Build #1453  - Engineering Case #232226)=================

Windows CE will report IO errors to an application if it accesses files 
on a removable storage card, (eg. Compact Flash), during a short period 
after the CE device is powered on or if the card is ejected. When the 
IO error was reported to the engine/server, an assertion would fail 
and the engine would shut down with a fatal error. In other words, turning 
a CE device off and back on could cause an assertion failure in the 
engine if the database was located on a storage card.

Now when certain IO errors related to removable media occur on CE, 
a message is displayed in the server  window and the operation is retried 
for some duration and, if not successful, a message box is displayed 
to  prompt the user to re-insert the storage card and to ensure that 
storage space is available. The user has the option to retry the operation 
or cancel. Canceling the operation will likely cause an assertion failure 
in the engine.

The engine will now also display a message in the server window whenever 
a PC card has been inserted or removed. Note that these messages will 
also appear when the power is turned off and back on since CE effectively 
ejects and reinserts each PC card when the device is turned on.

This change also corrects a problem where a number of writes could 
be performed by other threads after an assertion failure or a fatal 
error. In rare circumstances, writes occurring after an assertion failure 
or fatal error could conceivably cause database corruption.


=================(Build #1453  - Engineering Case #232317)=================

If a 'create view' statement contained a call to a user-defined function, 
where the owner of the user-defined function was specified, then the 
error '<user-defined function> is not a java class, and cannot be referenced 
from SQL.'  In other words, the user-defined function syntax would be 
misinterpreted as a java method call.


=================(Build #1456  - Engineering Case #233905)=================

When starting the network server from the command line or windows without 
specifiying any additional parameters a GUI would pop up where the parameters 
could be specified.  When a server name was specified the resulting 
command line that was generated would be incorrect.  The server name 
would also be used as the database name.  As a result, a user would 
be unable to connect to their database unless they specifed the new 
database name explicitly.  This would only occur when the server name 
specified is different from the default server name.


=================(Build #1456  - Engineering Case #234085)=================

Returning a string from an external function that used the original, 
(old), external function interface could have crashed the server.  This 
has been fixed, but a work around would be to rewrite the external function 
in the dll to take an out parameter and write a stored procedure/function 
wrapper with the the original interface that calls the new external 
function.



=================(Build #1457  - Engineering Case #234275)=================

If the RAND() function was called with a seed argument, subsequent calls 
with no argument should have returned a consistent sequence of values. 
The sequence returned was not consistent if the server was run on a 
multi-processor machine with multiple OS threads, or if other connections 
were making simultaneous calls to RAND().


=================(Build #1458  - Engineering Case #231290)=================

Transact-SQL allows temporary tables to be created using "SELECT ... 
INTO #temptab ..." or "CREATE TABLE #temptab(...)".

Three problems were fixed:
1 - Temporary tables, created as above, inside an IF statement would 
be dropped at the end of the IF.
2 - Temporary tables, created as above, inside an EXECUTE IMMEDIATE 
would NOT be dropped when the compound statement or procedure containing 
the EXECUTE IMMEDIATE completed, leaving the table accessible.
3 - No error was given for attempts to create duplicate temporary tables 
at the same scope level. If a table was declared twice, the second definition 
would hide the first until the second table was dropped; at that point 
the first table would again be visible.


=================(Build #1458  - Engineering Case #234413)=================

If the server was started with -m switch, (truncate on CHECKPOINT), 
the statement
	BACKUP DATABASE DIRECTORY 'xxx'
would report an error when trying to back up the transaction log. This 
has now been fixed.


=================(Build #1458  - Engineering Case #234443)=================

In the following SQL statement, the "g:" causes a problem. The error 
is "Mismatch between external function platform specifier and current 
operating system [-618] ['WW004']".

	CREATE function gpfTest()
	returns char(50)
	external name 'GPFTEST@g:\src\db\test\tst\extproc.dll'

The work around is to write the external name string as:

	external name 'WindowsNT:GPFTEST@g:\src\db\test\tst\extproc.dll;Windows95:GPFTEST@g:\src\db\test\tst\extproc.dll'

Note that both platforms must be specified if you want it to work under 
both WindowsNT/2K and Windows95/98/ME.



=================(Build #1459  - Engineering Case #213595)=================

Calling the java method System.currentTimeMillis() when running on a 
NetWare server would always return 0 for fractions of a second. On NetWare 
4.x and 5.x, it now returns the correct number of milliseconds. NetWare 
3.x does not support this capability.


=================(Build #1459  - Engineering Case #228111)=================

Attempting to insert a datetime value into a column of a remote table 
that had less than 3 digits in the millisecond field, (eg '2000-11-21 
01:23:23.050'),  would result in an incorrect value being inserted. 
For the example, the value that was inserted was '2000-11-21 01:23:23.500'. 
Simililarly, attempting to insert the value
'2000-11-21 01:23:23.001' , would insert the value '2000-11-21 01:23:23.100'. 
This problem has now been fixed.


=================(Build #1459  - Engineering Case #232779)=================

Queries or view definitions containing GROUP BY with an alias to a subselect 
in the GROUP BY clause could return a syntax error (SQLCODE -149).


=================(Build #1459  - Engineering Case #234102)=================

Using dbxtract -an or dbunload -ac/ar/an, on a database with a page 
size of 8k or higher, caused, the engine to go into an infinite loop 
with high cpu usage. This has now been fixed.


=================(Build #1459  - Engineering Case #234801)=================

The TDS interface for the ASA engine would return a "Syntax error near 
'(end of line)'" message to the user if it attempted to execute an empty 
string.  This behaviour was causing problems when an ASA engine was 
particpating in a Replication Server environment, since Rep Server occasionally 
sent down an empty string to be executed.  The engine now retuns immedaitely 
with a successful return code.



