Sybase数据库技术，数据库恢复专家

sp_passthru in a multi-session environment in which some of the sessions may hang

in 'remote i/o'.

If users do not upgrade their ASEPlugin, the latest JS Template instance documents should not be upgraded.

The changes were required to support localization for the Job Scheduler interface within the ASEPlugin.

"Domain error during implicit conversion of BIT value '0' to a BIT field."

ASE 12.5.1 running on win2000 adv server

1965, ... exceeds maximum limit of 1964. Run update statistics on this table after upgrade." may be

reported in rare circumstances during an upgrade from a pre 12.5 server.

raised during ALTER TABLE add <column> command on a table which has a column using a constraint name

of 30 characters.

entry <value> ..." or "WARNING: Pss found with open sdes ... system table entry

<value> ..." when a session hits a non-fatal internal error and a system or user table

has been left open mistakenly.

BMC Bug #388263

or a SELECT ... FOR BROWSE on a view, the base table column

names are returned instead of the view column names.

install auditing capabilities. Terminating configuration." on AIX platform.

statements and the datatype of the parameter is specified

as UNICHAR, conversion from UNICHAR to CHAR may not be correct

if the server's default character set is configured to something

other than UTF-8.

contains a select into statement using a view created on table that

is recreated.

of system type names.

subtraction).

2> when the HAase Agent online scripts start ASE, it starts as user "root" and not as user "sybase" and when Agent offline script trys to stop ASE , it is not stopped .

Platform: Linux

core dumps, or causes ASE to hang after running sp_showplan.

value defined on a VARCHAR column and the file to load does

not specify value for that column, the DEFAULT value inserted

into the table is silently truncated to 30 chars.

when there are more than 5 engines bound to a certain engine-group.

Before this fix the XML builtin xmlparse retuned the result xml fragment as image datatype. The fix adds the ability to return the result xml fragment also as varbinary/binary datatype using additional 'returns' clause syntax.

Before this fix the XML builtin xmlextract returned the result xml data as either text or varchar. With the fix the user will be able to return the result xml data as java.lang.String also.

Before this fix the XML builtin xmlrepresentation accepted xml documents only as image datatype as one of its operands. With this fix user can now pass xml documents of varbinary/binary datatype also to this builtin.

having ANSI-style comments at the end of the command string sent from the client.

work ready for that engine.

ions' available", followed by the connectivity errors

11216 and 11206 will be raised, if trying to access an

MDA table from within a user's login script.

This means that repeating jobs only run from starttime to midnight.

or univarchar, sp_sproc_columns doesn't return any data type information, NULL is returned.

SQL queries with a huge number of IN() elements.

access the commit service to determine the status of a two-phase

transaction involving a replicated table.

space for object '<obj>' in database '<db>' because '<seg>' segment is full/has

no free extents. If you ran out of space in syslogs, dump the transaction log.

Otherwise, use ALTER DATABASE to increase the size of the segment."

LOAD TRAN command and during the redo phase of an index creation

for a Data Only Locked table.

stat_check_sampling_perct may be reported in the errorlog for queries requiring sort

executed in parallel.

the last chance threshold and got aborted due to a log full condition,

it may fail to rollback successfully raising a fatal 3475 error.

command may not work within a procedure or batch.

With this fix, CREATE EXISTING TABLE ... EXTERNAL TABLE/PROCEDURE/FILE and CREATE TABLE ...

EXTERNAL TABLE/FILE commands are now able to work in a procedure and batch.

listing all the files that will be installed by this EBF.

User information is being written to the DOS Console including ID and PSWD, when doing Monitor reporting.

This needs to be changed

sysattributes table related to a task level specification" will appear in the ASE error log upon

the termination of a process whose priority has been set via the sp_setpsexe stored procedure.

This message may be safely ignored.

'omni_code_gen_subq' and 'omni_opt_select' is sometimes printed in the errorlog when executing a

complex SQL query containing a SELECT using views. The stacktrace is printed after raising message

404, 'Too many ANDs or ORs in expression (limit 1024 per expression level). Try splitting query or

limiting ANDs and ORs.'

query exceeds the internal fixed limit of 14 worktables.

Customer would like this limit extended.

The Hardcoded limit is located:

/view/ase1251_cscope_vu/calm/svr/sql/generic/include/server.h

# define MAXWORK 14 /* max active SDESes for non-RI work tables */

dependent on how much in excess of the limit the query is. User

would like the 7789 error to include the exact number of worktables

needed to process the query in question.

many worktables a given query would require would be very desirable.

This is not possible, at present.

So that customer can predetermine how many Work Tables a Query would require.

sort orders to be selected.

giving spurious 605 errors in the temporary database or may fail with stack traces.

Stack trace will reference modules like exec_eop,run.

which includes the modules 'sendtext' and 'run' if a SELECT using ISOLATION LEVEL 0

selects from text/image columns and other users are updating these text/image columns at the

same time.

available to start a new process. Retry when there are fewer active users, or

ask your System Administrator to reconfigure SQL Server with more user

connections." may be reported in the errorlog due to upspawn failing with

the following message "upspawn: no more process slots available".

With 325777 changes, xp_cmdshell commands executing a shell command which tries to redirect output to a file using the '>' file I/O redirection does not work as expected. Instead the output is returned to the ASE client. Fix for 325777 will be reworked under 348878.

may be hit during the execution of SP_RENAME when the given object

is not yet cached. The stack trace will also report modules 'exec_dbcc'

and 'd_refresh_ides' in the error log.

systemtable SYSDATABASES when using bracketed indentifiers in the CREATE DATABASE statement like

CREATE DATABASE [<dbname><space>...<space>].

an absolute path from root. It is not beginning of

a path relative to the sub-path preceding the filter.

Linux supports a variety of file systems such as EXT2, EXT3, and XFS.

The recommended file system for Adaptive Server is EXT2 or EXT3 with

journaling disabled.

which includes the modules 'logcommand' and 'grantrev' may be reported in the errorlog if a table

having an identity column is used in a stored procedure together with a GRANT or REVOKE command

for that same table and the REPLICATION FORCE_DDL option is set.

to use this character set in conjunction with eucksc, the sp_configure

parameter "size of unilib cache" should be increased to 144384 bytes

(or more) and the the sp_configure parameter "enable unicode conversions"

should be set to 1.

by a killed connection.

raised when executing the command WRITETEXT BULK <tablename>.<colname> <textpointer> and colname

is called "parameter".

(as it is by default) and character literals contain Unicode

characters requiring normalization, the parser failed to

perform the normalization. This was true for literals

promoted to the unichar data type, represented using

UTF-16, as well as char literals when the server's default

character set is configured as UTF-8.

CO55 The specified Compiled Object Item not found in the Server: <sproc_name>.

a stack trace which includes the modules 'omni_get_ojrelops' and 'omni_findjoins' may be reported

in the error log if a query with UNION and ANSI join involves proxy tables.

contain multiple main steps, such as UNION statements.

object '8' in database '<dbid>'" may be reported during boottime or LOAD DATABASE recovery.

may be printed in the errorlog, if a non-sa user tries to execute a database level dbcc command in a database that is not usable.

checkstorage will be raised when running DBCC CHECKSTORAGE on a database that

is marked suspect. Similar errors and stack traces are seen if the database is unavaliable

for some other reasons.

a character set or sort order definition if such a definition

is already installed in the server. The program loops until

a valid yes/no response is received from its standard input.

EOF is not considered a valid response, and will trigger

an indefinite loop.

improved. Further performance improvements may be achieved with the 702 boot time

trace flag though this comes with an additional 2% per-batch memory cost.

name of the object and some detailed object info is greather than 80 characters.

"cache_partition=<number>"' processed a change that went into effect when the

server was rebooted.

In ASE 12.5.1, running the same command returns messages (Msg 10850 and 10875, both

undocumented) that make it appear that it doesn't work. What is happening is a

confusion between what ASE 12.5.1 can do dynamically, and what it does do statically

DESPITE the messages. We need to make message changes.

more than ones with different case (ex: Col, coL), but the header of the result set

is the same for the columns.

trace which includes the modules 'q_oreop_fill' and 'substitute' may be reported in the error log for

an query at isolation level 0 that includes a HAVING clause with an OR.

in the different language cannot be installed.

materialized UNION ALL containing text/image columns.

and reset monitoring counters to 0 as needed.

"The table 'table name' is an inner member of an outer-join clause. This is not

allowed if the table also participates in a regular join clause."

consisting of multiple parts will be stored in syscomments using bracketed identifiers. For example

a columnname like <colname_part1 part2> is stored like [colname_part1 part2]. The same will

happen for table-names consisting of multiple parts.

Core file indicates that thread library function pthread_sighandler_rt() encountered Signal 11

due to corrupted memory address for thread handle.

with "EC0", "EC1", "EC2", or "EC3". In these cases ASE will apply the

system defined attributes for the user defined execution class rather than

the user defined attributes.

will cause a process to terminate with a stack trace if the source data is

not aligned on a data page.

first kerberos login and recovery or other highly concurrent

activity is also running.

Invalid spinlock sanity checks are avoided with this fix.

trace which includes the modules 'q_norm_const' and 'decision' may be reported in the error log if

a SELECT with two or more aggregates has a WHERE clause on a text constant of more the 450 characters.

Unicode surrogates to GB18030.

some performance issues in a multi-session environment.

In 12.5.1, the job timeout property was labelled to be in seconds. This value was actually treated as minutes. This fix changes the labels for the timeout property to say minutes in 12.5.2.

and txt_readchar, may be raised when trying to convert a text column to any

other datatype, if the text page was created in a version prior to 12.5.

errors a stack trace may happen. This occurs only with the statement cache enabled.

subquery in the select-list which is referencing a column defined in the derived

column list of a derived table expression in the FROM clause.

Tested and reproduced with:

SQLAdvantage 12.5.0.3/EF 10752, 12.5.1 EBF 11428, 11.5.1.3

ASE 12.0.0.8, ASE 12.5.0.3, ASE 12.5.1

3021 10 2

Unable to write statistics for all indexes of object %ld in database %l

d. Please run update statistics on all indexes after loading this database or pr

ior to using the indexes in the loaded database.

appear in the errorlog if configuration parameter 'enable rep agent' is not set to 0.

sp_configure "size of shared class heap", 3072

Which will double the amount of memory available to the shared class heap, assuming there is sufficient shared memory available.

Symptoms of this problem include a message similar to the following written to the ASE errorlog:

00:00000:00018:2004/03/29 16:16:55.18 server Java VM Host: Memory allocation request failed because of insufficient memory in Jvm Shared Class.

00:00000:00015:2004/03/29 16:26:48.06 kernel timeslice -501, current process infected

expected to be 1. Instead 0 was found." may be reported in the error log

in module 'des_rekeep' when the statement cache option is active. The

module 'des_lwpdescreate' will also be reported in the stack trace.

Alternatively the error 8203 "Expected to find the descriptor for object

<value> in database <value> in DROP state." may be reported in module

'des__markdestroy'. The modules 'ssql__rmplans' and 'des_objdrop' will

also be reported in the stack trace.

