Sybase数据库技术，数据库恢复专家

in the parallel and the query has a scalar aggregate. This problem may happen if

the showplan output shows multiple query steps with "Parallel result buffer merge"

execution strategy.

The query plans generated in the invoking session may thus be based on the previous 'optimization timeout limit' value instead of the newly configured value.

in wrong selectivity estimation.

ASE will report the following error when trying to use Unicode

characters in a table or column name:

Msg 102, Level 15, State 1:

Server 'ase_125', Line 2:

Incorrect syntax near '琴'.

all threads running a parallel query plan. The new command is:

stack family <alfa_spid>

Note, that alfa_spid must be the spid of the parallel plan's alfa thread.

into a T/SQL local variable or parameter (@variable), Adaptive Server

may return Msg 15445: "The datatype of the ciphertext data (NUMERIC)

is incorrect for column ''."

together with a stacktrace containing the modules 's_getTreeOrPlan', 'proc_setdb' and

'proc__setdb_tree' when a SELECT statement containing a derived table expression is executed.

ASE must be configured with user defined tempdbs and statement cache on.

A new message will appear in showplan output when dynamic partitioning is active:

[ Using Dynamic Partition Elimination ]

when Adaptive Server encounters some problems while distributing the work among worker processes

and ends up without proper clean-up of the worker processes in some rare cases.

is not handled correctly, and an index is unnecessarily rebuilt

by sp_post_xpload, if variable length index key was greater than

252 bytes. Also some page corruptions happen in DOL index root page,

in DOL non-leaf index page for UNICHAR NOT NULL index key in some cases,

and in a special case of slice control page if the database dump is from 11.9.x.

This may occur if the select list of the first SELECT contains duplicate

column references.

At the server level:

sp_configure "optimization timeout limit", <0..1000>

- this option is now valid for ad-hoc queries only

- the range has changed from <1..1000> to <0..1000>

- the value 0 represents "no timeout"

- the default is still 10

sp_configure "sproc optimize timeout limit", <0..4000>

- this option has been introduced for stored procedure queries

- the is from <0..4000>

- the value 0 represents "no timeout" at server level

- the default is 40

At the session level

set plan opttimeout <0..4000>

- this can be specified for both ad-hoc queries or stored procedures

- the range has changed from <1..1000> to <0..4000>

- the value 0 represents "no-timeout" at session level

- for ad-hoc queries, values specified above 1000 are discarded and default of 10 is used

At query level

select * from TabA ...

plan

"(use opttimeoutlimit <0..4000>)"

- only <0..1000> can be specified for ad-hoc queries

- <0..4000> can be specified for stored procedures

- a value of "0" represents no-timeout at the query level

followed by a stack trace may be reported in the error log when a session

is detaching from an external transaction and the revert to the database

that was active at the time of the attach has failed. The modules

'xact__detach_external' and 'closedb' will also be reported. As a result

an implicit shut down of ASE will occur.

partition condition and the specified index keys make it impossible

to enforce index uniqueness across partitions." will be reported

only when the user tries to create a unique local index; the issue

doesn't apply to a unique global index.

table based on a UNION ALL or a view containing a UNION ALL.

e.g. select * from T where T.a IN (1,2) and T.b IN (5,6) and ... some other predicates

suppose an index existed on T(a,b) and all rows of the index qualified with the above predicate, then a full index scan with lookup of all data rows will occur, which would normally be much slower than a simple table scan or some other index scan with the remaining predicates.

The rows to be inserted contain at least one LOB column (TEXT, IMAGE or UNITEXT);

The SELECT that generates the values to be inserted must access at least two distinct tables, each with a LOB column;

The query plan must use a HashDistinct operation, i.e. the showplan output contains:

HASH DISTINCT Operator.

The 14108 error can occur if at least one table in the query has no LOB columns. If all tables in the query have LOB columns, the 605 error can occur.

that has replicated stored procedures can fail on this error:

Error: 632, Severity: 20, State: 2

Attempt to move memory with an incorrect length of -166. Maximum

allowed length is 16384.

if server is shutdown and boot again, accessing semantically

partitioned table or roundrobin partitioned table with global

clustered/placement index may run into segv.

a hash distinct or a hash union operator in its plan.

This stacktrace contains the functions : LeHashOp::_LeClose() and soextdealloc().

and causes Job Scheduler not to function on Mac. This affects only Mac OS platform.

ASE reports in errorlog messages like "Incorrect Job Scheduler sequencer version found

within 'sybmgmtdb'", "Job Scheduler can not operate, Expected 0:27:27, Actual 0:26:26",

"Job Scheduler failed installing its required sequencer code",

"Job Scheduler Task failed to initialize", "Job Scheduler Task state set to stop".

and causes Job Scheduler not to function on Mac. This affects only Mac OS platform.

ASE reports in errorlog messages like "Incorrect Job Scheduler sequencer version found

within 'sybmgmtdb'", "Job Scheduler can not operate, Expected 0:27:27, Actual 0:26:26",

"Job Scheduler failed installing its required sequencer code",

"Job Scheduler Task failed to initialize", "Job Scheduler Task state set to stop".

severity 19. This severity is too high and should be 16.

session level.

...other scripts...

installmaster (instmstr on Windows) Script

installcommit (instcomm on Windows) Script

installsecurity (instsecu on Windows) Script

instmsgs.ebf Script

inst_ha_scripts.sh Script (Sun Solaris only)

installhasvss (insthasv on Windows) Script

installdbccdb (instdbcdb on Windows) Script

...after all the files for the EBF/ESD are placed on the hard drive.

hang.

a deferred update causes the cursor position to be lost, a signal 11 and stack

trace including LeDMLOp::_LeOpClose() may occur.

server tasks being blocked by a detached distributed transaction.

<PartitionNumber> for object ID <ObjectId> in database <DatabaseName> may not have

been done correctly ..." may wrongly be reported under the previous object section for DBCC checkalloc.

instructions for removing the "sybase" account from Windows system-defined group "Administrators"

after performing the steps needed to improve security is missing. This can leave the

"sybase" account with a higher privilege than needed.

The new XML has information that tries to bridge more gap between the text version of showplan and the xml version.

For all Objects/Nodes the following Important attributes are added

1) Abstract Plan

2) Overall Cost

3) Resource

4) Optimizer Metrics

5) Optimizer Statistics [Including object for which stats is missing]

is raised when the UPDATE PARTITION STATISTICS command is executed. The partition number is

omitted but instead the WITH SAMPLING <number> clause is used.

module 'senderror' may be reported in the error log together with a stack

trace when executing a prepared dynamic statement.

in dbid <value>." followed by the message "ASE is terminating this process." may be

raised when using datalength() built-in function on TEXT/IMAGE columns within a TRIGGER.

The stack trace being reported in the error log will show modules 'LeRun', 'textsize',

'gettxtpage' and 'open_partition'.

yet. Added an error message 11935 in rbld_main() to warn

the user.

while using Posix AIO. This can lead to ASE hang. Solution is to start ASE with trace flag 1630.

uses this same charset, then ASE may return the wrong number of characters

for the given column descriptor when a built-in function such as left()

is executed.

Chapter 2 Understanding Component Integration Services

Proxy tables

Using the create proxy_table command

...

This section should also include the

statement...create proxy_table...updates

system tables sysobjects, syscolumns,

and sysindexes

...

...making it similar to the previous section on create existing table.

fatal error 539 may be seen when out of log space.

on Veritas Cluster Server.

with text or image data using UPDATE or INSERT-SELECT command from a proxy XFS table

results in 605 or 806 error.

The syntax will be: sp_metrics 'filter', '@gid', '@predicate'

e.g. sp_metrics 'filter', '1', 'lio_max < 100'

which will delete all qpmetrics records in group 1 where

lio_max < 100.

value for a given user-defined cache, if it was created with a long name

initially.

cannot be acquired and is awarded on grace, a new error

message with severity 17, non fatal but requiring action, is printed to the

dataserver's errorlog. This error message is to inform DBA that license

will expire at a stated date.

Monitor counters are:

total number of parallel queries

number of run-time-adjusted queries due to thread limit ('max parallel degree' or

session 'parallel_degree' setting.)

number of run-time-adjusted queries due to a lack of available processes.

transaction, it results in partial commits on remote pre-DTM servers.

executed with subquery cache statistics enabled. The stack trace will

include printSqCacheStats.

the INSERT or UPDATE statement is executed inside a stored

procedure and a change in the underlying tables used by the

stored procedure caused more than one internal renormalizations.

table 'table_name'; column does not allow nulls" may be raised when executing

an INSERT statement inside a stored procedure and the option STRING_RTRUNCATION

is ON and the value to be inserted only contains blanks.

dbccs. The new syntax is shown below:

dbcc tablealloc(tablename | tabid

[, full | optimized | fast

[, fix | nofix

[, data_partition_name | data_partition_id ]]])

dbcc indexalloc(tablename | tabid, indid

[, optimized | fast

[, fix | nofix

[, partition_name | partition_id ]]])

dbcc rebuild_text(table | "all"

[, column

[, text_page

[, data_partition ]]])

undetected deadlock due to missing wakeup in a rare window.

has been set some unexpected errors may be raised and the procedure will fail.

not consistant with documentation.

maybe reported in the error log during ONLINE DATABASE, after loading the database

that is dumped by a server using a different charset id than the current server.

A table created in pre-15.0 system and then altered into range-partitioned table with

char/varchar partition key column(s) may be marked as SUSPECT and thus read-only during

sort order/char set change that normally wouldn't make the range partition suspect.

incorrect results.

1. Configurable timeout mechanism for

binds and searches to the LDAP server.

2. Configurable number of LDAPUA

native threads per engine.

3. LDAPUA error reporting improvements.

"current process infected with 11" in the module 'crc_ckconstr'

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

and 'crt_constr' may be reported during the creation of a table.

Similarly, the message "current process infected with 11" in the

module 'procrm' together with a stack trace which includes the modules

's_recompile' and 's_rmpbuf' may be reported during the execution

of a stored procedure.

of the union's first select statement and are the source for an insert

statement may return wrong result or stacktrace. The following union

provides an example:

insert foo (c1, c2, c3)

select c1, c2, c3 = null from t1

union

select c1, c2, c3 = c1 from t2

The problem may not occur depending on how columns are aligned in the

underlying data row.

http://sybooks.sybase.com/onlinebooks/group-as/asp1251e/uconfig/@Generic__BookTextView/1885;pt=2093;nh=1?DwebQuery=user+objects#X

(1) Under Table Specifications, "User Objects per database" has the maximum value of

"231 - 100". This is incorrect and the expression should be presented as

2 to the power 31 minus 100.