=================(Build #1459  - Engineering Case #234829)=================

dateadd( dw, ...) was not supported by ASA.  The workaround is to use 
dateadd( dd, ...) instead.


=================(Build #1459  - Engineering Case #234869)=================

If a permission denied error was received on VALIDATE INDEX, the message 
displayed was "you do not have permission to validate the table 'index_name'". 
It is now  "you do not have permission to validate the index 'index_name'".


=================(Build #1460  - Engineering Case #234876)=================

Using a java variable in a Remote Data Access query as in:

	select * from product where unit_price < Inv.lineItem2Cost;

where Inv is an instantiated java variable and product is a remote 
table, could have caused a server crash.


=================(Build #1462  - Engineering Case #235184)=================

Trying to start the utility_db on the network server would cause the 
server to crash. The personal server was not affected by this bug.



=================(Build #1463  - Engineering Case #235195)=================

If the total physical memory plus total available swap on a Unix system 
was greater or equal to 4 GB, then the server would fail to start - 
it would give an "Out of Memory" error.



=================(Build #1465  - Engineering Case #235450)=================

Starting and stopping the server a few times in quick succession, could 
have caused an assertion error or corruption to the database file. This 
only happened on fast SMP machines running the Linux SMP kernel with 
a version less than 2.2.17. The 2.2.17 kernel seems to have this problem 
fixed. A work around has been implemented, but will only take effect 
if the kernal is less than 2.2.17


=================(Build #1466  - Engineering Case #234477)=================

If two temporary tables were linked via a key relationship, the server 
could crash when closing a connection.  For
example:
- Create 2 temporary tables and link them with a foreign key relationship.

create global temporary table DBA.Tmp1(
    x integer not null,
    y char(2) null,
    primary key(x),
    ) on commit preserve rows ;

 create global temporary table DBA.Tmp2(
    x integer not null,
    primary key(x),
    )  on commit preserve rows;

alter table dba.Tmp1 add
    foreign key (x) references
    dba.Tmp2 (x);

- Exit from dbisqlc or shutdown to cause the the server to crash.



=================(Build #1466  - Engineering Case #235039)=================

A wide insert could have crashed the server if all of the following 
conditions were met:
1) The insert referenced a character column of width < 255.
2) A null value was initially inserted into the column.
3) A non-null value is subsequently inserted into the column, (as part 
of the same insert.)

As well, a wide insert could have crashed the server or given a spurious 
conversion error under the following conditions:
1) The table being inserted to had a primary key, and the primary key 
columns were not the first columns in the table (in order.)
2) A null value was initially inserted into any column, (other than 
primary key columns that appear at the start of the table in the proper 
order.)
3) A non-null value was subsequently inserted into that column (as 
part of the same insert.)


=================(Build #1468  - Engineering Case #236228)=================

If streams received data over TCP/IP that it didn't recognize as a valid 
packet, the packet was discarded and the connection was dropped. However, 
depending on the data received, it's possible that the engine would 
have crashed.


=================(Build #1469  - Engineering Case #236318)=================

When the server was busy with multiple concurrent client requests, it 
may have locked up and stopped responding to new or existing connections. 
The only way to unlock it,was to kill and restart the database server. 
This has been fixed


=================(Build #1469  - Engineering Case #236421)=================

When setting up an integrated login mapping using:

	GRANT INTEGRATED LOGIN TO ...

the NT userid was being treated as case sensitive. But NT userids are 
case insensitive. Hence integrated logins sometimes failed because the 
NT userid that we received from NT differed in case from the NT userid 
stored in the SYSLOGIN table. These userids are now treated as case-insensitive.


=================(Build #1470  - Engineering Case #236344)=================

The user_name() function would not always return userids with the same 
case. Even though userids are case insensitive, the function user_name() 
would not always return a userid with a consistent case. The following 
example illustrates the problem:

GRANT CONNECT TO uSErA IDENTIFIED BY password
CREATE TABLE userA.test(c1 int)
SELECT user_name(creator) FROM SYSTABLE WHERE table_name = 'test'

The result is: userA

Now shut down the engine, restart, and re-execute the SELECT query:

SELECT user_name(creator) FROM SYSTABLE WHERE table_name = 'test'

The result is: uSErA

With this fix the user_name() function will return the userid with 
the case that it was created with.


=================(Build #1470  - Engineering Case #236345)=================

For a procedure like:
    create procedure p( @p1 int ) as
    begin
        select table_id, table_name from SYSTABLE where table_id <= 
@p1
    end 
where the result set is generated by a query which refers to one of 
the procedure's parameters, the result set definition would not be stored 
correctly in the catalog. Depending on the application and API used 
to call the procedure, this problem might have gone unnoticed or might 
have generated the error "Not enough fields allocated in SQLDA" (-182). 
ODBC applications would not have noticed this problem if the data source 
was configured with "Describe Cursor Behavior" set to "If required" 
or "Always". The problem was also not apparent when calling the procedure 
from DBISQL.


=================(Build #1470  - Engineering Case #236533)=================

On CE devices, if the database character set was different from what 
the client requested, char type data would be mangled. For example, 
with a utf8 database on a Japanese CE device, the ODBC driver would 
return mangled Japanese data. A related problem was that 'select connection_property('charset')' 
would return 'cp932' instead of 'utf8' in this example. This was wrong 
because the ASA Server for CE would only send data to the ODBC driver 
for CE, in the database's character set. Please note that there is no 
way to turn on character set translation for ASA Server for CE even 
with -ct switch.