ASE sometimes raise error that the number of devices should

be greater than the highest device number used in sysdevices.

while printing the stacktrace after a JVM failure.

to ASE_MSGBODY_FORMAT and ASE_MSGBODY_SCHEMA respectively.

mapped address is passed as start address], ASE may use one block of memory as

two separete memory blocks. This can result in unpredictable results and data corruption.

The problem is in this section:

4.2 Adaptive Server 12.5.1 backward compatibility.

The problem is this sentence:

You can to earlier 12.5.x versions of Adaptive Server, but to do so you must take the precautions described in the technical note before using any of the new 12.5.1 features.

The sentence should say:

You can downgrade to earlier 12.5.x versions of Adaptive Server, but to do so you must take the precautions described in the technical note before using any of the new 12.5.1 features.

The "downgrade" is missing, the,

Release Bulletin Adaptive Server Enterprise Version 12.5.1 for IPF/HP-UX is correct.

would prevent corrupt primary server's master.

update, the first excution of the query works correctly. If the user

changes the "allow updates" state, the new state is ignored when the exact same

query is subsequently executed.

The following function are part of the stacktrace:

GarbageCollect (), WalkReferences (), IsFirstFrame().

new client connections are successful, but some commands do not run properly,

possibly because the blocking process is holding a lock on tempdb..syslogs.

This usually occurs when the server runs out of memory to launch the jvm.

When it effects tables in a primary database that is marked to replicate DDL?

returns successfully if the subscription 'xxx' does not exist.

With the fix to this CR it will raise an exception.

for a password and fails to connect to the server.

This is now fixed. For various other places where DDLGen needs user to have select permission on system catalogs, the appropriate errors will still be thrown.

Upgrading, or dump and load migration is not supported from 11.0.x to 12.5.x. You would have to upgrade to at least 11.5.x first.

The newest version of ASE you can upgrade 11.0.3.x to is 12.0.

However, if you look in the 12.5.1 System Admin Guide here, it states that you can't load a pre 10.x dump, which to ct means that we support loading an 11.0.3 dump, and it will automatically be upgraded by a 12.5 server.

ASE fails to boot rather than failing over to mirror device

when started with '-r' option.

Error 1204

Severity

19

Error message text

SQL Server has run out of LOCKS. Re-run your command when there are fewer active users, or contact a user with System Administrator (SA) role to reconfigure SQL Server with more LOCKS.

Should be 17 not 19.

2727, "Cannot find index ' '" is raised.

which includes the modules 'update_statistics_col' and 'ups__minorattributes' may be reported in the

error log when an UPDATE STATISTICS or an UPDATE INDEX STATISTICS command is executed.

that is encoded in a multi-byte character set, has been BCP'd into this

table.

the statement "set char_convert on" will cause incorrect

information to be reported via the global variables

@@char_convert and @@client_csid, and it may ultimately

lead to the process being terminated with a a stack trace.

clause is not forwarded to the remote server.

in a stored procedure and would also like to be able to get a count of variable

declarations in a stored procedure.

Also, under certain conditions when sp_monitor "connection" or sp_monitor "statement" is executed

the columns SPID, LoginName, ElapsedTime, CPU_Time, PhysicalReads, LocksHeld would by empty and

only the SQL Text column would be displayed.

different size during custom installation.

new client connections are successful. Pending task list shows a CREATE TABLE statement

trying to insert a row into tempdb..sybtabstats waiting for some event to complete.

order. This may happen if the server is configured to use backward scans.

overloaded and a process goes to sleep while holding the default data

cache spinlock. The call stack of this process will show the modules

'dstartio', 'kd_p_semaphore' and 'upsleepgeneric'.

execution of a select-into command in module 's_estmt_loopend' reported

by a stack trace in the error log.

for that database. The database would be bound to a named cache of sufficient size and would not

require physical devices to back the space.

The web pages that provided an interface to the Consumer did not work properly

and are no longer available.

Problem : To make this useful to customers, we need to enhance sp_depends to also

report the column level dependencies. Currently, sp_depends supports:

sp_depends <tablename> [, <column name> ]

By default, sp_depends lists all compiled objects that depend on

table <tablename>. This shows up as the 1st chunk in the output.

Solution : sp_depends will now filters the 1st chunk of output to restrict the list of compiled objects that actually use the specified column. This information can be obtained from the column ID map.

a procedure

may result in error 691. With the fix, the CREATE TEMPORARY DATABASE command cannot be used

with the WITH DEFAULT_LOCATION and FOR PROXY_UPDATE clause. The sp_defaultloc stored procedure

will also not be able to remap objects in temporary databases.

results in the configuration option being deleted from sysconfigures.

with a signal 11 (segmentation violation) under constructs where no external o/s command did execute.

isolation level 0 query involving TEXT/IMAGE data, could hang or fail

with the message 'illegal TDS token received'.

a stacktrace containing the modules 's_normalize', 'colnames_driver' and 'prot_desc' when

a SELECT * FROM [ ] is executed whereby the tablename consist of one or multiple spaces.

with unique index 'ncsysusers2'" could be raised by a migration tool

that tried to recreate user defined roles from a server where any of

the roles DTM_TM_ROLE, HA_ROLE, MON_ROLE, MESSAGING_ROLE, JS_ADMIN_ROLE

did not exist.

ASE received Alarm Signal (Signal 14) during select system call. Internal data structures

within Linux signal handling code are corrupted resulting in failure to process Signal 14

that results in recursive Signal 11 which leads ASE to dump core. This instrumentation

adds a worksround by blocking signals during select() system call.

Data Only Locked table with SORTED_DATA option may cause DBCC CHECKTABLE to report 9996 errors;

DELETEs on the table may fail with 644 errors.

the modules 's_handle', 's_normalize' and 's_renormalize' is written to the errorlog when executing

a stored procedures which gets recompiled. During the recompile Msg. 208 'Object <object> not found.'

is raised.

and inserts to this table fails producing stack trace in the errorlog file.

being cached in error.

is quiesced.

can access it.

records for updates of DOL tables that are replicated. This

only happens for replication log records and can affect the

repagent.

"You cannot include a union clause in a subquery unless it is part of a derived table expression within the subquery."

The Customer is running a Query with a Union inside a subquery See below. The Error Message Also Below Suggests that the parser needs to have the Parens match up before the UNION. This syntax seems to be support by DB2???

size > 4G, then it may lead to incorrect size calculation and may result in extended cache corruption.

expression containing SELECT and UNION whereby one side of the UNION contains a subquery.

SQL statement in the IF block is cached using the Statement Cache feature.

affected by the pass-through query.

that is qualified with the database name that is enclosed in double

quotes and SET QUOTED_IDENTIFIER is ON.

ASE 12.5.0.3 and later versions on IBM 64bit platform built on AIX 5.1 OS

(ASE version string reported by 'dataserver -v' has "AIX 5.1" rather than "AIX 4.3.3").

- Have added/modified some 20 stored procedures ( see CR description for list of SP )to support Long Identifier ( 255 ) and used sp_autoformat to format the output.

qpgalaxy_amats test PASSED.

in mount.c

This new default behavior change eliminates unused memory reserved for

EJB server when EJB server is not configured so that more user connections

can be configured for the ASE server.

ASE may fail to start due to a timeslice error in module 'dbt__getfree' after

the message 'Opening Master Database ... ' has been reported in the error log.

Traceflag 7843 can be used to reverse the changes from bugfix 269372 for situations where

an existing application relies on the previous (incorrect) behaviour of ASE attention handling.

during implicit conversion of INT NULL value '####' to a SMALLINT field".

This error occurs in databases where user id in SYSUSERS is greater than

32768.

'lm_internal_ctxchain_release' if a session has acquired a very large

number of locks. If the lock manager spinlock is held at this point,

the server will crash.

Edit the $SYBASE/RPL-12_5/bin/aserep.[bat|sh] script and add the "-Dfile.encoding=xxx" parameter to the JRE command line where "xxx" is a valid *Java* charset. For example:

"%JRE%" -Xmx64m "-Dfile.encoding=Cp850" "-Dinstall.root=%SYBASE_DIR%\RPL-12_5" -classpath

"%SYBASE_DIR%\RPL-12_5\classes\aserep.jar;%SYBASE_DIR%\jConnect-5_5\classes\jconn2.jar;%SYBASE_DIR%\jConnect-5_5\classes\jTDS2.jar;%CHARSETS%" com.sybase.replite.RepLite %*

The Java charsets that display the French unicode characters correctly are cp437, cp850, cp852, and cp858, specifically, the charsets that are MS-DOS compatible. Other Latin charsets do not work and display "corrupted" characters.

and we try to debug that stored procedure using the sql debugger, we may

hit a null pointer exeception.

Stacktrace with the following modules may be seen on the errorlog:

sqldbg__sendtrace, dbgrpc_control.

sql debugger will flag an error message like the following

(sqldbg) next

SQLException: JZ0SG: CallableStatement

返回的输出参数比应用程序为之注册的要少。

(sqldbg) next

SQLException: JZ006: 捕获到 IO

例外:com.sybase.jdbc2.jdbc.SybConnectionDeadException: JZ0C0: 连接已关闭。

where as Kernel Async disk I/O is in effect. This is because the O/S KAIO functionality

moved from the vsys_io driver making direct KAIO system calls from librt.

'print deadlock information' is active has been enhanced such

that the line number and the nesting level of the given procedure

are also reported. The application, host, and transaction names

will be reported.

or HAVING clause with GROUP BY aggregate may return incorrect results.

The same problem may be seen in queries with the referenced column coming from a join.

The worktable transitive closure optimization is now only enabled via -T4416.

Note queries using <COLUMN> <OPERATOR> <VALUE> construct are not affected.

from 5 to 25.

Edit the $SYBASE/RPL-12_5/bin/aserep.[bat|sh] script and add the "-Dfile.encoding=xxx" parameter to the JRE command line where "xxx" is a valid *Java* charset. For example:

"%JRE%" -Xmx64m "-Dfile.encoding=Cp850" "-Dinstall.root=%SYBASE_DIR%\RPL-12_5" -classpath

"%SYBASE_DIR%\RPL-12_5\classes\aserep.jar;%SYBASE_DIR%\jConnect-5_5\classes\jconn2.jar;%SYBASE_DIR%\jConnect-5_5\classes\jTDS2.jar;%CHARSETS%" com.sybase.replite.RepLite %*

The Java charsets that display the French unicode characters correctly are cp437, cp850, cp852, and cp858, specifically, the charsets that are MS-DOS compatible. Other Latin charsets do not work and display "corrupted" characters.

Edit the $SYBASE/RPL-12_5/bin/aserep.[bat|sh] script and add the "-Dfile.encoding=xxx" parameter to the JRE command line where "xxx" is a valid *Java* charset.

For example:

"%JRE%" -Xmx64m "-Dfile.encoding=Cp850" "-Dinstall.root=%SYBASE_DIR%\RPL-12_5" -classpath

"%SYBASE_DIR%\RPL-12_5\classes\aserep.jar;%SYBASE_DIR%\jConnect-5_5\classes\jconn2.jar;%SYBASE_DIR%\jConnect-5_5\classes\jTDS2.jar;%CHARSETS%" com.sybase.replite.RepLite %*

The Java charsets that display the French unicode characters correctly are cp437, cp850, cp852, and cp858, specifically, the charsets that are MS-DOS compatible. Other Latin charsets do not work and display "corrupted" characters.

sqlloc GUI tool fails to invoke the charset utility to install

a new sort order in syscharsets, prior to attempting to switch

to the new character set/sort order, and prior to installing

localized languages. As a result these latter steps fail.

file in the directory where the configuration file resides that has file name the same as

the configuration file and the suffix is larger than a long. For example, if the cfg

filename is ASE125.cfg and there exists a file with filename ASE125.2999999999, then every

time sp_configure is used to modify a configuration parameter, error 5854 is raised and no

configuration changes occur.