(2) Similarly, the following entries are incorrectly presented:

Under Adaptive Server Specifications, Database Specifications:

(A) Maximum database size

" 243(8TB) Minus overhead for system databases"

The above maximum size should read 2 to the power 43 not 243.

(B) Maximum size of a database device (disk partition)

" 224(32GB) If the Operating System supports file sizes up to 32GB, then Adaptive Server supports file system devices up to 32GB"

The above maximum size should read 2 to the power 24 not 224.

large private buffer pools. These pools could require a large

chunk of contiguous memory which could lead to 701 errors even

though there is ample free space in the procedure cache. A 701

error can be traced to this problem by rerunning problem query

with traceflag 751 enabled. If the resulting stacktrace contains

sort_auxbufinstall(), then this is the problem.

errorlog.

This can happen on windows 2003 or windows xp systems where system32\libeay32.dll is an older

version than the libeay32.dll released with tibco.

TIBCO recommends to copy libeay32.dll from TIBCO\clients\c\bin to system32.

query where some of the ORDERBY columns is not part of the INSERT

scalars. For example:

insert into t1 (col1) select col1 from t2 order by col1, col2 desc

In the above query, col2's ordering requirement may not be enforced.

This is fixed to show the right values

CR 406159 : Wrong Device Size in creation

and the current CR, i.e duplicate context menu items for a db created for load

These are fixed

was found." may be raised during LOAD TRAN or LOAD DATABASE, when the

configuration parameter "per object statistics active" is turned on.

to a proxy table can result in errors (for example, "Failed to seek to offset -NNNNNNNNNN").

cross reference CR 406116...DC/dcany/MSSQL...add sqlxformation to convert...Sybase

noholdlock...to Microsoft with (nolock)

When ASE Plug-in that is coded with JRE 1.4 is run with 1.3 JVM, it gives stack trace during

Create Table which results in a stack trace that says java.lang.NoSuchMethodError.

When ASE Plug-in that is coded with JRE 1.3 is run with 1.4 JVM,

focus error will be seen during Create Table.

"count_big",

"materialized",

"tracefile",

"semi_sensitive",

"insensitive",

"scroll".

SCL/Kerberos memory leaks on successful connections when using MIT Kerberos, CR 385928. After many hundreds or thousands of successful Kerberos connections, memory leak resulted in memory allocation errors in the errorlog and authentication errors returned to clients. The memory allocation errors reported: "Security Control Layer's memory allocation routine returned an error."

An API change occurred to OCS internal functions used by Security Control Layer (SCL) drivers called by ASE. The request ID changed to type long to handle 64bit request ID used by SCL and DCL drivers. CR 384600, the associated OCS fix, avoids a segmentation violation that can occur on 64 bit platforms when using Kerberos or Directory Services lookups.

treat block device the same as raw device. ASE assumes all writes

have been committed to disk, while the modified pages may be retained

in the OS kernel cache. In the event of a power failure or other

system crashes, the cached modifications may be lost leading to

data inconsistency.

Using a shutdown with nowait by itself would not cause an inconsistency.

Recovery would of course take more time during the next startup if we

do a 'shutdown with nowait' though.

There is also a possibility that if theres a system crash or a power

outage during shutdown, it could lead to a data loss. However, note

that a system crash has to occur for us to encounter

this sort of situation.

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

'exec_dbcc' and 'memusage' when a DBCC MEMUSAGE command has been previously run

against a busy ASE system resulting in ASE shutting down itself.

XML Schema def 1.0

In the XML model, the customer has used the Mapping Editor to create a mapping to the PDM.

In the Mapping Editor, they have added a user defined query.

When the generate the SQL\XML query, the where clause is not included.

Instead of the text message explaining the error that occured, the following message appears:

01:00000:00049:2005/10/12 15:59:25.84 kernel Failed to get the Connectivity Library (Security Control Layer) error message.

following occur due to attempts to exceed the limit of "user connection" resources, then a memory leak occurs.

server There are not enough 'user connections' 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.

server Error: 1601, Severity: 17, State: 7

After memory in the pool used by these authentication mechanisms is exhausted, memory allocation failures and segmentation violations occur with continued authentication attempts.

Memory allocation failures appear with message:

server Security Control Layer's memory allocation routine returned an error.

Segmentation violations due to attempts to use invalid addresses appear with message:

kernel current process (0x11200025) infected with 11

kernel Address 0xff130988 (memset+0x74), siginfo (code, address) = (1, 0x00000000)

These messages may be accompanied with or without stacktraces in the errorlog.

the configuration parameters 'enable surrogate processing' and

'enable unicode normalization' will not have the configuration

and run values set to the default value. Please see the system

administration guide for descriptions of the above mentioned

configuration parameters.

The work-around is to modify the configuration parameter values

in the configuration file and reboot Adaptive Server.

the job scheduler may fail to start and the message "Job Scheduler Task failed

to spawn an Agent" is reported in the ASE errorlog.

may encounter 701 errors. These queries can request large chunks of

contiguous memory in the sort manager for merge run processing.

The size of the requested memory is a function of the number sort

buffers. In previous releases, this was bounded to 500 by default.

The fix is to once again to bound the number of merge run buffers (how

many runs can be merged at a time) to 500.

This problem can be diagnosed from a memdump. The stack of the spid

encountering the 701 error will contain alloc_merge() in it.

fail to execute correctly raising error 2805 "Bad pointer 0x3294d04 encountered while

remapping stored procedure" or the execution may be broken and stack trace logged

in the errorlog file. Module "proc__mkworkrg" should be referenced in the backtrace.

trigger name on a table.

- CR401397

Previously DBISQL does not refresh database dropdown list when "create database" statement is inside a if...else statement. This has been fixed.

- CR404078

Previously in 15.0 GA, DBISQL launch script "dbisql.bat" (Windows) and "dbisql" (Unix platforms) contained invalid argument "-path" that caused the script to fail. This has been fixed. (The correct argument is "-Dpath".)'

- CR399694

DBISQL failed to import Excel 2.1 worksheet. This has ben fixed.

- CR400053

Previously Make Permanent button of Options dialog does not save preferences permanently. This has been fixed in build 3204.

- CR400479

Previously DBISQL command "output to" does not work for saving a single result set from a stored procedure, such as:

sp_helpartition

go

output to sp_helpartition.txt

go

This has been fixed.

- CR402435

DBISQL cannot be launched using the script file on Unix($SYBASE/DBISQL/bin/dbisql) and Windows (%SYBASE%\\DBISQL\\bin\\dbisql.bat).'

- CR386931

Previously DBISQL does not display query result similar to the that of command-line isql. This has been addressed by adding a "Text" output mode, in addition to existing "Grid" output mode. To turn on Text output, open Options dialog by selecting menu item "Tools"->"Options...". On Options dialog, select "Results" from the left-hand-side list and select "Text" radio button to enable text output.

Procedure cache memory request start with chunks of 2k. Large chunks, upto 16k,

are allocated based on the past allocation behavior for the query. The enhancement

reduces external fragmentation in the procedure cache.

in the log and the associated spid may keep looping , waiting for free buffers.

The spid cannot be killed.

with large estimated hash table sets (> 5.8 Gbytes).

pages. In rare cases, the unconverted page contents may mislead recovery

at some time in the future.

AND col2 = <value>, with an index in col1, col2 and a high number of rows with

col1 is NULL.

cross database queries on tables with columns based on user defined

datatypes.

and contain distinct clauses, unions, or group by clauses may

generate a 702 error if a hash operator is used.

intersection anymore together with an OR clause index union when

SET INDEX_INTERSCET is OFF.

Also, when index intersection properties (as partitioning) don't

match, abandon gracefully instead of continuing up to a failure.

are not padded at the end.

was marked for replication using sp_setrepcol.

resides in a temporary database whose ID is different from the current

session's temporary database ID.

declare @v

select @v = 100

select @v = @v + id from sysobjects where id <= 3

returns 106 instead of 103, because every value of sysobjects.id is added to

@v instead of just the last one. The expression, @v = @v + id, should only be evaluated

for the last qualifying row.

specified with the same column name in ORDER BY clause

and message 2601, "Attempt to insert duplicate key row in <object_name> with unique

index" may be reported when executing a CREATE DATABASE ... FOR PROXY_UPDATE

against MSSQL backend.

occur in LeOrScanContext::setOrVtuple().

updated, if a nonclustered index is safe (index columns are not

updated) and if the second nonclustered index has one row in the

root page, the update may fail with error 2601 if the index is

safe index is unique. If the index is not unique, duplicate

index row will be inserted.

alter table partition by... on table with local clustered

index and the number of partition >= 29 on 32 bit server

may run into error 632 .

table of an outer join, extra incorrect row(s) are returned. This is due

to some plan for an outerjoin query could miss a predicate on a column of

an outerjoin inner table.

expression in the select-list as well as an expression in the WHERE-clause can cause a

query plan to be generated that contains a RestrictOp as the parent of another RestrictOp:

One RestrictOp will evaluate the expression in the select-lis while the other will

evaluate the expression in the WHERE-clause. Although the query results are correct,

there can be a slight performance impact of having two RestrictOps when a single

RestrictOp could do the same processing.

1> sp_audit "table_access", 'all', 'all', 'on'

2> go

Msg 18226, Level 16, State 1:

Server 'EQTGDBLNUA_DS', Procedure 'sp_audit', Line 305:

'all' is not a valid user name.

(return status = 1)

hash table for table <internal LWP name = (* + number + alfanumeric suffix) >,

objid = <negative number>. Descriptor hashed by name = 0x0 and hashed

by id = <hexadecimal number>." together with a stack trace which includes

the modules 'lwp_create' and 'des_lwpdescreate' maybe be reported in the

errorlog if a client application connection has executed a high number

of dynamic sql statements without disconnecting from Adaptive Server.

In the same conditions and if the same connection is holding a cursor open

which uses internally a worktable then the message "current process infected

with 11" in the module 'proc__install' may be reported in the error log

and it could bring Adaptive Server down.

sometimes be generated.

by clause contains expressions. This has been fixed.

buffer pools, 644, 2628, 2511 or other similar errors may be

reported after an update to a non-leaf index page table is lost.

a proxy table containing clusterd index in an Allpages lock scheme, an erroneous

entry in SYSTABSTATS with column id 0 (object id) will be created and the ASE internal

housekeeper task may monopolize the CPU utilization to 100%. During ASE shutdown,

an error message 6008, "Unable to write statistics for object <objid> in database

<dbid>. Please run update statistics on this table after boot-up or prior to

accessing the table next time" will be reported in the errorlog. Once ASE that

contains this fix has been started/rebooted, the entry in SYSTABSTATS with id = 0

should also be manually deleted.