=================(Build #1472  - Engineering Case #235741)=================

"Join elimination" is a query rewrite optimization that eliminates a 
table from a query when conditions in the query's WHERE clause make 
that table unnecessary. The typical example is a FK-PK join when none 
of the primary table's columns are referenced, other than the ones in 
the join condition. For instance, the query

Select S.*
From sales_order S, employee E
Where S.sales_rep = E.emp_id

can be rewritten as

Select S.*
From sales_order S
Where S.sales_rep is not NULL

However, with any of the following conditions this optimization was 
not being performed:

- the join was specified as a KEY JOIN 
- the join condition was augmented with a user estimate
- the foreign key refered to a UNIQUE CONSTRAINT rather than a primary 
key

Similarly in 7.0, PK-PK join elimination was introduced to eliminate 
redundant instances of the same table, but the server would also fail 
to perform this optimization if the table had one or more unique constraints, 
rather than, (or in addition to), a primary key.

In addition, in some circumstances FK-PK join elimination could result 
in an improperly re-written query if the query contained a FULL outer 
join (supported in ASIQ only).


=================(Build #1472  - Engineering Case #236797)=================

When the server was running on a big endian machine, (e.g. Sparc Solaris, 
AIX, HP UX), any attempt to insert a double value into the database 
would result in a "Value <number> out of range for destination" error, 
when it should not have.


=================(Build #1473  - Engineering Case #234748)=================

If the OUTPUT was used to create a data file of the colmn values, and 
a binary column had a value of 0, the Java dbisql would put '0x20' in 
the file, whereas dbisqlc would have put '0x'. Now both dbisqlc and 
dbisql will output a 0 in a binary column as '0x00'. Also, if data from 
a binary column is empty (''), dbisqlc will now output '', (empty string), 
instead of '0x'.



=================(Build #1474  - Engineering Case #235455)=================

Calls to the SQL function RAND() in the Linux version of ASA would return 
random numbers in the range 0 to 0.000015 rather than 0 to 1.



=================(Build #1474  - Engineering Case #236657)=================

An attempt to insert a timestamp value using a SQLDATETIME host variable 
would cause assertion 103102 if the microseconds component contained 
non-zero values beyond the number of decimal places indicated by the 
Default_timestamp_increment setting. For example, if Default_timestamp_increment 
was 1000, then a microseconds component value of 123456 would cause 
the assertion, while 123000 would not. The value is now truncated before 
insertion.



=================(Build #1474  - Engineering Case #237063)=================

Assertions 101412 and 200601 could have been generated while deleting 
rows from a table.  The problem was that table pages where being deleted 
even if a temporary table existed that referenced the table page.  This 
could occur if a scroll cursor was opened on a table and then the rows 
were deleted from the cursor.  Under some circumstances the removed 
page could be requested resulting in the assertions.


=================(Build #1475  - Engineering Case #235301)=================

Serverside JDBC allows reusing JDBC statement objects to execute another 
SQL statement without explicitly closing it. The statement is implicitly 
closed on the reuse. If there was a warning associated with the statement 
object before it was reused, accessing the warning list of the reused 
statement would crash the server. This is now fixed. A work around would 
be to clear the warning or close the statement explicitly before reusing.


=================(Build #1476  - Engineering Case #237411)=================

When starting the server on Windows 95, 98 or ME, after a couple of 
minutes the warning message, "Warning: Unable to load dbmem.vxd.  Dynamic 
cache sizing is disabled." was displayed. The problem only occured when 
Adaptive Server Anywhere was installed to, (or is run from), a path 
that contained spaces --  the mechanism Win9x/WinME uses to load VxDs 
cannot handle paths containing spaces. Fixed by referencing the DOS 
compatible name.


=================(Build #1480  - Engineering Case #237415)=================

When joining a proxy table in a remote ASA database to a local ASA table, 
the result set could sometimes be incorrect. In one example, performing 
a join between two local tables was returning 4 rows, while performing 
the join with an identical remote table and a local table would only 
return 2 rows. This was due to rows containing NULL being dropped from 
the result set.


=================(Build #1481  - Engineering Case #237522)=================

If a connection was terminated due to the idle or liveness timeout expiring, 
the server could have crashed. The problem was more likely to occur 
on SMP systems and UNIX systems. It is unlikely to occur on single-CPU 
NT systems.


=================(Build #1481  - Engineering Case #237789)=================

The database could become corrupted while attempting to create a foreign 
key between temporary tables.   Assertion 101412 would be generated 
at the time of the database corruption.  When attempting to restart 
the database the error "<database name> is not a valid database file" 
would be generated.


=================(Build #1482  - Engineering Case #238177)=================

Attempting to insert a row with all default values through a cursor, 
would fail to insert a row and did not return an error.  This affected 
OLEDB's .AddNew call, as well as ESQL's PUT with all NULL data pointers, 
and ODBC's SQLBulkOperations or SQLSetPos with SQL_ADD and all SQL_IGNORE 
columns.



=================(Build #1482  - Engineering Case #238279)=================

If a checkpoint occurred while a CREATE INDEX statement was executing, 
but the server lost power or crashed some time after the CREATE INDEX 
statement completed and before the next checkpoint occurred, then recovery 
would fail. Forcing recovery by renaming the log and starting the server 
with '-f' would leave the database in a valid state, however this action 
is not recommended in a replication environment.


=================(Build #1483  - Engineering Case #238300)=================

When a new connection was made to the server, the message "TCP Listener 
shutting down (2:10038)" could have been displayed, after which no new 
TCP connections would be accepted by the server.