CASE expression contains a existence subquery, such as 'col1' in:

create view v1 (c1, col1) as

select c1, case when exists (select * from t1 where t1.c2 > 100) then 'a'

else 'b' from t2

A query that contains a reference to 'col1' in a OR-clause or IN-list,

such as:

select c1, col1 from v1 where c1 = 5 or col1 <> 'a'

May return an incorrect value for 'col1'.

reported to the errorlog together with a stack trace which includes

the modules 'sybase_asejdbc_Param_nt_1setBigDecimal' and 'j2c_setVal'

when executing a java method that uses ASE internal JDBC driver

and tries to convert from BigDecimal java datatype to NUMERIC SQL

datatype.

expressions in their target lists.", is raised when executing a SELECT from a view which has

columns defined and containw an UNION with a SELECT and GROUP BY clause on each union-side.

of "Bind Default".

may be corrupted and error 7928.

If program_name is not set by the client application, process name of the client will be stored in program_name column of Sysprocesses.

Starting with the 12.5.1/12.5.2 installation guides, the platform specific "Adaptive server specifications" only contain default stack size, minimum RAM and memory per user connection. The generic Configuration guide for Unix contains the rest of the specifications that used to be in the platform specific guide (however nothing for NT or Linux)

Suggestion to RE-add the missing information back to the platform specific guides, or document how to retrieve this information once the ASE server is running:

dbcc traceon (3604)

go

dbcc serverlimits

go

Here is what is documneted now inteh 12.5.2 install guide for NT:

Table 1-2: Adaptive Server specifications on Windows NT Hardware and memory

Minimum RAM required for Adaptive Server

48MB

21,504 2K pages

Minimum RAM per additional user

63K

With default stack size, packet size, and user log cache size. This value depends on the connection needs. See the System Administration Guide for information about configuring memory.

Default memory per additional user

402K 100K

with Java enabled. with Java disabled.

Java is enabled by default.

Default user stack size

41K

triggered by high network activity. Incoming network I/O is very high, ASE engines use 80% to

100% of CPU at Operating System level, but do not perform any useful work. New incoming

connections are most of the time refused, effectively ASE seems to be hung.

UNIT_NORMALIZE (20), RESOURCEMONITOR_VALUE (22) and SETCONFIG_NOWRITE (23).

more accurate information when exceptional behavior is encountered within ASE.

Make it possible to start replication agent with some delay time after starting ASE.

This could happen when a large number of "open objects", "open databases"

or user-defined caches are configured so that following an upgrade, Adaptive

Server may be unable to boot. The following message, "The value of the

'max memory' parameter (<value>) defined in the configuration file is not

high enough to set the other parameter values specified in the configuration

file...", is reported in the error log.

if more than one worktable is necessary to resolve aggregates from the view.

of tables/indexes, checkstorage may produce a timeslice whose stacktrace contains the

function 'pchk_pt_sort_grp'. Increasing the 'cpu grace time' parameter may alleviate

this problem.

- "Utilities" folder: launch Interactive SQL requiring new logon infomration

- Servers: launch Interactive SQL and connect to the servers

- Database: launch Interactive SQL and connect to the server of the database and switch to the database

- Tables and views: launch Interactive SQL to view table or view data

So after I have fixed the device issue and restarted ASE all if fine.

There doesn't seem be a need of marking the database suspect.

This nice change in behaviour is due to the parallel recovery feature.

table (table id ...) column does not allow NULLS. UPDATE fails."

may be raised when a java class running inside the ASE's JVM is

connecting to the server using the internal JDBC driver and then

executes a query on a table that has wide columns and a worktable

is necessary to resolve that query.

When generate an XML model from an OOM, cardinalities defined on an association are lost.

Steps to reproduce

New OOM : class diagram, java

2 classes with few attributes

one association between those classes

association properties

multiciplicity :

class_2

0..1

class_1

4

generate an XML

the concept of cardinality is lost

For the customer : the generated XML should reflect this with constrains

Request to implement one of the following:

1. Require a full path to be entered at disk init.

2. Append the current working directory to the device name when a full path name is not entered.

3. Raise a warning if a device is created without a full path.

time to stop when interrupted with Ctrl C or killed from another session.

if ASE executable sqlsrvr.exe is not found in the current working directory.

does not synch the password on the companion side.

To work-around, person with "sso_role" could run "sp_password" on the account whose

password has expired from either of the companion servers.

in "Chapter 2 Understanding Component Integration Services",under "Proxy tables" section, under the "Example of remote table definition" there is an issue that a customer has. The example shows how to create a remote table, under the assumption that the reader is login as "sa". We need a Note or a Step stating that if you are not logon in as user "sa' then you need to logon in as "sa" and under the "master Database" and "grant connect to USERID", then relogin as the USERID. Even thought this is documneted else where in the CIS DOC and in the Reference DOC under "Connect TO".

00:00005:00001:2004/06/02 12:47:20.75 worker Client message: ct_get_data(): user api layer: external error: Item 1 has already been read.

trailing spaces in the VARCHAR column if the source of the data is a table with CHAR column.

object id, index id and additional context information is printed

to the errorlog.

has been modified to 5 minutes. Any external transaction such as

XA transaction will be rolled back, if in detached state, for more

than 5 minutes. A value of 0 will disable external transaction timeout.

IPv6 support ASE_JAVA applets failed to establish IPv4 oriented network communications.

in an outerjoin-clause can return incorrect results.

'ap_generate' and 's_ap_capture' is printed in the errorlog when "abstract plan dump" is

configured and a SELECT containing a UNION with a simple SELECT like SELECT "<string value>"

is executed.

Chapter 29 - Restoring the System Databases under "Recovering the master database" are misleading. Instructions in 12.5.1 SAG to restore master device are misleading and there is also techdoc # 1028123 - Recovering an Upgraded Master Device and master Databases in ASE 12.5.x, that points to the SAG. (generic problem) Instructions should also differ based on 12.5/ebf versions. The Techdoc needs to be fixed, until the SAG can be corrected.

JRE follows the following steps:

1) The SYBASE_JRE environment variable

2) Search for JRE 1.4 at $SYBASE/shared-1_0/JRE-1_4

3) Search for JRE 1.3 at $SYBASE/shared-1_0/JRE-1_3

"KILL <spid> WITH STATUSONLY" will mistakenly report the message "Status

report cannot be obtained. ROLLBACK for spid:<spid> is not in progress.".

block in a timely fashion within the ASE.

the server's default character set is SJIS and the client's character

set is EUCJIS.

target database uses more than one database device and the dump file is created by backupserver

which is started with -m option and/or BLOCKSIZE option of DUMP command.

Since 12.5.0.3, The Drop login account needs sso_role not sa_role.

'max native threads per engine' controls the maximum number of native threads that an engine can spawn. If this limit is reached, the ASE thread that requires the native thread will sleep until another thread releases a native thread.

'rtm thread idle wait period' is the idle wait period (in seconds) that a native messaging thread will wait for work. If this period is reached without new work, the native thread will fade itself out.

Encyclopedia of Tasks", under "Other Useful Tasks", "How to Reduce the Size of tempdb"

8- Verify that tempdb has one entry that looks like this:

1> select * from sysusages

2> where dbid = db_id('tempdb')

dbid segmap lstart size vstart

--- ------ ----- ---- ------

2 7 0 1024 2564

The above information is incorrect, it was correct for 11.x, but as of ASE 12.X the above information is incorrect.

to a file via a proxy table mapped to an external directory results in stacktrace.

Stored Procedures to fail and make XP Server crash under any exceptional ESP.

stack size is insufficient to run a query. Instead the text "The transaction

was aborted because it used too much stack space. Either use sp_configure

to increase the stack size, or break the query into smaller pieces..."

should be reported.

bound to a named cache using the 'Relaxed LRU' strategy.

In addition to stack space, this new limit may result in a need to allocate more procedure cache space to the server.

The dataserver must be started with traceflag 404 ( -T404 ) to allow this new limit to take effect. Otherwise, the current limit of 1024 will remain.

This causes Open Server based applications to get the error "Output

parameter not found" and consequently makes XA incompatible with Open Server.

and 'sysindexes' when a session is in the process of dropping a replicated

user table, stored procedure or function.

ASE should not allow insertion of such values if they are invalid values.

SELECT INTO command, some premature out of data space conditions may

be hit by commands like INSERT, CREATE TABLE, SELECT INTO, BCP, ...

whereas 'sp_helpdb' shows that free data space is available.

Alternatively, for mixed log and data databases, 3475 errors may

be raised under the same space usage conditions.

while bringing a previous database online.

Support for HP ServiceGuard A.11.15 on HPPA 32 and 64 bit platforms.

In pre 12.5.2 release such queries would report an error that the maximum number of 16

subqueries has been reached.

but currently there are only <value> auxiliary scan descriptors available.

Either raise the value of the 'number of aux scan descriptors'

configuration parameter or try your query later." can be raised

during the execution of a complex query despite having increased

the 'number of aux scan descriptors' to the appropriate value.

sql.ini entry that represents this listener.

with a stack trace which includes the modules 'q_indexsel_preprocess' and 'q_select_best_index'

may be reported in the error log when a UNION ALL view is joined with another view that

includes a DISTINCT or GROUP BY clause.

'seq_schema_check_virtab' may be reported in the error log when

an execute cursor is chosen as the deadlock victim during the

open phase of the cursor with a stack trace showing modules

'curs_execute' and 's_checkforrecompile'. Alternatively the

stack trace may show modules 'mapsvs' and 'curs_fetch'.

of that name found." will be reported if a named transaction is rolled back

in which a SELECT that required a worktable (ORDER BY or DISTINCT for example)

preceeded any transactional activity and if the database has replicated objects.

'print deadlock information' is active has been enhanced such

that the database name and the index ID values are reported.

the sortorder specified by the "default unicode sortorder" configuration parameter. However,

it always uses binary.

non-existent database name.

($SYBASE/$SYBASE_ASE/sample/XML) generates different code for client and server,

to accomodate differences between jCommect and the native JDBC driver.

It is generating a SET STRINGSIZE command in the server that should only

be generated for jConnect. The fix to this CR corrects that.

The CR also adds two new options to the 'forxmlj' function:

xmlcheck={yes|no} and xsidecl={yes|no}

These are useful in nested calls of the forxmlj function,

as described in the documentation:

$SYBASE/$SYBASE_ASE/sample/XML/Using-SQLX-mappings.htm

longer blocked, if a new lock request comes in then it may be queued behind

the re-queued locks although it is compatible with them. This situation is

rare, but it will result in unnecessary blocking situations, for example a

'shared intent' table lock that is granted to task A may block a 'shared intent'

compatible table lock requested by task B.

message when the remote server is not available. With this change, more meaningful

error messages 11206 and/or 8013 will be reported in the event of a transaction

failure caused by the remote server shutdown.

may appear in the error log followed by a stack trace including the module

'ubo_object_from_slot' and 'smget' or 'smdestroymbox' when a query or a

sort operation was executed in parallel and there was a lack of 'Mail Boxes'

objects.