UPDATE statements using the TSEQUAL() function may fail

and the client server connection gets broken with TDS

protocol errors.

This is due to the extra result set sent by tsequal()

builtin which may not be correctly handling by

DBLibrary API.

(i.e. not all pages in the database are present in the database dump), the pages

that were not overlaid by the backupserver may present a recovery issue later.

An initialized page with a timestamp greater than the highest timestamp in

the database when allocated may cause recovery to skip this allocation if ASE

crashed before this new page gets flushed to disk. After recovery, the object

to which this page was allocated may become inconsistent.

not reporting 946 error "A user interrupt has prevented Sysusages changes

for database <name> from being copied into memory. Please run sp_dbremap

to complete database space changes.". The side effect of this is the

inability to subsequently run ALTER DATABASE with success, 2601 error

"Attempt to insert duplicate key row in object 'sysusages' with unique

index 'sysusages'." is reported, until such a time ASE is restarted.

or a password is changed.

prohibits creation of multiple instances of the table.

User will see error 2714 with the message "There is already an object named 'constraint name' in

the database." when attempting to create a temporary table.

conditions even when memory is available. The failure to allocate

memory may manifest as failure to allocate large network memory pools.

executing an INSERT or UPDATE query to a proxy table in a remote ASE.

database transaction log if the SELECT command is audited on a table and the

transaction mode is 'chained'.

to return to the sybmon prompt rather than dbcclog prompt. Then all the loaded transactions need to be loaded again.

Msg 20200, Level 15, State 0:

Server 'textsvr':

Error E0-0720 (I/O Filter): Could not load filter 'flt_kv -recognize' which is

named in your style.uni file.

"Order-by items must appear in the select-list if the statement contains set operators."

if in the parent query ORDER BY clause is used on the view column which is not present

in the parent's query SELECT list.

cross platform dump load may be treated as regular table

after clustered index is created on the table before the

suspect partition is fixed.

you may get the wrong result from selecting the non-materialized computed column.

If the "allow nulls by default" dboption is on for tempdb, the installmaster script will raise error 2718 'can't specify Null values on a column of type BIT' in multiple places.

Fix: bit columns were defined as null

120,and 454 are the number of rows in each partition respectively) will result

in Error: 691. Severity: 20. State: 1.

message indicating the correct range of possible values for a given

configuration parameter.

parameter <@parameter_name>, which was not supplied" maybe

be reported by an Open Server application when Adaptive Server

is executing the procedure as a Remote Procedure Call (RPC)

requested by a client application and the value for the

parameter was sent by the application as a NULL value.

ASE does not eliminate duplication from the result set.

without data copying, reorg rebuild on the table or "alter table

partition by..." on the table lead to column offset corruption.

using BEGIN TRANSACTION @variable whereby the @variable contains a transactionname greather then

67 bytes. Maximum allowed for a transactionname is 30.

A column can not use both 慹ncrypt� and 慶ompute� options

These changes will be seen in Column Properties and Table Editor.

when the inserted value is an expression.

"The table 'table_name' in database 'db_name' is being created

and the transaction has not been committed yet. Please retry the query later."

may be raised if the parallel query is executed on a table created

in the same transaction.

ORDER BY clause can generate wrong results. Only the rows from the first UNION ALL side in the

derived table expression will be inserted.

target tables are in different databases could show <x> rows affected when the UPDATE

statement is finished. The affected rows however do not show any value being changed.

Grant null on dbo.titles to guest

go

Grant null on dbo.titles to guest

go

Below is the output of @@version for several SQL Servers versions:

SQL Version @@version output

1) 11.0.3 SQL Server/11.0.3.3/P/Sun_svr4/OS 5.4/SWR 9122 ESD 7/OPT/Sun Jun 18 19:50:32 PDT 2000

2) 11.9.2 Adaptive Server Enterprise/11.9.2.6/1294/P/SWR 10276 ONE-OFF/Sun_svr4/OS 5.5.1/FBO/Fri May 10 02:09:30 2002

3) 12.5.3(32-bit, Linux) Adaptive Server Enterprise/12.5.3/EBF 12151/P/Linux Intel/Enterprise Linux/ase125x/1883/32-bit/OPT/Thu Nov 11 17:33:16 2004

4) 12.5.3(64-bit, Linux) Adaptive Server Enterprise/12.5.3/EBF 12897 ESD#3/P/x86_64/Enterprise Linux/ase1253/1916/64-bit/OPT/Wed Aug 3 18:14:45 2005

Customer is requesting that we change sysmon to either include both synchronous and asynch completes in the Completed output, or that we break them out into a section for synchronous and a section or asynch I/Os so we get a more accurate read on disk activity.

Customer is running Adaptive Server Enterprise/12.5.3/EBF 12417 ONE-OFF ESD#1/P/HP9000-879/HP-UX

11.0/ase1253/1899/64-bit/FBO/Tue Feb 22 19:21:47 2005 .

This problem is not version- or platform-specific, as it is written in the source for sp_sysmon_diskio, again as documented in CR 144252.

two or more different tables and that has an ORDER BY clause can return incorrect results for the LOB values.

An example query is:

SELECT c1, CASE when c1 = 1 then t1.c2 else t2.d2 end from t1, t2 ORDER BY c1

where t1.c2 and t2.d2 are LOB columns and c1 is a non-LOB column.

against a pre-15.0 installation and when sysdatabases has more than 1 data page

and if the master database's sysdatabases row is not on the first page.

The master database's sysdatabases row may not be on the first page if there

are other databases with the name less than 'master' in the sorted name order.

conditions even when memory is available. The failure to allocate

memory may manifest as failure to allocate large network memory pools.

fail to report referential integrity error. Instead of this a client

session may get killed and the message reporting TDS packet synchronization

problem will be displayed.

to execute.

VARCHAR column if the source of the data is a table with longer VARCHAR column.

This is usually caused when a non-TDS client connects to the port that ASE is using as its master listener,

and sends data that is not in valid TDS format. An additional message will now accompany this message

providing the IP address of the client in question, so that it can be tracked down.

In addition, traceflag 7845 can be used to suppress these messages if they cause a nuisance.

select * from t1 left join v1 on t1.a = v1.a

and the view is something like

create view v1 as select char_col + "au", ... /* character expression */

create view v1 as select 1, ... /* constant value or constant expression */

is incorrect on the following page ..." may be reported on Sysdevices when during the execution

of a DISK MIRROR command a buffer pool whose size is greater than 2K bytes is used by ASE.

configurable shared memory dump (CSMD) is being generated. This problem does not apply to ASE running on the Windows platform.

to loss of data rows. This problem will only occur if the table was partitioned in the

pre-15.0 server, and was then unpartitioned during the upgrade process.

may fail with error 208 "Object not found" referencing the temporary table if

the system is configured to use multiple temporaray databases.

unusually large security token is given by the client to ASE during authentication.

understood by Adaptive Server version 15.0, during upgrade to 15.0 from

prev-15.0 installation.

r.a <relop> s.a where <relop> is one of > >= < <= !=

were not being costed in most cases (except when an OR predicate was involved) so that plans that could be benefit from these predicates were given no preference over other competing plans. This could result in sub-optimial plan selection.

Also, a new theta join is now recognized

(r.a = s.a OR (r.a IS NULL and s.a IS NULL))

is really a join in which NULL = NULL, and is now recogized as such for nested loop join operators only.

in the errorlog together with a stacktrace which includes one

or more of the following modules: 'proc__free', 'procrm' or

'proc__grab' if many processes were concurrently executing

tasks that requires some quantity of memory and to satisfy

those requests some unused procedures had to be removed from

Procedure Cache for reusing their memory.

"Database '<DBNAME>' is unavailable. It is undergoing LOAD TRANSACTION"

followed by the message, "Forced flush of database id <DBID> failed".

This may happen under very rare circumstances where the LOAD TRAN command is

used many successive times to recover a huge set of log records representing

around 1 billion page changes.

Each node of the tree now displays information as to which partition is in use Vs which ones are eliminated

stack traces in LeVtuple::LeVtGetColumn().

maybe reported in the errorlog together with a stack trace which

includes the modules 'ns_handle' and 'ns_prep_stmnt' if a batch

including together a CREATE TABLE statement and a SELECT executing

java was being executed and an error condition happened.

several allocated pages may fail with an error 3935 reading "Fatal protocol

error. xact_beginupdate() API was invoked in the wrong context. The transaction

is currently in 'Command-attached' state."

is also used as the source table in one of the union sides.

if CS_RES_NOSTRIPBLANKS is turned ON.

be enforced. For example, a foreign key constraint on three

columns is missing the check on the second column.

the nocase sortorder, ASCII and accented characters are sorted in case-insensitive

manner, CJK characters are sorted in Unicode binary order. Please run sp_helpsort

to get the sort order id.

a stack trace which includes the module 'ksmasknserver' may be reported in the error log.

This stack trace may be repeated a few times followed by ASE listener failure with a

stack trace accompanied by the messages in error log such as

'kernel nopen: accept, Software caused connection abort',

'kernel nopen: accept, Invalid argument',

'kernel Listener with protocol tcp, host <host-name>, port <port-number>,

engine 0 is quitting after a fatal error.'

metadata cache to obtain the current last page of a data partition.

cache, and the looping SPID can not be killed.

's_reinstall_lwp_tree', 's_getTreeOrPlan' and 'exec_proc' is written to the errorlog when

executing a stored procedure with a name which starts with '*' using the execute immediate

feature.

'ubo_object_from_slot:Sybproc Invalid slot id 0' followed by

'kernel current process (0xb000b0) infected with 11'.

"set statistics plancost" or traceflag 9500 for performance

analysis. For large input sets, the number is approximately

50% more than it should be.

So, during runtime rollback, the pages are being fetched into the default data cache.

This is leading to the condition that a page is in two caches at the same time.