=================(Build #1485  - Engineering Case #237666)=================

A java RMI program would run very slowly the second time it was invoked.This 
was actually due to two problems:
1. When the  program was run the first time, it allocated Java Heap 
memory close to the maximum heap size. Running it the second time caused 
the Garbage Collector to be invoked invoked for every Java Heap allocation, 
which slowed the program substantially. The fix was to coalesce the 
heap after garbage collection near the cache page limit.
2. The Java VM was sometimes hanging on 'shut down', because the number 
of running java threads was incorrectly counted. This has now been fixed


=================(Build #1486  - Engineering Case #238290)=================

The server could have crashed when processing a number of simultaneous 
connects.  This problem was most likely to appear at startup on an SMP 
machine, with more than one database being specified on the command 
line, when processing the transaction log in recovery mode


=================(Build #1486  - Engineering Case #238291)=================

The first time a table was used by an SQL request after startup, the 
engine would sample every index on that table to determine the physical 
characteristics of each index (depth, number of leaf pages, etc) and 
utilize those statistics during index selection. With this change, the 
optimizer now only samples relevant indexes for this query, where relevance 
is determined by the sargable predicates in the query's WHERE clause.	


=================(Build #1486  - Engineering Case #238879)=================

DBTran and DBRemote could have hung when scanning a log file containing 
a delete of a row containing java objects.  If the row belonged to a 
table that had articles or had no primary key, then the log scan will 
fail if the deleted row contained a java object that referenced a java 
class that had not been referenced since the previous checkpoint.


=================(Build #1487  - Engineering Case #239015)=================

The time required to start a database containing many table/view definitions 
was longer for databases created with older versions of the server than 
for an equivalent database created with current software. An upgraded 
database would also exhibit this problem. The time difference depended 
on the number of tables and views defined in the database, the number 
of foreign keys, and the speed of the server machine. The server now 
avoids two cases which caused sequential scans of system tables to be 
done.


=================(Build #1487  - Engineering Case #239192)=================

Attempting to evaluate an IF EXISTS query on a proxy table joined with 
a local table, may have resulted in a correlation name not found error.



=================(Build #1488  - Engineering Case #220999)=================

The error:
	Definition for alias 'alias-name' must appear before its first reference
could be reported erroneously when an alias was referenced in the ON 
condition of a join. For example:
	select pk1 + 10 as c1
	from t1 join t2 on (c1=c2)

It could also be reported incorrectly when a column defined with a 
CHECK condition was referenced in a view, and a query on the view further 
aliased the column. For example:
	create table t(pk int primary key, c1 int check(c1 < 5));
	insert into t values(1,1);
	commit;
	create view v as select * from t;
	select pk,c1+10 as c1 from v;

If there was a circular alias reference, the server could hang, or 
crash. From build 1459, until this fix, referencing alias' only after 
they were defined was strictly enforced, even in cases when an alias 
loop would not result. This would have lead to queries which had previously 
worked, receiving this error. The mechanism used to detect this error 
has now been changed to allow forward alias references which do not 
result in loops.



=================(Build #1488  - Engineering Case #238632)=================

When the OS ran out of memory, and a new client application attempted 
to connect, the application may have hung during the connect.

Generally, allowing the OS to run out of memory may cause many different 
effects ranging from a hung server or client, to many different server 
assertions (including an "out of memory" error).  Most of the server 
allocations come out of the configured memory cache, (designated by 
the -c switch), but some of the server's allocations, including the 
communication buffers, use system memory.  Normally this is not an issue 
since the server will do an initial set of allocations during the first 
few client sessions, and then reuse this memory.

Periodically under heavier loads, the server will need to do more allocations 
from system memory, and thus hit a newer "high water" mark.  If at the 
time that the server needed to increase this "high water" mark, due 
to an increase load, the server cannot allocate any more memory, (or 
extend its size of shared communication memory), due to a system wide 
out-of-memory, (or out-of-address space), problem, the server would 
assert or possibly hang.

This fix now causes the server to issue a fatal assertion error when 
such an internal fatal error occurs (#105600), or runs out of memory 
(# 105601).


=================(Build #1489  - Engineering Case #235876)=================

Inserting an invalid date value with year greater then 9999 through 
TDS would succeed.  For example, a Java application using jConnect could 
insert the value '20000, 01,01'.  Fetching the value from the database 
without going through TDS would display the date as '****-01-01'.  

The error message 'The specified date is invalid' is now returned when 
attempting to insert the invalid date.


=================(Build #1490  - Engineering Case #240161)=================

If a table contained a column defined with a CHECK constraint, a DELETE 
of the form:
	DELETE FROM table FROM table WHERE ...
would cause the error:
	Correlation name 'table' not found
when no error should have been given. The CHECK constraint must have 
been defined with a placeholder variable (e.g. @col). This would typically 
happen if the constraint was inherited from a user-defined type used 
to define the column. The DELETE must reference the table twice in two 
FROM clauses to cause the error to be given.


=================(Build #1490  - Engineering Case #240252)=================

If a proxy table was created which referred to a table with a column 
whose default value was greater than 16 characters in length, (eg, a 
timestamp column with default value of current timestamp), then the 
name of that column would be corrupted in the proxy table and any queries 
that make use of that column would have failed.