Either use sp_configure to increase the stack size, or break the query into smaller pieces."

may be reported when executing some some basic queries in 64 bits platforms,

if the stack size is configured with the minimum value allowed.

Likewise, upgrade to 12.5.x 64 bits version from older versions may fail if

the stack size is configured to that minimum value.

maybe reported when executing a SELECT INTO statement using

views in the FROM clause together with a UNION and a GROUP BY clause. If the same query is placed in a stored procedure, it will endlessly get renormalized until the database runs out of space.

punctuation. Formerly implemented under command line traceflag -T109,

this is now the default behavior. Traceflag -T109 is now ignored.

A new traceflag, -T114, is introduced to invoke the former default

behavior of allowing equivalence mappings of non-ASCII characters

to participate in SQL language parsing.

subqueries that calculate MAX or MIN aggregates and the aggregate

is on a table that is not correlated with the outer table.

Error message "Invalid parameter value specified. Refer the documentation for

the correct value(s). A signal(11) was raised while executing this ESP.

If this is an user DLL check the code else contact Sybase Technical Support."

will be sent to the client.

results from a SUBQUERY, to a VARCHAR variable.

may not work correctly.

with a stack trace which includes the modules 's_reinstallProcTree' and 's_getTreeOrPlan'

may be reported in the errorlog when executing a stored procedure that calls functions

from java classes and the stored procedure had to be recreated internally using the text

stored in 'syscomments' system table.

containing the modules 'exec_eop', 's_execute' and 'sequencer' when an assignment SELECT is executed

using a VIEW with a UNION clause and the "set fmtonly" option is set on.

when the LDAP server did not accept LDAP V2 protocol.

LDAP User Authentication failed to find the result set with "no results"

from an LDAP server when it did not return the result as an attribute

name and value.

page chains, is reported when dbcc tablealloc is run on a single data partition.

a stacktrace containing the modules 'apa__struct_plan', 's_ap_apply' and 's_compile_stmt' when

a SELECT INTO statement is executed based on a join together with the PLAN-clause using

an AP join operator describing the AP-plan for the join.

(*1*, *2*, *3*, *4*) these work as expected...

Additionally, some messages related configured shared memory dumps do not print the correct

unit (i.e. "bytes").

contraint and the SELECT part of the query has a subquery which is resolved using reformatting strategy.

a stack trace which includes the modules 'norm3_driver' and 'vu_cleanup' may be reported in

the error log when a nested view with more than 16 subqueries is recompiled.

nested top action, a divide by zero error may occur resulting in

the failure of the recovery of a database.

the modules 'clean_process', 'kill_proc' and 'terminate_process" is printed in the errorlog

after killing a spid in the server using the KILL command.

Chapter 26:Checking Database Consistency,

Checking consistency of databases and tables.

Part of the problem doc listed below, we aslo see the same problem in ASE 12.5 Chapter 25. The doc has a sentence stating: "At least two workspaces to use during the check operation. See "dbccdb workspaces" in Chapter 59, "dbccdb Tables" in the Reference Manual." There is no Chapter 59.

In SAG, the examples are correct.

It should be in Chapter 13 under the dbcc_types table write-up in Reference Manual

when a spid within ASE is killed ('kill <spid>') it could result in ASE hang.

Running pstack (unix command) on ASE pid would show functions kill_proc, drop_connection, tli_nclose,

_tx_getinfo, _t_do_ioctl, ioctl on the stack. This problem is due to Solaris bug # 4924525.

If you are using tli type entries in the interfaces file that is used by ASE contact SUN

quoting Sun BugID 4924525 and request a patch (110955-05 or later) to fix the problem.

Shared JVMs within ASE

Request Title:

Implement shared system and database level JVM processes within ASE

Product Relevancy:

ASE 12.5+ and ASE 15

Request Description:

This request is to implement a JVM feature within ASE that would support shared static java attributes. The current implementation of Java within the ASE database is such that each database connection requires its own instance of the JVM process. Unfortunately, this implementation prevents the full support of static attributes in accordance with the Java specifications.

Implementation:

Ideal Approach - Ideally the implementation would consist of a single JVM running at the server level, utilizing a port of ASE threads as a "native" threading model. Java supports substitution of threading models since 1.1.81. This would allow for a 1:1 mapping of the database threads to the JVM threads. The hierarchy of class loaders would provide for application level "sandboxing". For example: a "server level" class loader (loading the classes at the master database level (not visible to everyone), followed by the "shared class loader" loading all the classes at the sybsystemprocs level, and a class loader at the database level inheriting the sybsystemprocs class loader and loading the classes installed in this database. Switching the database context, would append the java class loader references for the new database to the class loader chain (if it is not there already). A new configuration parameter for the database will determine the order of the class loader overrides (parent first/child first).

Alternative Approach - Alternatively, the implementation would consist of a configurable number of sharable JVM instance (similar to how named caches are implemented). This would provide the ability to have logical groupings of JVM's to provide specific functionality in addition to providing for a sufficient number of JVM's to by-pass any potential JVM threading limitations.

Specified classes (or an entire database of classes) could be bound to a 搒hared� JVM. Rules would need to be defined to handle naming conflicts when trying to load the same class name from two different databases or the classes will need to be restricted to a single location (i.e. database).

These 搒hared� JVM instances would coexist with the current connection-oriented JVM implementation. Classes not bound to a shared JVM would act as they currently do (execute within their own running JVM processes).

Overall Benefits:

Performance and Manageability (Administration)

Specific Benefits:

If we can get this working in ASE in accordance to java specifications, this will resolve a few extremely important issues and outstanding enhancement requests (customizable Identity implementations, Global Variables, etc) that were made through ISUG and also exist as CRs.

Customizable Identity Implementation:

Through the use of static object attributes, it would be possible to implement functionality which assigns (and reserves) sequential values similar to the ASE Identity but be flexible enough to support customization. The functionality could also be wrapped in java functions making it possible to assign using batch INSERT/SELECTs.

Example #1 - A java function could be coded to identify and reserve the next sequence number for a table. The function call might be called next_sequence (<table name>, "reserve").

Example #2 - A java function could be coded to also partition the values based on the application requirements. The function call might look like next_sequence (<table name>, "reserve", "Cumberland").

Example #3 - The same java functions could be coded to also backfill gaps in sequences instead providing the "one-up" (max+1) values. In other words, when assigning the next sequential value, it would scan the table to identify the next available gap in the sequence.

Global Variables:

Through the use of static object attributes, it would be possible to implement global variables. The global variables would not just be simple variables, but could also include structures and arrays. This has been a long awaited request. There have been many requests for this feature in ISUG and in CRs (Feature Requests).

Real Time Messaging Services (RTMS):

If shared JVM's would optionally load classes on startup, hence invoking constructors, this would allow some JVM's to function as running business processes somewhat detached from a user spid. This may be really useful in view of RTMS. The process could start looking at a message bus for pending messages upon server boot.

1 see JNI Programmer's Guide and Reference on the Sun's java website.

given the ID of the index and the IDs of the database and the object on

which it is defined.

A number of MDA tables return the database ID, object ID and index ID trio.

This built-in would simplify the process of analysing these MDA tables for

index usage.

644 "Index row entry for data row id (<value>, <value>) is missing from index page <value> ... Drop and re-create the index." may be reported.

Customer would like to have a feature request such as a "with log" option for the minimally logged commands so he has greater control of the logging:

change ... retry the command later." may be reported when running DBCC

TABLEALLOC or DBCC CHECKALLOC in fix mode, resulting in the reported page

being incorrectly deallocated.

Addresses the errors 2821, 2812 and 223.

may fail with message "current process (<value>) infected with 11" in module 'bufdlink',

or 'des_add_dirty_chain_elmt'; this is another variation of the bug reported under CR 360414.

The 'CREATE TABLE' command is not allowed within a multi-statement transaction

in the 'tempdb' database." will be raised when the stored procedure sp_modifylogin

is used to change the login script in a HA environment.

rows into sysloginroles, syssrvroles and other catalogs that

have special SSO_ROLE privilege checking.

zh_CN.GB18030, dataserver/isql command fail.

in the module 'pipe_sqltext' together with a stack trace which includes

the module 'sqlpars' may be reported in the errorlog when "enable monitoring"

(or "sql text pipe active") configuration parameter is activated.

ubo_object_from_slot() function as the instigator. These messages are harmless but should not be produced.

failover for an ASE Active-Active HA configuration.

Product Relevancy:

ASE 12.5%2B and ASE 15

Request Description:

During a failover event, cluster software can be configured to

switch a floating IP address over to a different host. In an

ASE Active-Active HA configuration, the floating IP address

currently cannot be loaded into Active companion ASE server.

This feature request is to request that the floating IP address

be loaded into the Active companion (ASE 12.5%2B) or a surviving

ASE server (ASE 15.0).

Implementation:

As part of the failover event, the OS cluster will switch the

"owner" of a floating IP address to a surviving host. After the

OS cluster has completed this task, it should be configured to

trigger the Sybase HA scripts to execute the new ASE dynamic

listeners feature within ASE to load a floating IP address.

Overall Benefits:

Availability and Manageability (Administration)

Specific Benefits:

Automated Failback (improved Availability):

Failback could be automated and not require a scheduled downtime.

Since a surviving ASE would "know" which connections were failed

over (the connections associated with the dynamically loaded

floating IP address), the process to failback would be the same

as the process to failover except in conducted in reverse. The

biggest (and probably only) difference would be that the

connections to be failed back would be intentionally disconnected.

Easier driver setup / administration (improved Manageability):

Several existing 3rd party tools that connect to ASE are not HA

aware and rely on underlying connectivity drivers

(ODBC, JDBC, native, etc) for migrating connections. For the

ASE 12.5 drivers, this usually requires setting them up with

two network addresses (primary and failover). Administration

could be kept simpler if the drivers did not need to be setup

with the network address of each HA companion (or cluster) node.

Clusters w/more than 2 nodes (improved Manageability):

Although ASE 12.5 supports a two node Active-Active cluster

(companion), ASE 15.0 is being redesigned to be capable of

supporting more than two nodes. Therefore, administration

would become more difficult if each connectivity driver needed

to be configured for each node within the cluster (especially

as nodes are added / dropped from a cluster). In addition, if

the connectivity driver is setup with fewer number of network

address than nodes in the cluster, there exists for the

possibility of system outages (application downtime) and

thereby reduce availability.

the modules 'logcommand', 's__altertab' and 's_execute' is printed in the errorlog

when executing an ALTER TABLE statement while DDL replication for that database

and statement cache are enabled.

java version "1.3.1_10"

Java(TM) 2 Runtime Environment, Standard Edition (build 1.3.1_10-b03)

Java HotSpot(TM) Client VM (build 1.3.1_10-b03, mixed mode)

I tried using Java version:

java version "1.4.2_03"

Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_03-b02)

Java HotSpot(TM) Client VM (build 1.4.2_03-b02, mixed mode)

And the mouse seems to function properly in Sybase Central. The customer can use this as a workaround until such time as Sybase Central supports JRE version 1.4.

fails with the message "Device activation error. PHYSNAME '<filename>' may be incorrect".