So, further attempts to fetch the page throws 693 error.

jConnect: System stored procedures like sp_tables and sp_stored_procedures do not allocate sufficient space for the input paramater @table_name. [CR 406276]

jConnect: DatabaseMetaData.geProcedures() does not return procedures owned by non-sa_role users if the current login holds sa_role prileges. [CR 394097]

jConnect: Request to support Savepoint feature as described in the JDBC 3.0 specifications. [CR 350092]

ODBC Driver by Sybase: Installing the Driver Meta Data SQL scripts will leave the Adaptive Server with 'allow updates' enabled. [CR 406722]

ODBC Driver by Sybase: System stored procedures like sp_tables and sp_stored_procedures do not allocate sufficient space for the input paramater @table_name. [CR 406276]

OLE DB Provider by Sybase: Installing the Driver Meta Data SQL scripts will leave the Adaptive Server with 'allow updates' enabled. [CR 406722]

OLE DB Provider by Sybase: System stored procedures like sp_tables and sp_stored_procedures do not allocate sufficient space for the input paramater @table_name. [CR 406276]

Configurable Shared Memory Dump (CSMD) completes, causing an inability to connect to ASE.

This issue does not apply to ASE running on the Windows platform.

in strings constants over 255 bytes, stack trace might occur in

LeRun() -> concat().

with "SELECT DISTINCT"

in the errorlog together with a stacktrace which includes one

or more of the following modules: 'proc__free', 'procrm' or

'proc__grab' if many processes were concurrently executing

tasks that requires some quantity of memory and to satisfy

those requests some unused procedures had to be removed from

Procedure Cache to free up memory.

has a constant projection and a Having clause with only

constant in it. Here is an example:

select 'x' from mytable group by mycol having 'abc'='abc'

Also the process folder is now split into two tabs normal & blocked.

the table are increased. The performance degradation is due to redundant

memory allocation during DMLs.

As a result of Job Scheduler task terminating, messages such as "Job Scheduler Task set to stop", "Job Scheduler Task was not stopped by a Task shutdown request" etc. appear in the errorlog file.

support dynamically claiming more buffers from the tempdb

cache (if it is a named cache). This is done provided there

are still buffers available against the "max buffers per lava

operator" configuration parameter for the given operator.

This can significantly improve performance and reduce the

tempdb space used. Use "set statistics plancost" to check for

I/O against these operators or if you get an 1105 error.

be processed correctly by ASE Web Services Consumer. The

consumer.log file would contain two log messages similar to:

INFO [Thread-1] - Begin parsing WSDL File: [<URL for WSDL file>]

ERROR [Thread-1] -

there is a system attention, there could be a

stack trace printed due to the attention. This

doesn't mean corruption or any other problem with

the server, it happens only with certain exact

timing of the system attention.

however, the next page ... has not been marked as an overflow page."

may be reported by DBCC CHECKTABLE on an All Pages Locks table that has

a clustered index whose overflow pages deallocations needed to be redone

by the recovery process either at boot time or after the LOAD TRAN command

was run.

the house keeper and log information about IOs in an event buffer.

Usage: pssinfo(<spid> | 0, 'isolation_level')

1965 bytes is larger than the maximum size (1964 bytes) allowed for this table." or 9557 error "The

server failed to create or update a row in table 'sysstatistics' because a varying-length column would

start at offset 8192." is reported for sysstatistics during ONLINE DATABASE command after the

database was loaded across platform.

RepAgent(8): Missing datarow in TEXT/IMAGE insert log record.

Transaction log may be corrupt.

when TEXT/IMAGE is inserted from XFS proxy table.

unless ASE is restarted, if a session failed to execute this stored procedure

because of error 216 "Attempt to automatically drop temporary table failed."

that was hit due to disk space shortage in a temporary database and ASE has

also aknowledged an attention event (Control-C) emitted by this session

following these errors.

623 or 806 error is reported by ONLINE DATABASE command after loading a database with a

cross-platform dump and the database is in read-only mode.

on syslogins while waiting for the client application to send

the password causing concurrency problems.

after load database, if the dump file is created by backupserver which is started with -m option

and/or BLOCKSIZE option of DUMP command.

'ind__get_sysindrow' together with a stack trace may be reported in the error

log if a trigger containing "inserted" or "deleted" is fired on a table that

has been renamed and that the old table name is used on another object

(e.g. a view).

To fix this problem, the ASE will check if the client could support large identifier or not, if the client does not support large identifier and the identifier name is longer than 30 bytes, the ASE will raise an error msg; otherwise, sending the large identifier.

error (overflow, divide by zero, format error, etc.) OR subquery

error like 512 OR a deadlock is hit, then a subsequent cursor FETCH

may fail with error 803 resulting in the session being killed with a

stack trace reported in the error log showing modules 'curs_fetch'

and 'bufkeep'. With this fix, the default behaviour is to end the cursor scan.

However, if Adaptive Server is started with the 589 trace flag, then

the cursor is implicitly closed instead.

an 'interrogate' event by utilities like "IBM Director" or sc.exe. Event viewer will show

"error 7016: The Sybase XPServer_JEK2K1253_XP service has reported an invalid current state 0."

The signature of the stack trace will include: s_getTreeOrPlan()->proc_get()->

proc_read()->alloc_proc()->kisignal()

maybe be raised when executing a stored procedure containing a SELECT

CONVERT( <java datatype1>, <java datatype2>) statement in a database that

was loaded from a dump of a database with different dbid.

If the source database is dropped errors 913, "Could not find row in sysdatabases

with database id <dbid>. Run DBCC CHECKTABLE on sysdatabases." and 11060

"This statement has exceeded the maximum number of recompilations (10). This

indicates an internal error. Please note the error prior to this one and contact

Sybase Technical Support" are raised instead.

We should log seperate restamp log records for the following pages.

1. OAM pages

2. last page in the datapage chain of partition.

This is because these pages get modified after restamp operation.

For recovery to proceed properly, they should be logged seperately.

while the nested loop join may show better performance.

The fix is: For a large DOL table, use SQL "create existing table ..." to create proxy table. Unlike "create proxy_table", "create existing table" allows definition of lock scheme. This ensures the proxy's table's lock scheme is the same as the source table's.

for queries with hash operator's to get into an infinite loop in cm_cache_wash()

(as seen from sybmon stack dumps). This was caused by the hash operator

incorrectly grabbing buffers from the tempdb's cache's largest pool. This

has been fixed. Note, that in these cases, the session could not be killed.

clustered index on descending column

can run into error 1530.

type is different, if UPDATE [INDEX|ALL] STATISTICS command was executed on sysstatistics, sysusers or other system

table before the DUMP DATABASE.

with a stack trace which includes the modules like 'log_to_virt', 'virtmap ', 'bufgrab ',

'getpage_with_validation', will be reported if the requested size for master database

is not multiple of size of allocation unit. The similar error will be reported in

case of model database also.

may cause 'remote i/o' hang on the spid.

(without changing the server's default character set at the same time),

and before rebooting, a subsequent change is made to nearly any parameter,

then after the reboot the new sort order appears to have taken effect

(according to the errorlog and sp_helpsort) but it doesn't in fact work.

msg 1767: "Number of variable length columns exceeds limit of 254 for allpage locked tables.

CREATE TABLE for '<proxy table name>' failed." if the remote table contains more than 254

variable columns.

metadata descriptors. Increase the serverwide '%s' parameter or the '%s' parameter." is incorrectly

raised when multiple spids are running the same LWP in a multi-engine ASE environment.

coming from an inner nested left outer join. Example:

SELECT T1.PNO,

T1.PNAME,

T2.PNO,

T3.PNO

FROM PARTS T1

LEFT OUTER JOIN (PARTPRICE T2

LEFT OUTER JOIN SUPPLY T3

ON (T2.PNO = T3.PNO

AND T3.PNO > 'P1'))

ON (T1.PNO = T2.PNO)

T3.PNO may be incorrectly missing NULL substitution.

to grab a group of identity values reserved for that process.

if at least one of SQL commands is using TEXT or IMAGE host variables and

the statement cache is active on the server.

jConnect: The column names used to describe ResultSet MetaData do not conform to the JDBC specifications. [CR 419579]

sqlupgrade: Errors encountered when installing system stored procedures in Japanese environment. [CR 418053]

jConnect: Request to support getDatabaseMajorVersion(), getDatabaseMinorVersion(), getJDBCMajorVersion() and getJDBCMinorVersion() functions on the DatabaseMetaData class. [CR 416759]

unichar column that is shorter than the actual length of the column,

and where the unichar column is not aligned on a 2-byte boundary in

the page may cause the process to terminate with a bus error.

items, XML model will create a logical item without symbol.

However, it is reported that this external shortcut will affect the items

ordering of the XML Model in PD12, PD11 and PD10

I hope you can appreciate my point: if I've explicitly told job

scheduler to go down, I think it's reasonable to expect it to stay down

until told otherwise.

may sometimes return incorrect results.

It may cause performance drop after upgrading from pre-15.

a Namespace's item , is still allow to map to OOM or PDM item.

However, in XML model of PD12, the "Mapping" Tab of items will disappear as

long as it become a "shortcut" by defining the "reference" property.

limited to a configurable percentage of tempdb's pagesize buffer pool.

This percentage is controlled through the "max resource granularity" configuration

option. This value defaults to 10%. Before this checkin, each instance of

these operators could keep upto 50% of the MRU area of the pagesize buffer pool.

and session 1 executes DDL and/or DML commands while session 2 is interested in

printing the execution plan for session 1 through SP_SHOWPLAN.

for a single query, performance may be affected adversely due to lengthy

chains.

will print some spurious multi-attribute sets that cannot be corrected by update statistics

e.g. references to the "Rid" attribute, as well as references to BIT, IMAGE, TEXT attributes that cannot be placed in update statistics commands

this fix will limit the hints provided by the tracing to just those that can be corrected by update statistics

could result in a very different query plan, sometimes a very bad

performing query plan.

"Both the READ ONLY and the FOR UPDATE options have been specified for cursor"

might be raised.

consume significant portions of the procedure cache and spend significant

amounts of CPU time in memalloc() (as seen from sybmon stack dumps).

This was the result of extraordinarily long subordinate PROC_HDR chains

where the subordinate PROC_HDR is effectively empty (as seen from sybmon

"procmem prphdr <main_phdr_address verbose" command). This has been fixed

by freeing and recreating the PROC_HDR as processing proceeds from one input

set to the next. Sessions in this state can significantly impact overall

system performance.

If a table t1 has an index i1 on columns (a1,a2,a3) and multiple IN lists are defined e.g.

select * from t1 where a1 in (1,2) and a2 in (10,11) and a3 in (100,101)

the optimizer would only consider costing subsequent IN lists if the previous IN lists were reasonably selective. The change is to cost all IN lists, since strongly selective IN list can follow weak IN in lists, and then determine whether to avoid some less significant IN lists due to poor selectivity.

stored procedure, may unexpectedly survive a server re-start.

with error number 14 (Bad address). Refer to your operating system documentation for

further details." or error 4.139.2.22: [105] "The 'uncompress' call failed for device

'device_name' with error number 0 (Error 0)." is returned, when a database dump file

that was generated with 'with compression=<compression_level>' clause is loaded

across platform.

in a value a few increments lower than in specified behavior. The practical impact is

virtually nil, and any theoretical problem quickly resolves after a few new runtime

updates.

database devices during failover if PID of dataserver process(es) is not removed.