=================(Build #1490  - Engineering Case #240259)=================

An ALTER PROCEDURE statement could cause a server crash if all of the 
following conditions were present:
- the definition was mostly in Watcom-SQL dialect
- no RESULT clause was present
- at least one Transact-SQL statement was present in the procedure 
(e.g. SELECT @var=value)
- some part of the procedure definition contained a keyword allowed 
only in Watcom-SQL (e.g. ATOMIC, EXCEPTION)



=================(Build #1492  - Engineering Case #240440)=================

If a trigger was executing, dropping or altering a view may have caused 
the engine to crash. The same problem could occur when doing a grant 
or revoke, if the server was running in C2 mode (i.e. with the -sc switch).




                      ******************************                      
                      ** Adaptive Server Anywhere **                      
                      **                          ** 
                      **  Sybase Central Plug-in  **                      
                      ******************************                      

=================(Build #1450  - Engineering Case #227120)=================

When creating a publication with a large number of tables using "Add 
Publication", a syntax error could result when the Finish button was 
clicked. This problem appeared under Windows 95/98 only.


=================(Build #1450  - Engineering Case #230641)=================

Unload and Extract menu items were missing from the database's context 
menu items.  This occurred
on Unix machines only and not on Win32 machines. It has now been fixed.


=================(Build #1451  - Engineering Case #230945)=================

The last panel of  the "add procedure or function" wizard had a line 
with "yes", but no label in front of it. The missing label has been 
added. For functions, the label says "Returns", and the value will be 
the return type. For procedures, the label says "Result set", and the 
value is "Yes" or "No".


=================(Build #1451  - Engineering Case #231492)=================

Creating a column with a user domain name with a space in it would cause 
a syntax error. Now, quotes are put around user domain names, so the 
syntax error will no longer occur.


=================(Build #1451  - Engineering Case #231494)=================

Domain properties always reported the domain as using the database default 
for allowing nulls.  This has been fixed.


=================(Build #1451  - Engineering Case #232739)=================

After using "View Data" in Sybase Central to view a table's data, then 
viewing a second table's data, Sybase Central would not  terminate, 
even after all windows were closed. The process scjview.exe would still 
be in the task list, and the JAR files would be locked. This has been 
corrected.


=================(Build #1472  - Engineering Case #236249)=================

Table creation could have failed as the SQL generated was incorrect.  
This has now been fixed.



=================(Build #1474  - Engineering Case #235964)=================

If a procedure had a comment with a question mark in it, attempting 
to save it would cause  an error.  This only occurred if the connection 
to the database was made throught jConnect.  This has been fixed.



=================(Build #1481  - Engineering Case #229422)=================

When using the Create Domain wizard, size and scale would not be accepted 
if they contained leading or trailing spaces. The  value is now trimmed. 
Also, the partition field is only enabled for the "global autoincrement" 
default value.


=================(Build #1481  - Engineering Case #236209)=================

When running the Unload Database or Unload Database wizards, and the 
'Keep the connection in Sybase Central' option was left unchecked, the 
operation would fail with 'SQL error: Specified database not found'. 
This has been fixed.


=================(Build #1481  - Engineering Case #236451)=================

The Translate Log wizard now allows you to choose a log file with any 
extension. Previously, attempting to translate a logfile in with an 
extension other than .LOG, caused an error. 



=================(Build #1482  - Engineering Case #216682)=================

When using the Extract Database wizard and specifying a connection to 
a database with no replication info, (ie an empty database), two error 
dialogs were displayed, the wizard would then advanced to the second 
panel even though there was a problem.  Now the wizard stays on the 
connection page. 



=================(Build #1482  - Engineering Case #232417)=================

If two or more table editors were opened on the same table and each 
was closed, the primary key would be deleted.  This has been fixed.


=================(Build #1482  - Engineering Case #233585)=================

The '@' character in table and column check constraints is no longer 
converted into the ':' characters.


=================(Build #1488  - Engineering Case #237653)=================

The upgrade context menu item did not appear if the database was only 
a point release older than the current version, (ie 7.0.2 vs 7.0.1).  
The upgrade option now appears for all databases.


=================(Build #1488  - Engineering Case #238456)=================

Connecting to a database created with the EUC_TAIWAN collation would 
fail with the error "JZ015: An unrecognized CHARSET property was specified: 
eucb5."  This has been fixed.




                      ******************************                      
                      ** Adaptive Server Anywhere **                      
                      **                          ** 
                      **        Utilities         **                      
                      ******************************                      

=================(Build #1448  - Engineering Case #229433)=================

The START DATABASE ... ON statement would inadvertently disconnect the 
current database connection if it was unnamed and you were running in 
batch (console) mode


=================(Build #1448  - Engineering Case #229546)=================

The STOP ENGINE statement would fail with a syntax error, in Java ISQL, 
if an engine name was specified.


=================(Build #1449  - Engineering Case #229621)=================

The following two OUTPUT problems have been fixed:
1. When exporting character data to an Excel spreadsheet, if the string 
lengths were longer than 255 characters, the time to export them was 
unacceptably slow. Note that the Excel file format limits string values 
to 255 characters.
2.  When exporting in HTML format, data which contained ampersands, 
less-than signs, greater-than signs, or quotation marks were not encoded 
correctly in the output file. This could prevent some HTML renderers 
from correctly displaying the data.





=================(Build #1449  - Engineering Case #229750)=================

If an error is encountered in an included (READ) .SQL file, and the 
user elects to stop processing, then ALL script processing is halted, 
including the calling file (and so on). Previously, execution of the 
outer file continued, which was incorrect. Also, the included file was 
being left open by ISQL so that it could not be deleted / renamed by 
other applications, even after script processing was aborted.