ASE errorlog also reports "kernel dopen: open '<filename>', Invalid argument".

variables defined by the caller of the execute immediate.

"Adaptive Server has truncated parameter number %d passesd to rpc '%.*s' on server '%.*s'.

The remote server does not support character or binary data wider than 255 bytes."

a stacktrace containing the modules 'run', 'exec_eop' and 'execute' when using

the STR() built-in function when the parameter 'length' is not a constant but

an expression evaluating to the value bigger than 255.

data having long running transactions suspended on the LCT.

fails with an error "Device activation error. PHYSNAME '<device path>' may be incorrect"

as expected, but zeroes out the page on the device that would have been the last page

of the new device as described by the disk init. This results in data corruption.

may hit a Security password issue.

if the dealloctaion is done during trigger or internal sql execution.

in the LOAD ... WITH LISTONLY=FULL command when loading across platform.

(2)On linuxia32 and ibmplinux platforms, if a dump file size on a tape or a capacity value

exceeds 4GB, an append of a dump file to the tape could fail or fail to report an appropriate

error message which tells there is no room to append on the tape.

attempt to dump on the '/dev/null' device will be successfull.

Ideally any dump on '/dev/null' should fail.

has not been recovered, then when recovery bypasses that step with message 12544,

"Not recovering temporary database ... ", it will print a unpredictable database ID

instead of the database ID of the user tempdb whose recreation is to be bypassed.

same column.

the stored procedure the session may be killed with the stack trace and error 216:

"Attempt to automatically drop temporary table failed." This error is spurious

and no temporary object will remain in the temporaray database. This problem may

happen only if a stored procedure creates non temporary table.

for cross-platform database load, if the database has many disk fragments in Sysusages.

The message "current process infected with 11" in the module 'lddb__stat_translate'

together with a stack trace which includes the modules 'lddb__unscr_stat_page'

may be reported in the error log, when loading a database across platform,

if the database has an orphan row in Sysstatistics.

the module 'proc_invalidate_plans' with a stack trace in the error log

when the value of the configuration parameter 'max parallel degree'

or 'max scan parallel degree' is reset to its default value of 1.

This will result in an implicit shut down of ASE. The modules

'cfg_notify_max_plldeg' or 'cfg_notify_max_scanplldeg' will also

be reported.

is added. A user can specify authentication mechanism as 'kerberos' as follows,

sp_modifylogin 'user1', 'authenticate with','KERBEROS'

2> a user can get authenticated against specified mechanism, even if configuration options 'use security services','unified login required' are set to 1.

with password has been activated.

to GRANT the same privilege WITH GRANT OPTION to an individual user.

It should be "The length of a unichar/univarchar ....."

metadata descriptors for the number of tables and procedures in use by Adaptive Server. Configure

the serverwide 'number of open objects' parameter to accommodate more objects." is raised when

a procedure is executed using a name like "*<name>".

such as Linux 32 on Intel, Solaris 32 or 64bit, etc.

As a side effect of this fix, compatibility from older dump on Tru64 is not kept. To

workaround incompatibility issue, you could use trace flag -D128 with the backupserver

binary. This way you would be able to load older dumps.

sent via PRINT statement are silently ignored without printing.

process, Adaptiver Server error handling may not be graceful and may fail to recover

subsequent databases.

in the LRU/MRU chain in data caches using "strict lru" strategy.

join order might have been underestimated.

the bug fix 354768. This is an informational message only and can be safely ignored if EJB

is not configured. ASE 12.5.3 alters this behavior so that this message appears only when

"enable enterprise java beans" parameter is set to 1 in the config file.

Also, this fix prevents the creation of a Scheduled Job when a duplicate Job name is provided in the first page of the wizard. Previously, it may have looked like the user created a new Job and a new Scheduled Job when what actually happened is that a new Scheduled Job was created using an existing Job. To create a Scheduled Job using an existing Job, the user should use the Job Scheduler wizard.

raising the expected message "Msg. 128, The name 'MS' is illegal in this context. Only constants,

constant expressions, or variables allowed here. Column names are illegal.".

to improve the performance of DROP TABLE in temporary databases was active

by trace flag 3706 at boot time or via DBCC TRACEON(3706). This enhancement

now become the default ASE behaviour. Consequently by default ASE will release

during the drop of user tables in tempdb the system catalog locks when not

in DDL-IN-TRAN mode after the commit of the transaction but before post commit

work started. The trace flag 3706 is therefore obsoleted. Should there be a need

to turn off this enhancement then trace flag 3708 can be used.

leading to stack overflow errors. Module name qremap__tree is printed

in the errorlog.

The problem is generic but is likely to happen more frequently on linux platform.

mid-way through an ALTER TABLE UNPARTITION operation. These errors can also

occur while accessing tables after loading a transaction dump containing

uncommited changes for an UNPARTITION operation to such tables.

(Fixes issues raised in 400988.)

return correct result if the statement cache option is turned on and the parameter

passed to to the builtin is host variable.

dictionary.srt, espdict.srt, espnoac.srt, espnocs.srt, nocase.srt, noaccents.srt and nocasepref.srt.

the stack trace may be written to the errorlog when the server encounters an error during

the recompilation of the stored procedure using cursor on the temporary table.

statement cache configuration expectation.

The way the customer describes this it would possibly take a two step process...

1. exec sp_audit to start auditing for all logins

2. exec sp_audit to exclude individual logins

may occur indicating that the cursor is READONLY.

With this change, remote tables may need to be created with a unique index for table

scans on updatable cursor.

the grid tool does not support a char column with over 256 characters. Only a portion of the data is displayed. The issue is if you have results large then the window there is no way to scroll to see that data off the Window.

the host and login name relating to the lost connection. for example,

"extended error information: hostname: TONYI-XP login: sa".

security context will be partially lost and the user will not be able

to revert its identity. The output of sp_who will show that both the

original name and the current name are set to the faked user's name.

The next callout calculations were not correctly handling cases where the schedule starttime was after the schedule endtime, that is a schedule that wraps over midnight. The net effect of this error was that the schedule was being offset from the starttime when it passed midnight. The current fix handles this case correctly so schedules are rolled over midnight properly. The fix in effect recognizes that midnight, which is stored as a 0 in a CS_DATETIME, effectively needs to be viewed as the largest allowable value when approached clockwise during an operational schedule rather than as a 0.

In addition to the direct fix there are a couple side effects to this fix that users may see:

1) End times are now enforced correctly. This means that when a recurrence interval falls on the end time the job will not run. The current situation is that a job can actually run past the endtime because the next callout calculation was allowing the next callout time to be outside the endtime boundary so the job would be run past the endtime and then the following callout time would move up to the next starttime. This was clearly incorrect and has been fixed, however this represents a change in current behavior.

2) Schedules where starttime equals endtime, either as default values or explicitly set are treated correctly. That is the schedule will run in continuous intervals without interruption unless constraints (day of week/date of month) are set.

The internal date is used for things such as the crdate field in sysobjects and sysindexes, the loggedindatetime in syslogins, WAITFOR commands, and the datetime stamps in begin tran and commit tran log records.

IN ASE 12.5.1 Doc "Reference Manual Tables"

This is the current description for the currauth column:

A bit mask that indicates which roles were active for suid at the time

the threshold was added or most recently modified. When the threshold

is crossed, proc_name executes with this set of roles, less any that

have been deactivated since the threshold was added or last modified.

Determining which roles were active based on the systhresholds.currauth

is not clear. Customer requests a documented explanation in the ASE user

reference manual for the bitmask values assigned to systhresholds.currauth.

leads to a variety of problems related to communication between the ASE Job Scheduler task

and its agent. Symptoms include a Job Scheduler hang (no response), ct lib connection

failures (as noted in the Job Scheduler agent log file), the Job Scheduler task error

'spurious states from jsagent', and log messages indicating a failure to receive a

communication buffer from either the Job Scheduler task or agent.

when trying to generate the DDL on a Unique Constraint object:

"DGP1 The Object type is invalid null.".

Add Option Types of: group, role

Examples:

sp_audit @option="group", @login_name="all", @group_name="mygroup", @object_name="all", @setting="on"

Turn on auditing for all users in all databases that are in group "mygroup"

sp_audit @option="role", @login_name="all", @role_name="myrole", @object_name="all", @setting="on"

Turn on auditing for all users in all databases that have the role "myrole" active.

These are suggestions not necessarily how to implement the feature.

access the primary device, mirroring succeeds but later ASE fails during the subsequent

operations with the errors such as :

- Error 692: "Uninitialized logical page <page id> was read while accessing object <object id> in database <database id>

or

- Error 822: Could not start I/O for request BLKIO < ...buffer details>

the size of single page of 16384 bytes." may be raised together

with a stack trace reported in the error log which includes the

modules 'pll_exec', 'copy__constant' and 'memalloc' when executing

a query statement that uses the CONVERT() function for converting

to TEXT (or IMAGE) datatype and the query is executed in parallel.

the procedure name is stored in a variable enclosed in squared brackets,

prefixed with the database name and executed from a different database.

created in ASE with the aid of sp_maplogin, an audit event for LOGIN COMMAND (103) is

not generated.

Chapter 5 Upgrading Adaptive Server --> Pre-upgrade tasks -->

Step 17:

Copy the following files from earlier versions of Adaptive Server to their corresponding Adaptive Server 12.5.x installation:

The issue is that Step 17 needs a Note specifying that coping the files are not required if the customer uses the SQLUPGRADE program to upgrade there server because SQLUPGRADE will copy these files over automatically, SQLUPGRADE needs these files to start the new server.

techdoc 1012256 needs to include a description of this particular situation and if a customer has this particular situation, describe that this recovery procedure will fail.

may be reported when a INSERT...SELECT or SELECT INTO with join and CASE

is evaluated to be true and Quickpass disabled.

trigger, or SQL batch..." may be reported after running many client or

language cursors in a server configured to use Abstract Plans in "capture"

or "apply" modes.

In order to avoid this problem, please specify ObjectID > 0 in WHERE clause.

instructions to do a reserved word check without upgrading

ASE in a non-GUI environment.

The installupgrade script is used to create sp_checkreswords in the

ASE that will be upgraded.

will be printed instead of the actual error message.

"Schema for table '<value>' has changed since compilation of this query.

Please re-execute query." may be followed by a segmentation violation

in module abortcmd() with a stack trace reported in the error log.

srvbuild created sybsystemdb 2MB on master and 3MB on its own device.

repeat the value for the "Current Failed login attempts" property for the

following item "Authenticate with".

1. With AIX and LINUX the initial dialogue in srvbuild and 5 dialogue screens deep within sqlupgrade...checkbox/radio_buttons options are unclear...the customer makes a selection and it appears deselected...the boxes change from gray to white(it may be yellow but it is so small that the open square looks white). The white(it may be yellow but it is so small that the open square looks white) makes them look empty and thus it appears that item is deselected.