Msg. 11031 is raised. At the same time the error 'Invalid pointer

param number 4, pointer value 0x3' .*s invalid pointer param number 6, pointer value 0x2' is also

raised.

select 1 from r left join s on r.a = s.a where s.a = NULL

one of the equi-join columns also has "col=<ct>" in the WHERE clause,

the optimizer will not consider using the MergeJoin algorithm,

although it's legal under these circumstances.

"collocate" when a session runs a subquery whose execution requires a work

table to be sorted in parallel by ASE. The modules "psort_producer", "subst"

and "subconst" will be reported also in the error log.

on syslogins while waiting for the LDAP server to respond,

which causes sp_addlogin to hang

manually or by sp_post_xpload, and after a database dump has been loaded across platform

and brought online.

predicate that is always false (in this case 1=2) in the first SELECT of the UNION may not

return all of the rows of the result set if the query plan chosen by the optimizer contains

a hash-based UNION operation. The presence of the hash-based UNION operation in the query

plan is shown in showplan by the line: HASH UNION Operator.

a user with negative suid.

that contain unknown constants at compile time (i.e. local variables or substituted columns

of a join clause) can be used to eliminate scans of some of the partitons of the table based

upon the runtime values of the constants. For example:

create table t1 (c1 int, c2 int) partition by range (c1) (p1 values <= (50), p2 values <= (100))

declare @v1 int

set @v1 = 49

select * from t1 where c1 < @v1

The select statement contains a range predicate (c1 < @v1) that is unknown at compile time.

This predicate will be used at the beginning of execution of this statement, when the

value of @v1 is known, to determine

which partitions of t1 could possibly contain qualifying rows, and will cause the

elimination of the partition, p2 from the scan of t1 because its values of c1 are all

greater than 49.

of systabstats statistics for an index or an index partition.

are non-null values and all the columns require translation and all translation type is same length,

when database dump file is loaded across platform.

partition ID <ptnid>; at offset <indoffset>; and data page <data pno> (row ID <rnum>), partition ID

<ptnid>. Drop and re-create index ID <indid> of table '<tabname>' in database '<dbname>'." and

abort silently, or return "Invalid parameter" error with 2510 error message.

clause (and having 32 or less distinct group-by columns) will have a

predicable ordering similar to that of ASE 12.5.x. Namely, in the absence

of an explicit order-by clause, ordering will be that of the group-by columns.

joining relation(s).

"There was an error reading the results of the SQL statement. The displayed results may be incorrect or incomplete."

after calling stored procedures (e.g. sp_cacheconfig) in console mode. This has been fixed.

WHERE clause and whose SELECT list columns would make a row too large to store

into a worktable.

at the connection level, through the SET XXX ON/OFF/DEFAULT, where XXX is

nl_join, merge_join, etc. There settings should be available also at the

query level using AP language extensions (as currently the case for the

optgoal and opttimeout settings).

6/26/2006

This issue has been resolved in the PowerDesigner/PowerAMC 12.0.0 EBF4.

7/31/2006

This issue has been resolved in the PowerDesigner/PowerAMC 12.1.0 release.

an All Pages Locking table are dropped, ASE does not incorrectly remove

the status value "Index supports primary key/unique declarative constraint"

from Sysindexes. This results in incorrect SP_HELP output.

which has an expression in the select-list, but also has an expression in the

WHERE-clause that disqualifies all rows will still cause the select-list expression to

be evaluated. In the example, the expression is invalid, so an error will be returned,

even though no rows qualify.

should also use quoted identifiers around the table name.

Error 318: "sysstatistics catalog was read and an inconsistency was found, please run update statistics and contact Sybase Technical Support."

It doesn't tell what table need to run update statistics.

It doesn't tell what database need to run update statistics.

Customer would like to have error 318 giving more details on database/object level where there error occurs.

and 'amc_finish_command' may be reported in the error log if ALTER TABLE ... MODIFY a

column to null on a proxy table with any text or image column

sysusages row that controls the allocation page <pageno> while doing the extents accounting for

OAM page <pageno>, object ID <objid>, index ID <indid>, database ID <dbid> (OAM entry nn).'

printed in the errorlog during ONLINE or 692 error during sp_post_xpload, after 15.0 database dump

file was loaded across platform if there is an un-upgraded allocation page in the 15.0 database dump

file. This problem happens only after a pre-15.0 database was upgraded to 15.0 and was dumped again

on 15.0, then loaded on 15.0 again across platform.

bulk copying data to a replicated text column with DB-Lib bulk copy routines may cause

error 7105.

the error log in module "ind_dbinfologspacevalid" when the LCT_ADMIN built-in

is executed with option "logsegment_freepages" on a temporary database that

has not fully recovered yet. The modules "th_lct_admin" and

"th__logsegment_freepages" will also be reported in the stack trace.

is set to on.

a usage error message and prevents ASE startup from proceeding.

creating the manifest file with a process that holds a lock on a system

table or the log semaphore and it's suspended because it tried to write

to the database being quiesced.

"kernel os_find_good_keyfile: Found shared memory segment: key 0xce0013ac, id 1048593

kernel os_find_good_keyfile: If SQL Server using it is not running, remove it by hand

kernel os_find_good_keyfile: Too many dangling memory segments, nead a clean up

kernel kbcreate: couldn't get shmid for kernel region.

kernel kistartup: could not create shared memory"

may be reported in the error log.

in the underlying table have been eliminated during compilation.

generating a RAISERROR statement in the script

following out of sp_addreplicateart.

despite the needed ordering being cheaply available form the existing indices.

This fix adds a user settable parameter in ASEP->Settings Tab, to control the login timeout period. By default it is 30secs.

"current process is infected with signal 11" in the module "ind__ides_getnext()"

The stack trace may include the modules, open_system_table(), ind_get_ides()

and ind__get_sysindrow().

non-clustered index scans when sargs were being applied to data rows.

For the case of non-clustered indices, the benefit of sargs being

evaluated in the access layer was costed correctly, but this was

not the case for clustered indices. The end result was that

a clustered index scan was not usually chosen when strong filtering

predicates were applied after a weak limiting scan.

The end result is that for APL tables a slower non-clustered

index scan could be chosen over a faster clustered index scan.

if the given table resides in a temporary database.

and 'default character set id' may not work correctly if the old and new values for first parameter

in sp_configure are same.

offline, online for standby access, read only." if the database is set to readonly and an upgrade of

the database that was loaded from pre-15.0 version of ASE has not happened yet.

This fix now adds "if/exists clause", and the object if already exists is dropped, before creating a new definition.

due to which those mappings in the table have references to non-existent users.

with a subquery as the argument.

turn on both 'statement cache' and 'enable literal autoparam' configurations.

- the query is unconnected, in that not all FROM clause tables are

connected with eachother through join clauses;

- a subset of the FROM clause tables form a STAR query, in that there is

one table that is equi-joined with each of the other ones;

- another subset of the FROM clause tables contains at least 2 tables,

that join with eachother but don't join with tables in the STAR subset.

select * from titleauthor where au_id ="409-56-7008" or title_id ="PC8888"

If you have enabled index_union, this example uses an index on au_id to find the row IDs (RIDs) of all titleauthor tuples with au_id =409-56-7008", and uses an index on title_id to find the RIDs of all titleauthor tuples with title_id =PC8888". Adaptive Server then unions all RIDs (to eliminate duplicates). The resulting RIDs are joined with a RidJoin to access the data tuples

If index_union is disabled, Adaptive Server does not use an index union strategy in a query to limit the table scan. Instead it uses other access paths on the table (in the example above, it would use a table scan for table titleauthor), and applies the or clause as a filter in the scan operator.

that is in the process of auditing may block each other without ASE being

able to solve this blocking situation.

Interface is "sp_monitor 'deadlock'", followed by various arguments. An archival facility is

provided to store deadlock events in a user-specified table. Detailed usage information can be

obtained from: "sp_monitor help, deadlock" comand.

Connection monitoring is now enhanced with new command "sp_monitor procstack",

that reports on a task's procedural execution stack.

Use "sp_monitor help, procstack" for more details.

If device size is not specified, DISK INIT command automatically recognizes size on raw devices.

of index id 2 of table 'syscolumns' ..." may be reported when a syscolumns row is

being updated in the deferred mode as shown by sp_showplan and the row being updated

is the forwarded location.

statements will be sharing the same query plan:

select * from unitb where unicol = U&'\20ac'

select * from unitb where unicol = U&'5493\4e54'

bcp 'out' command used with the '-C' flag fails with Msg 225

when copying out encrypted data.

"Cannot drop or modify partition descriptor ...

as it is in use. Please retry your command

later. Total reference count ... Task reference

count ..."

May be seen rarely when create placement index

command is rolled back for resources reason.

reports missing stats on columns with BIT datatype, which is noise, since update statistics cannot be done on BIT datatypes

An example of such a query would be:

SELECT *

FROM (SELECT 1 AS Col1) TableA

LEFT JOIN

(SELECT 1 AS Col2) TableB

ON 1=2

could occur in terminate_process() effectively shutting down

the server engines.

owner <ownername>" will be reported in the errorlog when a user with sso_role

specifies a column for encryption on a SELECT INTO statement.

2) When the "CREATE" audit option is set, encrypted column names and key names

are not added to the extrainfo column of the audit log for the SELECT INTO statement.

on disk <n> for both data and the transaction log." may be raised in error

during CREATE, ALTER or LOAD DATABASE if the database has any disk fragments

allocated on virtual devices between 248 and 255.

selectivity of predicates which did not have limiting sargs.

A query such as... select count(*) from t1,t2 where t1.a2 LIKE "%abc" and t1.a1=t2.b1

would have no limiting sargs since the wildcard at the

beginning of the LIKE expression precludes this, yet the

histogramming can still be applied, by evaluating each

histogram boundary value with the LIKE expression.

In 15.0 expression histogramming was extended to all predicates

that were not sargs, including LIKE predicates, to estimate the

selectivity.

The improved selectivity estimate should result in better join

order selection.

criteria. For example, use allrows_oltp optgoal and turn off nljoin at the

same time could sometimes lead to the case because the only viable

join operator for some query under the optgoal would be nljoin.