=================(Build #1450  - Engineering Case #230496)=================

Some DBISQL options were added, but they are not created in new databases 
by default, which could cause problems when rebuilding a database.  
They are now created with default settings when a new database is initialized.
These options are:
ISQL_field_separator
ISQL_escape_character
ISQL_quote
ISQL_plan
ISQL_command_timing


=================(Build #1451  - Engineering Case #231841)=================

DBISQLC could crash when trying to execute a statement after a failed 
DDL statement such as ALTER TABLE. This has been fixed. In the cases 
affected, the cursor being used to display the previous result set is 
now closed and the data window cleared.



=================(Build #1457  - Engineering Case #226136)=================

On Windows-based systems, the DBCOLLAT tool would write collation, map 
and insert files using only linefeed (0x0A) characters at the end of 
each line, instead of carriage return and linefeed (0x0D and 0x0A) as 
required. The resulting files were difficult to edit, since the line 
breaks would usually not occur properly in most text editors. On Windows-based 
systems, DBCOLLAT will now write carriage return and linefeed at the 
end of each line.


=================(Build #1458  - Engineering Case #234637)=================

If no existing .odbc.ini file is found, the UNIX dbdsn -w would attempt 
to create /.odbc.ini (i.e. in the root directory of the machine).  It 
is now created in ~/.odbc.ini if no existing .odbc.ini is found.


=================(Build #1461  - Engineering Case #234343)=================

Setting ISQL database options with boolean values, using DBISQL, would 
cause errors about the values "true" (or "false") being invalid, if 
the database was latter started with DBISQLC . This has been fixed.


=================(Build #1464  - Engineering Case #235282)=================

The procedure debugger would throw a "ClassCastException", which would 
mention the "DebugPosition" class, if two breakpoints were set on the 
same line -- one for a connected user, and the other for the synthetic 
"All Connections" user. This has now been fixed.



=================(Build #1466  - Engineering Case #235013)=================

On Linux only, after typing the user id and password information into 
the logon dialog and hitting OK, dbisql may have core dumped with a 
message similar to :

SIGSEGV 11* segmentation violation stackpointer=<memory_address>

The Procedure Debugger may also have inhibited the same behaviour.


=================(Build #1470  - Engineering Case #234723)=================

Running dblog on a database file larger than 2GB would give the following 
error message:
Unable to open database file "{database name}"-- Unknown error (79). 
This has now been fixed.



=================(Build #1474  - Engineering Case #225626)=================

Running the Backup Utility, dbbackup.exe, with the -x switch, (delete 
and restart transaction log), caused the transaction log to be copied 
to the backup directory, deleted and a new transaction log started with 
the same name. This is the correct behaviour, but if the database was 
running with a mirror log as well, the mirror log was renamed using 
the name yymmddxx.mlg and a new mirror log was created. The yymmddxx.mlg 
was not deleted after that even though the -x switch means deletion 
should be done. Now the mirror log with the name yymmddxx.mlg is deleted.


=================(Build #1474  - Engineering Case #232425)=================

DBISQL would report an internal error executing the INPUT INTO command, 
if there were more columns in the file than the table and the file was 
in LOTUS format. The error message contained  "ArrayIndexOutOfBoundsException" 
in the "Stack Trace" portion of the message.

Now, attempts to import a file with more columns than the table, will 
only inport the first "n" columns, where "n" is the number of columns 
in the recipient table. Data for the other columns in the file are ignored.


=================(Build #1474  - Engineering Case #236519)=================

In files which were processed via a READ statement, parameters that 
were in double quoted strings were not being expanded. With this fix, 
now they are. For example, if you have two files called F1.SQL and F2.SQL 
whose contents are as follows:
	
	F1.SQL:    READ F1.SQL occam
	
	F2.SQL     PARAMETERS tableName;
	                   SELECT * FROM "{tableName}";
		   
DBISQL will now interpret the SELECT statement in F2.SQL as 

        SELECT * FROM "occam"
	    
This is the same behavior as DBISQLC.



=================(Build #1474  - Engineering Case #236894)=================

Result sets from stored procedure calls would not be displayed if DBISQL 
was run in batch mode. For example, a procedure called "my_procedure", 
which returned a result set, run with the following command:
	
	DBISQL -c uid=dba;pwd=sql "my_procedure"
	    
now prints the result set. Previously it did not.


=================(Build #1476  - Engineering Case #237502)=================

The NetWare version of dbisql would crash if trying to connect to a 
server running on the same machine, but specifying the wrong server 
name. Note that specifying no server name at all would work. Only versions 
of isql.nlm as of build 1450 had this behaviour.



=================(Build #1482  - Engineering Case #237755)=================

Attempting to upgrade a database, that had already been upgraded to 
version 7.0.0 or 7.0.1, to 7.0.2, would fail with a primary key violation 
on table 'ml_script_version'. The database was then unusable afterwards. 
This has now been fixed


=================(Build #1487  - Engineering Case #236717)=================

Clicking the "Show Columns" button on the Table Browser dialog, by a 
users that was not the owner of the selected table, would give an empty 
list of columns. Now the list shows the column names regardless of ownership.



=================(Build #1490  - Engineering Case #238287)=================

New connections were prevented from opening if the following were done 
in dbisql:

            1. Connect to a database
            2. Disconnect from the database
            3. Connect by clicking the "SQL/Connect" menu item

At this point, regardless of the userid and password, the following 
message was displayed,  "Connection failed. Parse error: Invalid or 
missing keyword near 'CHARSET_CONVERTER_CLASS'" The following line was 
also displayed on the "Advanced" page of the "Connect" dialog:

            CHARSET_CONVERTER_CLASS=com.sybase.asa.logon.SunIoConverter

A workaround is to manually remove this line from the text field on 
the "Advanced" page before clicking the "OK" button in the "Connect" 
dialog.