2. Solaris and HP use a larger square box and the boxes turn yellow...so there is an inconsistant GraphicalUserInterface between UNIX platforms.

java.lang.ClassCastException: java.math.BigDecimal

at com.sybase.sybmigration.ObjectSelectForm$NextTask.doWork(ObjectSelect

Form.java:2082)

at com.sybase.sybmigration.MigrationToolThread.run(MigrationToolThread.j

ava:28)

it requires more than 24480 pages of memory" may be reported during CREATE INDEX on

a large table with the 12.5.3 server.

leaving core files due to SEGV and BUS errors.

read '<loginname>' is not a valid login name.

These locks may be left around as a result of an ungraceful termination of a connection.

bounds" together with a stack trace which includes the modules 'srchpage' and 'srchindex' may be

reported in the errorlog of an server with a 8 Kb or 16 Kb page size during a SELECT using an index which includes variable length columns.

gets interrupted by the end-user, a segmentation violation may be hit in

module 'sortclean' with a stack trace in the error log. The modules

'cri_physbuild', 'sort' and 'sam_gen_distmap' will also be reported.

self deadlock for a process that is trying to flush its user log cache when

the cache replacement policy for the given cache is "relaxed LRU replacement".

This allows you to customize it by setting the RPL_JAVA_MAX_MEM in either the instance RUN script or in your environment.

error--end of buffer missed" on the ASE console, while executing a query using the Abstract Plan

syntax.

the modules 'oh_init', 'yyparse' and 'sqlpars' is written to the errorlog when executing

an INSERT or UPDATE statement and supplying large amounts of text for a column of type text.

or use sp_help to check whether the object exists (sp_help may produce

lots of output)" maybe be raised when executing a stored procedure that

creates inside that object and it has been executed and recompiled

concurrently by several sessions.

Under the same circumstances sometimes a 3703 error is reported following by

one or several 216 errors and resulting in the termination of the session

executing that stored procedure.

1- Startime typo should be starttime

2- missing quote in sp_sjobcreate

3- Create a job the from step 1 to step 2 the jobname does not follow

In Chapter 2 Understanding Component Integration Services under "Proxy databases".

This statement below, is misleading, the last "." in the string identifier >>servername.dbname.owner.<< maybe confused as the end of sentence not rquired for proper syntax. it maybe better to quote the string

Example: >>"servername.dbname.owner.".<<

(and,and).......

Creates a proxy database named proxydb and,and brings over all remote tables from a remote database regardless of who created them:

the name of the previous file is truncated to the first 3 characters of the actual file name.

server, the warning messages 11294 "The remote server ('server_name') class

<id> is not supported by UPDATE STATISTICS in import mode." and "UPDATE

STATISTICS will not run in import mode." will be displayed. This causes

sp_text_notify on EFTS text index table to produce an error "RMTQUERY,

Execution of query 'update statistics <dbname>.<owner>.<text_index>' failed."

in the EFTS log.

is used in the constant expression being a part of OR clause. An example of such

statement would be :

select @var = value from tab where (@var is NULL and col = value ) OR (expression)

is accessing system table.

(msgsend(), msgrecv(), etc) The following error is reported

Incorrect syntax near 'Invalid pointer param number 2, pointer value 0x10'.

executed on RedHat Linux variant RHEL 3.0 version

if ASE executable sqlsrvr.exe is not found in the current working directory.

installmsgsvss script is not included in ASE 12.5.3 release.

Also, instructions for post installation tasks on hppa64 platform

for RTMS is missing in ASE 12.5.3 Coverletter.

transaction isolation level 3 (holdlock) may return an incorrect result set.

in database '<temporary_database_name>'", may be reported if a session is interrupted

with ^C during the dropping of a temporary table and the 3706 trace flag is turned on.

The interrupted session is terminated abnormally and may not have its resources cleaned up.

corrupt" may be reported in a server with a multi byte character set when data inserted in a varchar

column is longer than the definition of the column. This can happen for an insert using binary values

for the character column.

operating after having reported the error 9289 due to an inconsistent

log record found. Instead it will attempt to continue the replication

after having reported the error 9290 in the error log.

stored procedure with the recompile option. The stored procedure contains a derived table expression

which is referencing a temporary table column. The message is raised when the procedure is executed

having a different @@tempdbid from the one the procedure was created under.

1. The comment at the beginning of sp_helptext is wrong. It states that 18406 is a problem with "allow select on syscomments.text". However, 18406 is raise when sp_hidetext was run for the object NOT when "allow select on syscomments.text" is set to 0. Note that the quoted text is also incorrect; there is no "allow" in the configuration argument.

2. There is a next comment:

/*

** If the configuration parameter 'allow select on syscomments.text' is set

to

** 0, then the user can use sp_helptext ONLY in the following cases

**

** 1. if the user has sa_role

** 2. if the object is a system stored procedure

** 3. if the object is owned by the user

**

*/

It should include dbo as well because dbo or dbo alias could also run sp_helptext.

3. The text for error 18180 says: "Configuration Parameter 'allow select on syscomments.text' is set to 0. Only the object owner or user with sa_role can run sp_helptext on this object." Again there should be no "allow" in the parameter name and the message should includde dbo as well.

errorlog containing the modules 'curs_open', 'execute' and 'exec_eop', when a cursor statement

contains a SELECT with a UNION- and a JOIN-clause using the REFORMATTING strategy, when

the OPEN of the cursor is executed. The set option FMTONLY must be set on to see this issue.

DBCC CHECKCATALOG command run for the SYBSYSTEMPROCS database.

Chapter 19 Transactions: Maintaining Data Consistency and Recovery, under

Transaction modes and stored procedures, sp_proxmode should be spelled sp_procxmode.

http://sybooks.sybase.com/onlinebooks/group-as/asg1251e/sqlug/@Generic__BookView?DwebQuery=sp_proxmode

This will occur sometime after a different task encounters a fatal error,

such as a timeslice error, while performing disk I/O. The stacktrace for the timeslice error

will include functions such as aioread() or aiowrite(), kaio__daiostart(), and basis_daiostart.

data-retrieval operations.' is raised when the result of derived table expression is assigned to

a local variable. With fix the behavior is changed and it's now allowed to do so.

will be raised if a procedure fired when a threshold is crossed tries to access an MDA table.

this parallel query. Increase the value of the configuration parameter 'memory per worker process'."

may be reported in a server with a low number of worker processes and a high number of user

connections.

but the bit is no longer used. The internal upgrade process will reset it..

granted.

For example, "select col1 from tabA compute max(col1)" will fail if there

are no rows in tabA.

together with a stack trace which includes the modules 's_compile' and

'sequencer', may be reported in the error log when the user has a soft

binding with an user temporary database and both this user and the system

temporary databases are not available.

(enable unicode conversion is set to 0 or 1) is used, 4847 error is raised with

BCP IN command.

the remote statistics is not available, error message 11290 揇istribution

statistics for column '<column name>' in index '<index name>' is unavailable or

incomplete in the remote server.� shows incorrect <index name> against the

<column name>.

proxy text data results in Msg 605 error.

containing the modules 'sequencer','lwp_create' and 'sqt_deallocate' is written to the errorlog

using dynamic SQL against the server.

logged operation was that prevented a dump tran.

before you can continue." can be raised at the secondary site in a HA environment if the password has expired in the primary, and sp_password is run to change it.

The correct location for this file is $SYBASE/$SYBASE_OCS/config

name>' in index '<index name>' is unavailable or incomplete in the remote

server. UPDATE STATISTICS will not run in import mode." is displayed if UPDATE

STATISTICS on proxy table detects no entries in the remote systems statistics

catalogs. This warning message will no longer be displayed.

http://docs.hp.com/en/netsys.html

http://www.software.hp.com/portal/swdepot/displayProductInfo.do?productNumber=B8843CA

http://www.usenix.org/events/osdi2000/wiess2000/full_papers/subramanian/subramanian.pdf

concurrently with a CREATE TABLE command or other DDLs.

optimizer uses a "Positioning at index start" instead of "Positioning by key"

in a subquery.

different from the current database and no mode specification results in its mode

getting set to "Chained"; if the procedure exists.

NULL for unsupported events.

create config option--revert to pre 12.5 method of assigning spid numbers to

new connections

prior to version 12.5 ASE assigned spid numbers to new conncetions using a ..."lowest free spid #" method. Now with ASE 12.5 the method has changed to..."least recently used spid" method. The config option would give the customer the option to switch back to the pre 12.5 method of ..."lowest free spid #" method.

area

No more Monitor Server selection in Server Properties or display of Monitor

folder.

or produce incorrect result.

command will fail with Msg 911 and 11206.

modules 'sequencer','lwp_create' and 'sqt_deallocate' are found in the stack trace for the terminated connection in the errorlog.

continues to appear as such. This can have different effects, for example, that

DBCC CHECKTABLE continues to unnecessarily check for replication information.

is raised together with a stacktrace in the errorlog including the modules 'memalloc', 'q_oreop_fill'

and 'substitute' when a user executes a query with a very large IN-list in the WHERE-clause.

if a parallel select into fails due to lack of memory.

INSTRUMENTATION: ASE running on multiprocessor NT platforms with heavy use of ct_cancel

by client application may miss acknowledging one or more cancel operations leading to a

hang at the client end. Print diagnostics under traceflag 9999 to confirm the problem

was encountered and corrective action take by ASE.

NOTE: Fix for this problem is in ASE 12.5.3 ESD#2 under CR 377437

queries against DB2 backends.

Instead, the job is run when another stored procedure request is handled or the next scheduled job comes due.

This may happen if the Dynamic Index strategy is choosen by the optimizer

to execute the query.

error log for a query with two GROUP BY clauses; in the main query and in a subquery, when it is executed in parallel.

client session terminates. The connections can now be dropped and detached from an engine after

executing a statement to the remote server by enabling Traceflag 11231. Exceptions are when the

statement is participating in cursor, transaction or stored procedure operations; or when ASE

is in either HA failover or failback states.

a hard fault OAM entry is present but there is no allocation." may be

mistakenly reported by DBCC CHECKSTORAGE when it is run on a freshly

loaded database dump and the given database dump was taken on a busy

ASE environment initially.

raising the error message "You cannot use the text workspace

'<TEXT_WORKSPACE_NAME>' for database '<DBNAME>'; required minimum size

is <X> KB. The workspace is only <Y> KB"

executing stored procedure if the stored procedure is being recompiled and the

procedure executor does not own the procedure being executed.

due to number of file descriptor issues, even if the configuration file contains low values

for the number of user connections and the number of devices. ASE boot failure

reports in errorlog "Verification failed for parameter 'number of user connections'".

the system catalog locks when not in DDL-IN-TRAN mode after the commit

of the transaction but before post commit work started. This enhancement

is triggered by -T3710 either at boot time or via DBCC TRACEON(3710).

This performs basic page header sanity checks (HEADER) and both

page header checks and intra-page row checks (FULL) on every page

that is dumped by the Backup Server.

in the module 'ind__get_sysindrow' when DBCC LISTOAM is run with an

object ID value that is not a table. The stack trace in the error log

will also report the modules 'ind_getval' and 'pg_list_oam'.

msgrecv(), msgpublish(), or msgconsume()), a stacktrace (signal 11) may occur. The

stacktrace in the errorlog will contain um_destroy_conn().

stacktrace containing the modules 'lwp_create', 'tdssend_dynparams' and 's_crtproc' is

printed in the errorlog when sending a batch of dynamic SQL commands which in total have more than

255 dynamic parameters.

mechanism is used by ASE which results in failure.