INSERT ... SELECT DISTINCT ... ORDER BY

which also satisfies:

1. there is some expression as part of the SELECT DISTINCT items.

2. some ORDER BY items are NOT part of the SELECT DISTINCT items.

and stack trace occurs when trace 2901 is on

when create placement index command fails due

to resources reason.

outermost stored procedure, will cause incorrect results when

multiple stored procedures are nested and returning output parameters.

overoptimistic on the PIO savings from large IOs, even though there

is a weak data page cluster ratio for the index.

The correction is that more PIO will be estimated due to a poor

data page cluster ratio that indicates that data pages are

not well clustered for the index scan.

may be reported in the error log in module "dbt_unkeep" during

the execution of an UPDATE command whose trigger has a cursor statement

that references TEXT/IMAGE columns. The modules "xact_commit" and

"xact_end_session" will also be reported in the stack trace.

number of subscriptions, 'sp_resumeXXX' command

may fail to resume a subscription or primary

connection; the status will not change from

'DOWN' to 'UP'. The condition may indicate that

the maximum number of threads used by Replicator

has been reached. The maximum number of threads

limit may be increased using 'sp_configrep'

command with the new 'max_threads' configuration

parameter.

of rows from the INSERTED/DELETED table and the associated INSERT/DELETE/UPDATE statement

may not be rolled back accordingly.

'invalid handle exception' during shutdown.

These deadlocks are exposed because the table locks are removed as part of the project.

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

the modules 'd_procbuf' and 'exec_dbcc' maybe be reported in the errorlog

when DBCC PROCBUF command is executed.

Error 7720 "Update/Delete from a table which is the inner table of an outerjoin is not

allowed. Command aborted." should be raised.

and caused stacktrace from LePipeRWState::LePrwsRWVtuple(LeVtuple*,pipe_client_rwmode,int)

during execution.

the query's FROM list (as for non-combinable aggregates, extended columns aggregation,

GROUP BY ALL, etc.), the Abstract Plan has no way to distinguihs between them. We need

to generate unique internal names in such a case to be able to apply lated such an AP.

Sun Solaris, IBM AIX, HPUX, Windows, Tru64.

select * from a, b, c where

a.a1 = b.b1 and a.a2 + b.b2 = c.c2

In this query, the predicate "a.a2 + b.b2 = c.c2" contains a virtual column "a.a2 + b.b2" as part of the equijoin.

In this case, the nary-NLJ might skip processing some rows of b.

Note that if the same predicate is written as a theta-join "a.a2 + b.b2 > c.c2", the correct result will be returned.

timeslice error may occur.

some search conditions to be sent to the remote server.

operations when processing large datasets. This could possibly

lead to duplicate rows in the distinct case and/or incorrect

row interpretation in the aggregation case.

"Arithmetic overflow during implicit conversion of INT value '576002052' to a SMALLINT field" in error log will be reported when

running "dbcc checkstorage" if installdbccdb script is not installed in master database.

1) Add an sp_configure option called 'restrict parallel execution' - 0 meaning no restrictions, 1 meaning only specific users could run in parallel.

2) Add a security role 'pqe_role' so that users could use the 'parallel' keyword on queries and 'set parallel_degree', 'set scan_parallel_degree' for sessions.

What would happen:

If 'restrict parallel execution' was equal to zero (0) then the system would function as it does now. 'max parallel degree' and 'max scan parallel degree' would determine whether or not queries ran in parallel.

If 'restrict parallel execution' was equal to one (1) then the system would not allow users without the 'pqe_role' to run any query in parallel. If a user had the role enabled they would then need to specify the 'parallel' keyword in the query or set parallelism using the 'set parallel_degree' and/or 'set scan_parallel_degree' for the session.

Sequence to Enable for SA user:

sp_configure 'restrict parallel execution',1

go

sp_configure 'max parallel degree',50

go

sp_configure 'max scan parallel degree',20

go

sp_role 'grant', pqe_role, sa

go

sp_modifylogin sa 'add default role', pqe_role

go

DBCC checkcatalog(<DBNAME>, fixall) has been used to correct Sysindexes

descrepancies.

you have active level0 scans on the target object. Retry the command

after terminating the scans." will be incorrectly reported when trying

to unbind an object from a named cache in a read-only database.

fail with the error message 691 "Encountered invalid logical page '0' while accessing object '32'

in database '1'. This is an internal system error. Please contact Sybase Technical Support."

if the query is exectued in the parallel mode.

Sybmail to support MAPI so that the default mail client can be used to send mail notifications.

- closed as dup of feature request 356901.

In SQL Advantage when you compiled SQL/query and an error was returned, if you double clicked on the error it would highlight the line in the SQL/query where the error was returned from, this is not the case with Interactive SQL.

add functions to respond to commands by context menu item

"Print" and submenu "Search".

a Replication Server session may report error 225 'Cannot run query--referenced object (NAME NOT RECOVERABLE) dropped during query optimization' when the query processes an encrypted column.

occurs on DOL tables due to the ILA "Index LookAhead" optimization.

This will only occur if the access pattern for rows switches on

ILA at the parent level and the next access moves to a new partition.

This would generally occur when the table is on the inside of a NLjoin

strategy. Note, the problem does not occur with the ILA optimization

is done at the leaf level. Problem can be worked around using trace commandline

flag 12611.

upgraded and a query had been executed. All users, other than the users with

sa and operator roles, are not allowed to access a database that is yet to

be upgraded to prevent errors in databases that have not been upgraded.

result sets, only the first result set is returned.

WITH COMPRESSION = <compression> and the AT <remote backup server> options

cannot be loaded using a local backup server (that is, without the AT

<remote backup server> option). In addition, the dump that is created

is not compressed despite compression having been requested.

could send incorrect or inaccurate information in the column format

descriptions for the user type id field. This only happens the first

time that the formats are sent after the DYNAMIC PREPARE command,

subsequent DYNAMIC EXECUTE commands return the correct format

information.

a select into query with nested union operators.

stored procedure whose execution time exceed the configured limit may hit TDS protocol

errors instead of the expected 11005 error reading "Transaction's running time exceeded

limit of <TimeLimit>."

an 11005 error may be unexpectedly raised after a transaction has been explicitly

rolled back.

correlate between multiple tables.

ObjectID, StatementNumber, OwnerUID,

CPUTime, PhysicalReads, PhysicalWrites

MemUsageKB,ClientHost,ClientIP,Status

ASE may report in errorlog messages like

"kernel secleanup: time to live expired on engine <eng#>"

"kernel Terminating the listener with protocol tcp, host <hostname>,

port <port#> because the listener execution context is located on engine <eng#>,

which is not responding."

decrypt access to that column.

select statement with a literal followed immediately by a correlation name in the select

list may get assertion failure in sqtcache.c.

normalization' is on, we may get wrong result.

reporting physical IOs. This was because it was accounting

for all physical reads rather than just the physical reads

used (in readahead cases).

contains statement like CREATE INDEX, UPDATE STATISTICS or DROP TABLE followed by a

SELECT statement from a view in a database different than the current database.

"Client message: ct_connect(): protocol specific layer: external error: The attempt to connect to the server failed." and "ct_connect() failed."

it is executed is smaller than the database being verified.

includes different modules like 'memalloc', 'make_orstruct', 'parscpy' or 'q_or_find' may

be reported in the error log when a query with a long IN list is executed.

to print partition-level and object-level housekeeper garbage collection queue statistics

logical page '<pgno>' in sdes for object 'sysprotects' - either there is no room in

sdes or buffer already in requested slot." followed by a process termination will happen when

granting access to a group that contains more than 11 users that already have the same access granted with grant option.

DDLGen generates ddl for all triggers of a table. The value to be supplied in the name parameter '-N' would be <dbName>.<tableOwner>.<tableName>.<triggerName>. Please note that if DDL is to be generated for all triggers of a table then in the above -N parameter '%' should be used for trigger name. Also note that -D parameter cannot be used when generating DDL for all triggers of a table.

therefore symbolic names are not printed in stacktraces when running diagsrvr.exe.

version 12.6 and above may not return any error messages to the BCP client when a constraint

violation error is encountered while inserting data into the Oracle database.

with a stack trace which includes the modules pss_getunkept_pss, sysxacts__fmtrow_from_xdes

end sysxacts_materialize may be reported in the error log during select from SYSTRANSACTIONS

in a ASE with high transaction load.

A new type 'LC' is added for Logical Cluster. The syntax for generating DDL for user defined logical cluster is:

ddlgen -U[userName] -P[password] -S[servername | host:port] -TLC -N logical_cluster_name

debugger just continues without stopping on the breakpoints set.

HK Chores system process to retry the drop.

a query with a lot of joining tables.

expected for varchar and char variables : result of expressions involving

local variables may be trunacted.

CR 435065 - Update the Meta Data stored procedures used by ODBC, OLEDB and jConnect drivers

and user connects to ASA Plug-in first before connecting to ASE Plug-in.

in module hk_accelerate() while it was in the process of creating a temporary table

and therefore holding exclusive locks on system catalogs. Other sessions that create

temporary tables concurrently will be blocked. The session could sleep due to ASE

having enqueued to housekeeper a request to flush to disk the systabstats statistics

for the being created temporary table. This proactive enqueue request is not deemed

necessary because the descriptor for a temporary table is expected to remain in cache

always. If this descriptor needs to be reused, ASE will flush its systabstats statistics

on disk first before reusing it for another object.

may hit error 8221 "Couldn't find the identity column for object <value>"

and hang forever while undoing the creation of the given target table.

without the requirement to either unmount or quiesce one or more databases.

in database '<value>'" may be raised mistakenly during the execution of

DBCC CHECKALLOC with the FIX option or DBCC TEXTALLOC with the FULL and

FIX options, if the object being corrected does not exist anymore in

the given database. This error results in the abort of the given DBCC command.

it is done on unichar or univarchar columns.

subqueries in the target list could return incorrect

values for those subqueries if the same subqueries are

used within predicates in the WHERE clause.

containing double-byte characters over 255 bytes, the audit records may be split in the middle

of one of double-byte characters.

may be reported and logged in the log. In addition, a job may be further processed even if the errors are reported.

capture and report the space usage characteristics, and an estimated

measure of fragmentation, for tables, indices, and transaction log of

a database.