=================(Build #1493  - Engineering Case #240260)=================

When using dbping to connect to a blank-padded database, the string 
displayed was also blank-padded. For example:
Adaptive Server Anywhere Server Ping Utility Version 7.0.2.1466
Connected to ASA 7.0.2 (1466)                            server "blank                                    
" and database "blank                                    ".
Ping database successful.



=================(Build #1493  - Engineering Case #240262)=================

The dbsvc utility would get an  "Access is denied" message when attempting 
to delete a service.




                             ****************                             
                             **  Mobilink  **                             
                             **            ** 
                             ** ASA Client **                             
                             ****************                             

=================(Build #1449  - Engineering Case #230023)=================

There are two possible ways for users to alter tables when the tables 
are involved in a synchronization definition.
1 - Within synchronization: Using hooks or the -i site script
2 - Outside synchronization: Remove the table from the synchronization 
definition. Alter the table and then add it back.

Both Case 1, with extended option LockTables=OFF, or Case 2 may have 
composed invalid upload streams, if some unsent row operations appeared 
before the alter table operation. An invalid upload may have crashed 
the  MobiLink server or corrupted data in the consolidated database.

Now, dbmlsync will detect and abort, if this kind of violation occurs 
while composing the upload stream.

A good practice to avoid the problem is to alter the table within the 
synchronization and keep the default ON value of the LockTables option 
for the sync that alter table is going to occur.



=================(Build #1456  - Engineering Case #234145)=================

DBMLSync accepted only Yes/On for the extended option lt, (locking tables).  
Any other settings, such as, On/Off, True/False, 1/0 were ignored and 
it used the default value of On and locked all tables.Now the behaviour 
is  to accept the settings,  On/Off, True/False and 1/0.



=================(Build #1459  - Engineering Case #234048)=================

When the ml_add_user stored procedure, (in syncasa.sql), was invoked 
in an ASA consolidated database, and the ml_user table was empty, there 
would be no user added. This has now been fixed.



=================(Build #1460  - Engineering Case #235041)=================

MLXtract could crash on UNIX platforms.


=================(Build #1462  - Engineering Case #232494)=================

When an ASA client had a syncronization definition that contained a 
table that had a foreign key relationship to a table outside of the 
syncronization definition, then dbmlsync might not order the tables 
correctly in the upload stream to prevent
RI problems on the consolidated database.


=================(Build #1463  - Engineering Case #235142)=================

When a read error occurred while MobiLink was waiting for the download 
acknowledgement, (or any read that wasn't a read of the upload), the 
misleading error text was as follows:
	"Incomplete upload stream."
This has been replaced with:
	"A network read failed. Unable to read data from the remote client."
Similarly, when a write error occurred while MobiLink wasn't writing 
the download, (as can happen when an UltraLite remote doesn't know what 
happened to the upload in the previous synchronization, for example), 
the misleading error text was as follows:
	"Failure writing download stream."
This has been replaced with:
	"A network write failed. Unable to write data to the remote client."



=================(Build #1490  - Engineering Case #240253)=================

DBMLSync on UNIX platforms, may have had permission problems when it 
was trying to access some system tables, such as sysuserperm, if the 
user ID DBMLSync used to connect the engine, had only remote DBA authority.




                            *******************                            
                            **   Mobilink    **                            
                            **               ** 
                            ** Scout Conduit **                            
                            *******************                            

=================(Build #1451  - Engineering Case #230640)=================

The product version field of the version resources for the Scout conduit 
always claimed to be "DEBUG" even when it was a production version.





                              **************                              
                              ** Mobilink **                              
                              **          ** 
                              ** Streams  **                              
                              **************                              

=================(Build #1451  - Engineering Case #229622)=================

The headers of the HTTP requests used during HTTP 1.1 synchronization 
were missing the "Host:" entry. This was OK for direct communications 
between a MobiLink client and server. However, when there was an HTTP 
proxy, firewall, or a Web server in between the MobiLink client and 
server, the request could have been rejected if the HTTP was being validated 
along the way. The "Host:" entry is now included in all HTTP 1.1 request 
headers.




                       ****************************                       
                       **        Mobilink        **                       
                       **                        ** 
                       ** Synchronization Server **                       
                       ****************************                       

=================(Build #1451  - Engineering Case #231572)=================

When the MobiLink Synchronization Server adds a connection to the connection 
pool, if the connection pool is full, memory usage could increase. Connections 
are added the connection pool under any one of the following circumstances:
1 - A worker thread needs a connection, and there is no free connection 
in the pool, and there is room in the pool for one more connection.
2 - A worker thread needs a connection, and the free connections in 
the pool are not associated with the script version being synchronized 
on that thread. A pooled connection with a different script version 
will be disconnected and removed from the pool, and a new connection, 
associated with the current script version, will be created.

It was the second case that may have leaked memory.


=================(Build #1467  - Engineering Case #235433)=================

When reading the upload stream, MobiLink wouldn't properly clean up 
the network layer when finished reading the upload. This could potentially 
lead to communication errors. There were not any known failures related 
to this change, but the potential for future problems, in either new 
or altered network protocols, (eg. TCP/IP, HTTP, etc), is now greatly 
diminished.


=================(Build #1472  - Engineering Case #236963)=================

When an updated row was uploaded, but the consolidated row with the 
same primary key no longer existed, MobiLink would have attempted the 
update anyway, resulting in either no warning (should have been an unable 
to update the row message), or an ODBC error. The actual behaviour depended 
on the ODBC driver used.