2) The sp_ldapadmin check_login 'login_name' fails.

command fails with some diagnostic errors in the ASE error log file.

parameter may be observed on a heavily-used multi-engine Adaptive Server.

the commands CREATE INDEX on a partitioned table, DROP of a partitioned table,

and SELECT-INTO that is run in parallel, may cause an undetected deadlock

scenario between the concurrent tasks, requiring an ASE restart to unblock

the situation.

statement: ct_dynamic(CS_DESCRIBE_INPUT)

printed in the errorlog containing the modules 's_crtproc', 'lwp_create' and 'make_negative_obj_id'

when sending large amounts of dynamic SQL statements to the server without deallocating these

dynamic SQL statements.

or used_pgs(), they are unexpectedly also written in the ASE's error log.

wizard should be self explanatory.

by the "heap memory per used" configuration parameter is reduced by

the addition of a spinlock specifically to synchronize access to

the heap block pool.

'constree' together with a stack trace which includes the modules 'com_resdoms'

and 'com_query' may be reported in the error log when referencing a

column more than once and that column is defined as a subquery within a view.

back to a save point that follows a save point to which a

rollback has already been performed.

was when it was set backwards.

"Your query is blocked because it tried to write and database 'testdb' is in

quiesce state. Your query will proceed after the DBA performs QUIESCE DATABASE

RELEASE."

because it tried to write and database '<dbname>' is in quiesce state.

Your query will proceed after the DBA performs QUIESCE DATABASE

RELEASE." The behavior is activated using trace 833 in the runserver

file.

object with id ..." may be reported by a table scan on a DOL table when a slow BCP IN command has

been previously run parallel on this table. Only expected for a 64 bits server with more than one

engine or a 32 bits server which is started with -T712.

having a clustered index and if the conditions are such that this table

can hold 256 rows per page.

DATE: Fri Mar 18 12:47:23 EST 2005

TSE Login Name: marinak -- Marina Vilensky

CUSTOMER NAME: Federal Express

BASIC INFORMATION:

Cust requeted to provide a new feature:

Parallel Union All (high need)

==============================

Currently in ASE, a union-all query is processed serially with each result set

materialized in order. FEDEX would like the union-all to be able to execute in

parallel - effectively with the same result set processing that a parallel query

on a partitioned table in pre-15.x servers (in other words, results are set

immediately through a family synch vs. held until previous query is complete).

This would be controlled by %93max parallel degree%94/%94set parallel_degree%94 setting

as well as by the parallel query controls (another Feature Request addressed by

the subcase 11122438-11).

ADDITIONAL INFORMATION:

CR 362551 addresses a similar issue (Query with more than one union all is not

run in parallel). It is a defect CR and is fixed in ASE 15.0.

If the fix for the existing CR 362551 covers cust's request please associate

this subcase with CR 362551. Otherwise, please open a new feature request.

PLATFORM: Sun, HP.

Operating System: Solaris, HP-UX

Customer's ASE Release Version: 12.5.1 ESD#2 for 64-bit release.

disk i/o load can result in ASE client hang. In some situations it could also

lead to server shutdown.

errorlog containing the modules 'execproc', 's_compile', 'vu_viewcopy' when statement cache is

enabled and a statement containing a derived table expression is send to the server.

the following role(s) to execute this command/procedure: 'sa_role'".

Server may attempt to unslice just created table even if this table is not sliced.

While in majority of cases this action will not do any harm sometimes it may lead to errors

like 622 ("Opentable was passed a varno of 2. Object already has that session descriptor in use.",

216 ("Attempt to automatically drop temporary table failed. ") and 3702 ("Cannot drop the table

because it is currently in use. ")

ASE crashes and denied access with high numbers of concurrent Kerberos

authentication attempts from OpenClient applications.

When using MIT Kerberos client library version 1.3.1 or earlier

you can start ASE with traceflag 7844 that limits concurrent Kerberos authentications

to avoid segmentation violations. The traceflag is not needed with

CyberSafe or MIT verion 1.4 client libraries.

when a segmentation violation or a call to the C library routine

"abort()" has occurred within the 3rd party Kerberos

client libraries, results in error message:

"Internal error - Security Control Layer (SCL) is unavailable

on this engine and cannot support any external security mechanisms".

The error path also disables Kerberos operations on that engine.

The fix intercepts execution in the error path to release resources

and allows Kerberos operations to continue on that engine.

running." may apppear in the errorlog during a configured shared memory dump.

stored procedures during upgrade.

trigger, or SQL batch. Retry later, ..." may be reported for queries that need to allocate

memory chunks larger than 2 Kb (for example queries with long strings), while there is still

memory available in each engine's local memory cache.

currently in use." may be raised followed by error 216 "Attempt to

automatically drop temporary table failed." when sp_spaceused or OAM builtins

are used concurrently with sessions dropping their temporary objects.

TSE's ask for server level configuration from client, they always need to see data cache configuration as well. So they end up asking outputs of three commands sp_configure, sp_cacheconfig and sp_helpcache.

and remove "invalid" entries from this file without having to reboot

Historical Server, as in the case of inactive views, they can remove it

using hs_delete_data, but in the case of "invalid" views they cannot. They

want to clean up the hs.ctl file and keep it as small as possible with

little information from the past.

statement if the server configuration option "send doneinproc tokens" is set to 0.

CS_DESCRIBE_OUTPUT if the dynamic SQL executes multi-statement batch and SELECT

statement returning the data is not the first one in the batch.

results or fail with a stack trace on 64bit platform.

process (ase_monitor) dies, a parent ASE monitor process restarts ASE.

syntax errors or generate incorrect results if one of the strings used to built the command

string has a NULL/empty value.

may be reported on a heavily-used 32bit Linux system that is making use of large memory support

(the "extended cache size" configuration parameter is non-zero). Now, the wash size for the

<server page size> pool in all user-configured caches (not the extended cache),

is set to the configured wash size; prior to this change it was set to a fixed value.

This performs basic page header sanity checks (HEADER) and both

page header checks and intra-page row checks (FULL) on every page

loaded. If VERIFYONLY is used, the pages are read and checked from

the dump archive but they are not copied onto the database devices.

may encounter remapping problems when the statement is executed using the pass through mode

on a remote server and the local server has multiple temporary databases defined.

The problem will manifest with the message "current process ... infected with 11"

in module proc__reset_wktbl().

indid <value> in dbid <value>." may be raised during the recovery

phase of a database with a stack trace reported in the error log

with modules 'recovery', 'pg_zap_dealbit_db' and 'pg__recalc_oams'.

as being a replacement of sp_resumesub ... withskip option

2K pages and Large Memory Support(the "extended cache size" is set to a non-zero value).

with a stacktrace containing the modules 'mdarpc_exec' and 'exec__native_rpc'

may be written to the errorlog when ASE monitoring tables are enable and the

'number of open objects' parameter is not configured high enough.

in the HAVING clause may lead to different errors indicating query tree corruption.

containing the modules 's_compile_stmt', 'execproc' and 's_execute', when executing a stored procedure

which performs a select from a view having a UNION clause. This statement must be the second statement

in the stored procedure.

sybatch_ase.res

sybatch_bs.res

sybatch_ms.res

12.5 System Administration Guide

...to help with configuring 8 page I/O buffer pool for checkstorage

Chapter 25 Checking Database Consistency

...

Preparing to use dbcc checkstorage

...

Configuring Adaptive Server for dbcc checkstorage

...

Configuring an 8 page I/O buffer pool

The dbcc checkstorage buffer pool requirements for the

different page sizes in 12.5 are as follows:

(2k page server) * (8 extents) = 16k buffer pool

(4k page server) * (8 extents) = 32k buffer pool

(8k page server) * (8 extents) = 64k buffer pool

(16k page server) * (8 extents) = 128k buffer pool

<device name>: volume mounted out of order, expected volume 0001, got volume 16777216.",

is reported by Backup Server when running load database with a dump file that was generated

on little-endian platform.

in errorlog similar to 'File positioning failed (4).Seek offset: 828964864'

together with a stacktrace containing the modules 'd_upgradeobj', 's_normalize' and 'u_view_driver'

when executing the command DBCC UPGRADE_OBJECT() on a stored procedure containing a view having a

SELECT * on a table which is dropped and recreated before executing the actual DBCC command.

(7930 error:)

Table Corrupt: Keys in the left child page do not precede the parent key; check page %d, partition ID %d.

when the sysindexes system catalog and the database where it resides are bound to

different named caches, and when the clustered index of sysindexes is not explicitly

bound to a named cache.

is larger than the maximum size (<value> bytes) allowed for this table..." may be raised during

the execution of a stored procedure using the reformatting strategy and the creation of a

DOL (Datarows Only Locking) work table. Alternatively the error 9561 "An insert or update to table

'temp worktable' places column '(unknown column)' at offset <value>..." may be raised.

negative values.

the server's default character set is configured as UTF-8.

2) Running sp_encryption when there are no keys in the database produces no output.

3) sp_encryption enhancements to display keyname, tablename, column name and key details.

locally in ASE as proxy tables is not pushed entirely to the ORA-side although

this is completely possible

"Running installcommit script", likewise the installcommit script may report error 2714, "There is

already an object named 'spt_committab' in the database."

in error 11056 "Exceeded tempdb space limit of <value> pages." being raised.

This may happen due to some inaccurate accounting of tempdb space during

the execution of DROP TABLE, TRUNCATE TABLE and CREATE INDEX commands.

which includes the module 'ct_get_data' may be reported when accessing parsed XML documents

using a local proxy table mapped to a remote proxy table that has an external directory defined

on it.

single batch

may fail with Error Message 11291, Level 11, State 3: 揌istogram(step)

statistics for colidarray <hex colidarray> is unavailable or incomplete in the

remote server.� together with diagnostic error in ASE errorlog stating 換uery

in fetch_histogram() has error in fetched col = (varbinary column). Expect datum

length ...�.

Since Mac OS X 10.3, the Rendezvous (BonJour) API has been updated to allow domain-wide service registration and service discovery. To enable domain-wide service registration within ASE, first set an environment variable in the RUN_servername command file at $SYBASE/ASE-12_5/install. The environment variable is BONJOUR_DOMAIN. For example:

set BONJOUR_DOMAIN sybase.com. ; export BONJOUR_DOMAIN

This feature is also available for Backup Server and XP Server, and is triggered by the same environment variable.

It should be three times the size of logical page

during the insertion of rows with the BCP utility, if the target table had

its schema changed earlier to gain an IDENTITY column and one or more

non nullable column(s). This error would manifest only if the given columns

were added through a single ALTER TABLE command.

printed twice in the ASE errorlog for "model", "sybsystemdb", and "tempdb"

system databases.

This can result in failback operation failure with error

"Agent is unable to offline resource. Administrative intervention may be required."

ASE waits indefinitely for the server to complete all transactions on shutdown with wait.

Customer would like to be able to do checkpoint more often than 1 minute...they would like to checkpoint to run after so many SECONDS rather than MINUTES.

be reported in the error log when executing the 'ALTER TABLE' command for a

table having an 'auto identity' column and if the command needs a data copy.