STR, SPACE) in the select-list and that have literal values as arguments to the builtin

function and have literal auto parameterization enabled and have the statment cache enabled

can get output results from these builtins that are blank-padded up to a width of 255 bytes.

row <value> with status <value> must be the forwarded row location instead of

the home row location." and the call stack of the given session as opposed to

silently continue.

or buffer already in requested slot." may be raised with a stack trace in module

"pg__allocate_completion" leading to error 1127 "Internal Error: The OAM has no room

for new inserts ..." in module "pg__insertoam" during the execution of an INSERT command

in a table that has more than one OAM pages allocated to it.

there is a possibility of deadlock.

rows will return too few rows if the optimizer selects a query plan with a hash-based

UNION operation. The presence of the hash-based UNION operation in the query

plan is shown in showplan by the line: HASH UNION Operator.

The following query is representative of such a union query where the second

select statement will return no rows:

select 1 union select a from x where 1=2 union select 2

irrespective of it being a dedicated log sergment or not, may take more time to complete

compared to a pre-15.0 ASE system. The recovery process will spend cycles in module

'pg_logdevusage' in reading each syslogs allocation page while establishing statistics

information in module 'systs_fill_catalogs' as opposed to only consider the statistics

information that was retrieved from the systabstats system catalog already.

> This percentage is controlled through the "max space granularity" configuration option.

The above should refer to the session config prop "max resource granularity".

statistics" may be reported after an upgrade to 15.0. On little-endian platforms error

2601, "Attempt to insert duplicate key row in object 'sysstatistics' with unique index

'csysstatistics'" followed by error 511, "Attempt to update or insert row failed because

resultant row of size 1967 bytes is larger than the maximum size (1964 bytes) allowed for this

table." may be reported during an upgrade to 15.0.

a join to at least one additional table. This may occur if the

chosen plan uses a merge join, and the rows are deleted in 'direct'

mode (does not occur with deferred deletes). This may only happen

when deleting from APL tables and requires duplicate values in the

joined column.

trigger, or SQL batch. Retry later, or ask your SA to reconfigure ASE with more procedure

cache." while allocating memory from the procedure cache when the procedure's cache

memory is fragmented.

than 1.

causing bad performing index scan when the ON clause was involving

expression on one side of the join. Also false join could be

added to the plan causing wrong results when there were nested

outerjoins in the query.

stages of the execution. In certain cases, a stack trace will be encountered

containing the method icp_crecvmsg().

over-written by a work table oam page, in some rare cases.

followed by either error 18140 or error 10855 when respectively an attempt to create a logonly

cache or an attempt to drop a cache is made.

during the execution of builtins like data_pages(), reserved_page(), row_count() etc.

The traceflag 1116 can be used to suppress these errors in non-temporary databases.

In temporary databases, these errors are suppressed by default.

mode for DELETE and UPDATE queries where there is a join or an RI constraint.

This functionality had not been migrated to Galaxy. This is fixed.

feature. This results in two statements that vary only by the amount of whitespace around

their literals caching two separate statements in the statement cache and two plans in

the procedure cache. For example, the following two statements will both be cached

with their own plans:

select 1, 2 from sysobjects where id < 10

select 3, 4 from sysobjects where id < 100

With the autoparam feature, these two statements should be considered equivalent.

minimum row length of 64" together with a stack trace which includes the modules 'collocate' and

'translate_varcols' may be reported in the error log when a database dump from a previous ASE

version is loaded in a 64 bits ASE on IBM AIX.

session being killed when ASE receives an external attention "Control-C"

while it was instantiating a database information in cache. The modules

"dbt_get", "dbt_removeall" and "des_discardall" will also be reported

in the error log.

error:

The UPDATE WHERE CURRENT OF to the cursor '<cursor_name>' failed because column '<column_name>' being updated was not specified in the SELECT list.

The error can occur if the optimizer chooses to join the tables in an order that is

different from the order of the tables in the FROM clause of the DECLARE CURSOR statement.

after sp_chgattribute "<TABLE_NAME>", "identity_gap", 0 has been run for a

table whose identity column has already reached its maximum value.

DBCC CHECKTABLE on sysdatabases." may be reported after loading a database and

executing a stored procedure that inserts explicit NULL values in columns

declared as java classes.

e.g. If there is a subquery that needs to be flattened to the outer query which has

an order by clause.

at '<buffer_address>' with a buffer pageno of '<page_no' and database id '<dbid>' with

HASHED status set failed - buffer was not found ... " may be raised in module cm_bufunhash()

recursively resulting in ASE shutting down itself.

in the error log in module "dbt_cacheremove" during the execution of dbcc dbcacheremove()

when execute cursors exist and are not closed in the database being uncached.

query involving remote tables.

This happens if the UPDATE is executed in the stored procedure and the value

to be set is passed as the procedure parameter and represents a string of spaces.

This problem happens only if the option string_rtruncation is set.

may be raised followed by the message "Illegal attempt to rollback transaction (xactname

'$chained_transaction') that has already committed" when committing a transaction that has

a select into command followed by a drop of the table just created. The problem is restricted

to use of large page allocation, objects with Data-Only-Locking, databases with "ddl in tran"

active and "set chained on" option.

warning being sent to the client. This option has only effect on informational warnings, error

messages when necessary are still being raised.

virtpage '<PageNumber>' from virtual device '<DeviceNumber>' for object '<ObjectId>' in database '<DbId>'. The page

was not read successfully. You may have a device problem or an operating system problem." may be raised

for a page belonging to the log segment.

complex constrainst in the same statement.

exhaustion of time slice during rollback.

'LeHashContext::HashVtuple' together with a stack trace which includes the modules

'LeHashContext::GetNext' and 'LeHashContext::ProcessTuple' may be reported in the error

log when tempdb is bound to a name cache and the ASE is using a HASH operator to execute

the query. Other stack traces including modules like 'LeAlignedDataRow::readRow' and

'LeHashBufferReader::GetNext' are seen as well. The diagserver reports a 685 error.

of bulk insert from host, please use the proper tools for this command."

may be raised followed by the message "current process ... infected with 11"

in module bufauxuninstall() during a bulk insert operation.

select * from

(t1 left join t2 on t1.a = t2.b) -- OJ #1

left join t3 on t2.b = t3.c -- OJ #2

where t3.c = 5

In this query, the where-clause predicate "t3.c = 5" will eliminate all null-supplied rows from OJ #2. This in turn will result in all null-supplied rows from OJ #1 being eliminated. This problem has been fixed.

optcriteria cr439123 needs to be on to see the change.

if it is accessed in a batch together with a statement that gets cached.

"How do I save multiple licenses to a single file"

if procedure cache request results in a procedure cache pool growth after

a new shared memory segment allocation.

sp_helpdevice shows "directio on" for the device as if ASE supported directio on the platform,

but ASE will activate the device using dsync on and deactivate using directio at startup.

an outerjoin inner table.

into a unicode character column on a proxy table.

comparision to avoid the use of login name with mixed upper and lower case letters in the password.

to $SYBASE/$SYBASE_ASE/locales/{language}/{charset} directory in ASE 15.0.2 release.

1) Current usage, maximum usage, minimum usage

2) Number of allocations

3) Type of allocation

4) Who/what made the allocations (including size, date/time, spid, user name and command name).

5) Average size (mean) of the allocations

6) Standard deviation from the mean

The stored procedure should be called sp_procmon with the following parameters:

sp_procmon <duration> [sample rate] [output table]

Duration is required and represents time in minutes.

Sample Rate is optional and by default would be one tenth of the Duration time so that only ten samples would be taken.

Output Table is optional and is for the raw data collected by sp_procmon.

This is just a suggestion and as log as the outcome for the customer is a simple way to monitor and configure procedure cache size I don't think they would care how we implement this feature.

DBISQL do not pickup the 2 Print in a batch like this.

print "Print test 1"

select "Select test 1"+

print "Print test 2"

go

external authentications to failover to ASE logins, if authenticated by ASE 12.5.4 or later

the configuration option 'enable ldap user auth' is on.

compared in a where clause.

during a create database command.

4096 or greater.

a correlated subquery and the correlation involves some

join between the union projection and an expression,

following compilation error could happen:

Msg 403, Level 16, State 2 Invalid operator for datatype op: type: VARCHAR.

there are literals in the where clause.

will return a status which contains the switch properties.

and performing dynamic index strategy to resolve OR predicate may fail with different

errors related to sort operations including error 1509 "Internal error encountered in merging rows;

resubmit query batch." Sometimes query execution may fail with stack trace logged in the errorlog.

The module shown in the backtrace is bld_dynindex.

in ksct_startio when accessing proxy table mapped to the remote ASE.

The document needs to be updated to reflect the installation advice for Tru64 Unix.

stacktrace which includes the modules 's_execute', 'exec_eop' and 'run' maybe reported in the

errorlog when running a SELECT with a UNION ALL clause and ORDERBY selecting text or image columns

from a table.

'lddb__stat_translate' together with a stacktrace which includes the modules

'lddb__unscr_stat_page' and 'lddb__unscr_usro' may be reported in the errorlog

when running LOAD DATABASE across platform where the endian type is different.

Other errors, like 'Invalid column length' error, 1142, 692, or 12315 error may be

reported.

a subquery when the predicate involves an expression with

correlated column(s). This could cause less performing

plan.

This may lead to improper rollback of AMC operation

(In our case, the partition rootpage is not rolled back properly).

A wrongpage error (605/696) will be thrown upon subsequent access of the table

housekeeper 'HK_WASH' system task and a session that performs a DML

command in a table whose database has mixed log and data segments

when that table uses a cache that has a memory pool configured.

constant_expression(s) only. For example,

select count(distinct 1+2) from t

Giving a correct usage would avoid this problem.

i.e. declare @var int select @var = <a value>

+ No mclib in java.library.path

+ This application has failed to start because javaai.dll could not be found. Reinstalling the application may fix this error.

despite ASE having granted the lock to the session. This is a side effect

from the fix of CR 399737. The modules "lock__semawait", "log__lock" and

"lock_psema" will be reported in the call stack of the hanging session.

using unichar format U&'***', subsequent insert into the table may fail or insert

into incorrect partition.

predicates. As a result, they might raise user exceptions that could

otherwise be avoided if the evaluation was deferred until just before the

result is returned. This way, filtering predicates might eliminate the

row before the exception occurs. This is similar to the problem fixed

in CR 436738 (where the problem was fixed for scan, remote scan, or,

and restrict ops).

while the semantic partitioning feature is OFF.

above even though the IQ server supports the syntax.