=================(Build #1474  - Engineering Case #235458)=================

When invoking the error handler after a database connection has gone 
bad, the MobiLink server could crash or consume 100% CPU.




               *********************************************               
               ** SQL Remote for Adaptive Server Anywhere **               
               **                                         ** 
               **        Database Tools Interface         **               
               *********************************************               

=================(Build #1448  - Engineering Case #230858)=================

The receiving phase of SQL Remote may have become much slower, if errors 
occured when it was applying
incoming messages.


=================(Build #1448  - Engineering Case #231199)=================

In verbose mode, SQL Remote may have shown the same transaction multiple 
times.  For instance, if an error (deadlock, conflict update, etc.) 
occurred when it was applying the third transaction from an incoming 
message, it would re-apply transaction without grouping or buffering 
the operations.  But in the retry, SQL Remote would show the first and 
second transactions as well.  This behaviour was confusing, so now SQL 
Remote will only display the SQL statements that will be applied or 
re-applied.



=================(Build #1451  - Engineering Case #232829)=================

Upgrading the Adapive Server Anywhere software from version 5.5.0x to 
6.0.x or 7.0.x, or from version 6.0.x to 7.0.x, without upgrading the 
databases as well, would cause SQL Remote to complain and refuse to 
run. This restriction has been removed.



=================(Build #1459  - Engineering Case #234982)=================

SQL Remote may have crashed if the command line switch -l was used to 
set the maximum message size to a few megabytes and the compression 
level was set to 0.


=================(Build #1461  - Engineering Case #233841)=================

If a 5.5.x remote database was upgraded to 6.0.x or 7.0.x, but SQL Remote 
had never been run against it before being upgraded, 6.0.x or 7.0.x 
SQL Remote may not have sent out any messages to the consolidated database. 
This is now fixed.


=================(Build #1474  - Engineering Case #236966)=================

With the option delete_old_logs = on, SQL Remote deletes all the old 
transaction logs that are not needed for replication. It should also 
have deleted all the corresponding mirror logs.  However, SQL Remote 
on UNIX platforms may complain "Could not delete ..." and the mirror 
logs may not be deleted. This has now been fixed.



=================(Build #1481  - Engineering Case #237220)=================

DBRemote could delete old logs on databases initialized with -b, (blank 
padding of strings for comparisons), even if the value set for delete_old_logs 
was 'Off'.


=================(Build #1482  - Engineering Case #238106)=================

If an Adaptive Server Anywhere database contained any synchronization 
definitions, templates, or sites and the option column for any of them 
was NULL, dbunload would complain "***** SQL Error: Cursor already open" 
and then exit.




             *************************************************             
             **   SQL Remote for Adaptive Server Anywhere   **             
             **                                             ** 
             ** File Messaging for Adaptive Server Anywhere **             
             *************************************************             

=================(Build #1452  - Engineering Case #233245)=================

With the 'FILE" message type, if the last character in the publisher 
/ remote user addresses was a Japanese character which had \5c, (backslash 
or YEN sign), as the follow byte, SQL Remote would not replicate.




                        ***************************                        
                        **       UltraLite       **                        
                        **                       ** 
                        ** Analyser Java Classes **                        
                        ***************************                        

=================(Build #1448  - Engineering Case #229429)=================

The UltraLite analyzer did not output binary constants correctly, such 
that the first byte contained the length of the binary constant.




                           ********************                           
                           **   UltraLite    **                           
                           **                ** 
                           ** Code Generator **                           
                           ********************                           

=================(Build #1451  - Engineering Case #231010)=================

Wheb defining tables to be used in MobileBuilder using a publication, 
if any of the tables in the publication began with a lower case character, 
compile errors complaining about missing class definitions would occur.


=================(Build #1462  - Engineering Case #235898)=================

When running UltraLite against a database older than 7.0.2, the following 
error: Procedure 'genTableDataCode' terminated with unhandled exception 
'com.sybase.asa.ultralite.ULFatalError' FATAL: Internal error running 
the analyzer, 14, could occur. This has been fixed.




                          ***********************                          
                          **     UltraLite     **                          
                          **                   ** 
                          ** Runtime Libraries **                          
                          ***********************                          

=================(Build #1448  - Engineering Case #229430)=================

When creating a service using dbsvc, using the -a switch to specify 
a network account, (i.e. domain\account),was not working . We were always 
prepending ".\" to the account name. Now we only do if there isn't a 
"\" in the account name already.


=================(Build #1451  - Engineering Case #230509)=================

In Java UltraLite, a Blob or Clob could be stored incorrectly in the 
persistent store, when the database is shut down.



=================(Build #1451  - Engineering Case #232984)=================

A row in a table with a multi-column foreign key that allows null values 
would be rejected if any of the columns in the foreign key was null. 
This row will now be accepted.


=================(Build #1455  - Engineering Case #233901)=================

In certain types of queries that required a temp table in the execution 
plan, the UltraLite runtime could have returned incorrect values to 
the user.  This would have manifested itself as returning negative numbers 
as positive numbers, random bit values or generally unexpected values.



=================(Build #1471  - Engineering Case #236652)=================

The HEXTOINT() function could return incorrect value on Unicode Platorms. 
This has now been fixed



=================(Build #1472  - Engineering Case #236880)=================

When reading from the network, UltraLite wouldn't properly clean up 
the network layer when finished reading. This would potentially lead 
to communication errors.

There aren't any known failures related to this change, but the potential 
for future problems, in either new or altered network protocols (eg. 
TCP/IP, HTTP, etc), is now greatly diminished.


=================(Build #1490  - Engineering Case #238284)=================

Assignment of a blob which was nullable would generate a compilation 
error.