This may happen if the session level 'quoted_identifier' option is on or if

the table name length is greater than 28 characters.

greater then 32 characters using [] to 'escape' special character _ for the LIKE clauses used

in sp_columns, will cause sp_columns to return no results for that tablename.

then the join estimations may not be accurate. This may affect

join orders in complex queries. In the query below, before the

fix, the row cound estimate of the result was 1892, but after the

correction, the estimate of 187 rows is much closer to the

actual join result of 184 rows.

select count(*) from t1, t2 where a1=b1 and a1<1893

go

-----------

184

==================== Lava Operator Tree ====================

Emit

(VA = 5)

r:1 er:1

cpu: 0

/

ScalarAgg

Count

(VA = 4)

r:1 er:1

cpu: 0

/

MergeJoin

Inner Join

(VA = 3)

r:184 er:187

/ \

Sort IndexScan

(VA = 1) i1

r:185 er:187 (VA = 2)

l:6 el:6 r:184 er:187

p:0 ep:0 l:2 el:4

cpu: 0 bufct: 24 p:0 ep:4

/

TableScan

t1

(VA = 0)

r:185 er:187

l:36 el:36

p:0 ep:36

============================================================

executable. The user must include the correct DLL path(s) in LD_LIBRARY_PATH or equivalent.

may be reported in the error log when a session tries to drop a dynamic

statement. The module 'lwp_drop' will also be reported in the stack trace.

object id of another recently created table, resulting in a 605 error when server is

configured for Large Memory Support (the "extended cache size" is set to a non-zero value).

size servers (4K to 16K) should be more than 256".

the process can now be terminated using the kill command.

to bypass the file system cache for file system database devices. This can result in

modified data pages not flushed to disk when ASE thinks the data is on disk. In cases

where there is a power failure or other system crash, these modifications may be lost

leading to data corruption.

1) dbcc checkalloc.

2) Multiple bcp's running against the same database.

if "disable disk mirroring" configuration parameter is 0.

with a stacktrace containing the modules 'vu_fixupview', 'vu_addviewqual' and 'pre_aggview' when

executing a SELECT containing a subquery in the select-list, a derived table expression in

the FROM clause and another subquery in the WHERE-clause.

Modify Utilities folder, add Wizard so user can add in external utilities

together with a stack trace which includes the modules 'des__systs_flushall' and

'des_checkpoint' may be reported in the error log during dump database when

the "number of open indexes" is under configured and Adaptive Server is configured to use

more than one engines.

You must be the System Administrator (SA) or the Database Owner (dbo) to execute

this procedure." will be incorrectly raised if the user is an alias to the DBO.

Related to the above:

Customer would also like a Warning message letting them know that he disabled the trigger on a Table. He thinks it's dangerous not to be reminded that a trigger was disabled.

minimum row length of <xx>." is printed in the errorlog together with a stacktrace containing

the modules 'insert_index_row', 'ncupdate' and 'obj_newrow' when executing CREATE TABLE <identifier>,

whereby the identifier is >30 bytes using a multibyte character set.

columns based on the timestamp datatype can lose the original selected values. New timestamp

values are inserted instead in the varbinary column.

then replication agent will send short ltl keywords of transaction log language

to the replication server.

compared to <column between constant and constant>

sample/esp/xp_echo.c does not work on IBM AIX 64bit platform.

member of the outer join and the view returns NULL constant.

"Migrate database data is not allow because migrate server data wasn't selected during the setup session." Should be 'allowed' instead of 'allow'

accurately calculated and claimed. This could result in infinite loops

in the cache manager where somewhere higher on the stack a Lava expensive

operator was attempting to grab a buffer.

traceflag 7844 is used, blocks all Kerberos activity on the engine.

A dataserver should be run with traceflag 7844 when MIT 1.3 Kerberos libraries are used

to workaround concurrency problems within the MIT Kerberos client libraries.

the recursion option may fail with error message Msg 11277 when a constant string

is concantenated to the filename.

when a query uses an aggregate function against a DATE column in a proxy table defined

against an ASIQ table.

ASE 12.5.3 ESD#1 or ASE 12.5.3 ESD#2 installed. The error at startup is:

Exception in thread "main"

java.lang.NoClassDefFoundError: com/sybase/util/ds/ldap/LDAPWarning

it finds a column encrypted by a given key in the same database.

It removes the sysencryptkeys system table in preparation for

downgrade without adequate checking.

if the view uses an encrypted column in the where clause.

if the given scan workspace is very large.

trace which includes the modules 'ss_chartoblob' and 'run' may be reported in the error log

during an UPDATE of a TEXT column using a J-connect program or a CT-library program that

uses host variables. In some cases a 510 error, "Bad eop type 0x.." may be reported instead.

note: if you use View > Redisplay or Zoom IN/OUT, the element symbol in the model diagram is refreshed and it's display in the diagram reverts back to the word wrapping setting.

stacktrace containing the modules 's_normalize' and 'sequencer' when executing a query

performing an INSERT into a temporary table based and on a SELECT with a UNION clause whereby the

last SELECT contains a subquery.

unichar or univarchar column.

a stored procedure, tsequal() no longer returns a timestamp value

unless the 4080 trace flag is being used. When an UPDATE that uses the

tsequal() builtin function is done outside a stored procedure, tsequal()

returns a timestamp value in the TDS_RETURNVALUE token.

is a Data Only Locked table and the SELECT contains a UNION ALL first, followed by one or more

UNION statements grouped together using ().

together with a stack trace which includes the modules 'rec_analyze_log' and

'rec_onlinetime' may be reported in the error log during ONLINE DATABASE on the HPUX 64 bits

platform.

not match <value> in Sysindexes" may be reported by DBCC CHECKCATALOG on a table that

has at least two bit columns, after ALTER TABLE was run to add/modify a variable

length or NULL column.

together with a stack trace which includes the modules 'lddb__unscr_stat_page'

may be reported in the error log, when loading a database across platform,

if the database has an orphan row in syssatistics.

when using asynchronous IO on file system devices.

This is liable to manifest as severe performance degradation on single CPU machines

and significant performance degradation on multi CPU machines under load.

This is due to a bug in Solaris kernel, Sun BugID 6302167.

Install the following patch from Sun Support for resolving this performance problem.

Solaris 10 SPARC: PatchID 120048-03

Solaris 10 x86/x64: PatchID 120049-03

are not correct. They should be 447 and 448 instead of 443 and 444.

Maximum allowed length is <maxlen>." will now be raised during garbage

collection of index leaf pages if the length of memory being moved

goes beyond a page boundary, instead of silently trashing the data cache.

'pg_oam_pagecounts' and 'pg_recalcoam' after loading a database dump which was taken

on a very active ASE environment.

column-names defined in a CREATE TABLE, CREATE VIEW statement and in the column-list of a derived

table expression are now checked against the rules for valid identifiers as stated in the standard

documentation.

This will print only 1 line summary of Deadlock giving only Deadlock ID.

syscomments for definition-time partition condition objects that were left behind

due to some errors that occured during creation of tables with partition conditions.

the Java type "double". This conversion will incorrectly

give an approximate value when the UNSIGNED BIGINT value

is large enough.

can run this command." will be incorrectly raised when a user without SA role different from the DBO executes DBCC CHECKVERIFY even if this access has been granted in the target database.

And the format of the backup config file will be xxx.000000001 instead of xxx.001.

(dbid: <dbid>, pageno: <pageno>) without releasing the <type> latch

it holds on the buffer." may be raised when an interrupt is received

by a query on a Data-Only-Locked table using the isolation level 3

(serializable) and performing a backward scan.

error log for a query referencing multiple times the same view and performing aggregates

on view columns. This problem will happen only if the query is executed in the parallel mode.

the install process.

incorrect values for the "Free" counters in the section

"Metadata Cache Management".

fragments which do not reference any segments.

1. For error 3019 replace severity level from 16 to 10 and also add info about severity level 18.

2. For error 3020 add info about severity level 18

3. Add information about error 3021 which is completely missing from documentation.

ddlgen -Usa [ no space after -U was allowed].

This did not match the standards of other applications like isql which used to accept spaces.

This fix solves this issue, and users can specify white spaces after the parameter types i.e.

ddlgen -U sa [now works!!!]

when executing a dynamic SQL statement which had to be

internally recompiled, for example, due to a schema change

for some of objects referenced by the statement.

locked table btree insert or delete log record is undone and if that page

is found to be corrupted, subsequent page fetches will fail with

Error 605: Severity: 21: State 1.

despite the session having send an interrupt request to ASE (Control-C).

This may happen if the session was waiting initially for the lock request

and ASE receives the attention request just before it awakes this session

in order to initiate a deadlock search.

DBT3_UNMOUNT_IN_PROGRESS and DBT3_DBSHUTDOWN_FROM_ACCESS

bits in dbtable are not cleared when we unkeep it.

So, the database is not accessible to anyone after this.

potentially breaking a client application. This problem may appear on all platforms

in 12.5.3 ESD#3 and the 12.5.3a GA releases.

may be raised when the configuration parameter "number of locks" is

under-configured.

if one or more databases fail to recover.

UNION ALL statement may fail to execute correctly. Stack trace may be logged in the

errorlog file. Following modules may be reported in the stack trace: execute(),

exec_eop(), run().

sp_helpartition

go

output to sp_helpartition.txt

go

This has been fixed.

=======

Delete: Can delete any logins except 'sa' and 'probe'.

Cut: Can not cut 'sa' and 'probe' and also those are bound by at least one user, with a SSO role is a must for this action.

For roles:

=======

Delete: Can not delete the following system roles, with a SSO role is a must for this action.

DTM_TM_ROLE

HA_ROLE

NAVIGATOR_ROLE

OPER_ROLE

REPLICATION_ROLE

SA_ROLE

SSO_ROLE

SYBASE_TS_ROLE

MON_ROLE

WEBSERVICES_ROLE

JS_ADMIN_ROLE

JS_CLIENT_ROLE

JS_USER_ROLE

MESSAGING_ROLE

Cut: Same condition as Delete needs.

a query under trace flag 289 and the query references a view with union and is

using convert function on TEXT data.

may be reported when executing a SQL INSERT..SELECT from proxy table mapped to a RPC. This

is resolved with a server-wide traceflag 11232 or a session level traceflag 11228. Either

traceflag has to be turned on to allow CREATE TABLE command in the remote procedure.

each row in the result set. The original builtin rand([integer]) was

only evaluated once for the result set.

This prevents further authentication requests from succeeding on any engine where

this occurs.

Messages such as "Waiting for native threads to be released, or for 'max number of native threads' to be increased." appear repeatedly in the errorlog after

a number of login requests become blocked. When this message appears in the errorlog, other use of native threads are also prevented.

This is extended to report the number of rows per partitions in addition

to the number data pages.

the remote procedure if it does not execute any SQL queries on the tables.

reported when using "batch updates" (sending several dynamic

sql statement executions in the same batch) and the statement

that was already prepared had to be recompiled because, for example,

the underlying table schema changed.

module name such as "terminate_process(), close_network() ... jobscheduler()" may be seen in the errorlog.

installation had system catalogs bound to named caches and if the caches are