The identifier that starts with 慳bcdefghijklmnopqrstuvwxyznowi� is too

long. Maximum length is 30. Sybase error code=103

In a query such as the following

SELECT

M.fiscal_period

, T.service_key

, sum(T.local_call_minutes)

, sum(T.local_call_count)

, count(*)

FROM

telco_facts T

, residential_customer C

, service S

, month M

WHERE

T.customer_key = C.customer_key

AND T.service_key = S.service_key

AND T.month_key = M.month_key

AND M.fiscal_period = 'Q4'

AND T.service_key in (02, 03)

AND C.state = 'CA'

GROUP BY

M.fiscal_period

, T.service_key

if the join order chosen used nested loop joins, and the T.service_key table was not at the beginning of the join order, then there was a possibility for an incorrect costing.

must not trigger the implicit start of a chained transaction.

in improved scans of the memory prochdrs. This improves response time in

queries where large chunks of memory are consumed.

are no longer part of a database and so should not be reflected in sysusages. Attempting to remove the

extraneous entries can encounter a segmentation fault and stack trace, causing disk refit to fail.

is producing empty result set.

columns (from an outer query block) may return incorrect results.

that involves several complex AND/OR predicates combination

involving multiple columns.

unichar->char, char->binary, or unichar->binary casting.

as the result of deadlock while compiling a procedure. A deadlock situation

during compilation should only occur when compiling a block that updates or

inserts into a table that contains computed columns and/or functional indices.

when executing a query that involves a join with complex

expressions for the join predicates and the optimizer decides

to apply "reformatting" (store index) technique to the inner

table (i.e. substituting it with a worktable with an index

on it)

should mention the following character set related limitations:

+ The XML functions in ASE 12.5.x versions do not support i18N (those in ASE 15.0 do)

+ The XML outputs generated by ASE 12.5.x are restricted to the first 128 byte-points only

user tries to access an MDA table from within a login script in

some 64 bits platforms.

key has been specified for encryption operations using an initialization vector,

then a paralellized update or select of the encrypted column may display the

following error message or a similar one, and the errorlog may show a

stack trace showing the function 'col_encrypt' or 'col_decrypt'.

"Error: 15414, Severity: 18, State: 3

01:00025:00021:2006/09/29 09:14:38.62 server An internal error occurred at end of symmetric encryption or decryption operation (error code 57606). Please contact Sybase Technical Support."

or parameters fails during the compilation of the statement. Stack trace is dumped to the

errorlog file and shows modules opt_like, opt_like_and.

identify correctly which columns are affected

by an UPDATE statement that includes a local

variable assignment and that assignment is the

first one in the SET clause.

used in the query. This happens if the procedure executes SELECT INTO with query producing worktables

(order by, distinct, group by, materialized views) and requires recompilation.

to correct error messages "There are <value> OAM entries with zero used/unused page counts.

Run DBCC TABLEALLOC FIX ..." when the OAM entries needing cleanup are the first in the list.

Because of this sp_extrapwdchecks fails to return error in case of failed password, if the raiserror is followed by any TSQL statement.

Pre-15.x, the "general OR strategy" was enumerated and costed by the optimizer when the WHERE clause contained OR clauses and all columns therein had indexes: each OR term was used in an index scan, then all resulting RIDs, after duplicate RID removal, were used to get the data tuples. This is actually an union-distinct mechanism and is implemented in 15.x as index union.

In 15.x, to request the pre-15.x general OR strategy on table T, i.e. the index union, an AP would use "(union_scan T)". To specify either the merge or the hash based union-distinct algorithm, use merge_union_scan or hash_union_scan. Index intersection is also provided for experimentation but not yet for production, through hash|merge_intersect_scan.

This functionality is used in the allrows_oltp "compatibility" mode (where the optimizer makes the best attempt to generate an execution plan similar to the pre-15.x one) when the legacy plan used the general OR strategy.

database '2'" may be reported when there are more then 128 disk fragements for tempdb database.

the dataserver.

the message "timeslice <value>, current process infected" followed

by a stack trace when ASE receives an interrupt (Control-C). The modules

"do_logtransfer" and "senddone" will be reported in the error log.

if a CTRL-C is issued while accessing the table. A merge-join with a very large

number of duplicate join keys in the inner stream is the mostly likely scenario

where this would occur. In the merge-join case, there is no way to kill

the looping session.

sorted or duplicate rows if special OR strategy is used to execute

the query. For such strategy showplan information shows "Using N Matching Index Scans"

for a query step selecting from the table. This problem may happen only

if some non equality search arguments are used in the query like > or >=.

are incorrectly set in all-pages locked tables, the command CHECKALLOC fixes them,

but TABLEALLOC, INDEXALLOC or TEXTALLOC don't do it.

one of devices is not accessible while creating the new device.

time (since server start-up) and when that operation requires extension of the ASE shared memory

space, the issue occures. The symptom is the impossibility to use the JMS messaging operations

with crash of the external JVM named JRTMSServer. The reason is because the JVM attaches only

to the first ASE shared memory segment in the RTDS 4.0 release.

Msg 9947, Level 17, State 10

An attempt to allocate memory failed due to insufficient SQL Server memory.

1> sp_tempdb 'create', 'newgroup'

2> go

Msg 18941, Level 16, State 1:

Server 'ase15_2k', Procedure 'sp_multdb_creategroup', Line 67:

Only the 'default' group is currently supported.

(return status = 1)

Customer would like the ability to add multiple tempdb groups and multiple databases to those groups.

from the outer scope does not work correctly producing stack traces.

fail with error 3935 reading "Fatal protocol error. xact_beginupdate() API was

invoked in the wrong context..." when it is run for a Data Only Locked table

that has a clustered index created using a pre-12.5 ASE.

Once the connection holding the SH_TAB lock exists, the cursor application continues to wait even though its EX_INTENT lock is compatible with the granted locks of the object.

and the inserted value needs a convert node (or perhaps any other builtin or expression)

we get a sig 11 and stack trace like LeRemScanOp::_checkConstraints()-> LeEvals::LeEvRun()

-> LeRun() -> bin__sqlhilo_X().

if the stored procedure triggers recompilation and the XML corresponding to a

requested query plan can not be found, server may stack trace in

LeInsertOp::_LeOpNext() -> LeDataRow::writeRow().

To reproduce:

1 Create PDM with a table and several columns

2 Generate from PDM an XML model using the XML generation option of "elements" in the Detail Tab

3 In the XML model bring up the Mapping Editor.

4 Set data source to your PDM source gen

5 Try to map table column to XML elements

The XML model only displays the 'database' element but no nested elements to map to.

If I change the embedded type on the database element from 'complex' to 'none', then all the elements are listed in the mapping editor.

flattened for reformating.

between sysobjects and sysprocedures for partition condition objects.

The stack trace will have the following signature at the top:

obj_com_col_ind_pro_cns_check()->confirm()->kisignal().

when no interfaces file name is specified in the backup server start-up command line.

in CIS generating query with the "NULL IS NOT NULL" clause,

which gets error 403 from the remote server:

> Error 403, State 1, Class 16

> "Invalid operator for datatype op: is not null type: VOID TYPE"

as escape character,

the cached statement may contain extra characters at the end.

The error 3250, "The devices '<backup api library name>::<filename>' and '<backup api library name>::<filename>' specified in the command should be different."

will be raised if the file names of two stripes are the same. This error should only be raised for normal or compressed files

and it is an error assuming that an external API cannot use duplicate names.

dbcc proc_cache (free_unused | flush_elc) where

- free_unused: free unused procedures from cache

- flush_elc: flush pages from Engine Local Cache.

This DBCC needs to used with care and when the system is not in heavy use.

reduces the need for join backs of non-combinable aggregates.

Previously there was a limit of the page width for an

aggregate row, which is now not the case.

if the keylength exceeds 600 bytes (on a 2k page system).

This stored procedure uses another stored procedure, for formatting its output. This new stored procedure was added as part of 15.0, and causes a slowdown, owing to an expensive computation with it.

For purposes of this fix, the formatting stored procedure was backed out, for returning the performance of sp_sproc_columns back to 12.5.4 numbers.

The stored procedure in question(that causes the slowdown) is being examined, for performance enhancements.

execution of a query with a string literal that can not be converted to the server's default

character set and has \ escape characters, may cause ASE to stack trace in

ssql_reparse().

cases, recovery may fail with Error 14621, Severity 21 and State 1.

The modules undo_anchor() and anch_process_logrec() will be seen in the

stacktrace.

Unicode support for Sybase ETL.

Resolution:

The level of Unicode support for Sybase ETL is as follows. Unicode characters can be extracted, transformed and loaded. Unicode characters in component properties are not supported. This includes file or directory names, metadata like table or attribute names, and connection settings like database, schema, user or password. Unicode characters in ETL object names or descriptions are not supported.

the table is from a stored procedure parameters and datatype for parameter is VARCHAR while the

base column is CHAR type, wrong result may happen.

if the stored procedure requires recompilation, ASE may stack trace in

LeStrmConst::write() -> LeStrmConst::read().

may stack trace in s_compile_stmt()->oh_init_from_ob()->oh_insert() when executing

long statements with constant literals.

occurs in distinct queries with an order-by clause. Further, the target list must

contain an expression, and there must be a character type column in the target list

or order-by clause, and the number of columns in the target list plus order-by clause

must exceed 32 columns. If all of these conditions hold, then the query may return

the wrong number of rows.

Customer would like to know the versions and date as soon as possible.

1) Secure login passwords stored in syslogins using SHA-256.

2) Store the last login time in syslogins depending on the

"enable last login updates" option.

the session assigned user tempdb can not be dropped after such

cursor is opened.

statistics could be added to help the optimizer.

In particular, the LIKE predicate could use statistics for

expression histogramming, which were being requested by the

set option show_missing_stats diagnostic.

feature. For previous version of PDesigner, allows to define Default Namespace without

prefix. Problem occurs when migrates those XSM file. The non-prefix Namespace will be

removed automatically after the migrate to PowerDesigner 12.0 or 12.1

Check the NameSpaces of the Model:

URI - http://www.w3.org/2001/XMLSchema with Prefix - xs

URI - urn:oasis:names:specification:ubl:schema:xsd:UnspecializedDatatypes-1.0 with Prefix - <empty>

The Second one will get loss after migrate.

if the optimizer is using reformat strategy to execute the query.

may not cleanup properly if interrupted by Ctrl C. As a result, Sysusages

and Systhresholds transactional locks may be unexpectedly left behind.

"with recompile" may fail the error 225 if it accesses a non

temporary table created inside the procedure.

Ex.1) A column is defined using bigint implemented in ASE 15.0, ASE 12.5.x CIS does not recognize bigint data type and will map to varchar.

Ex.2) If you maps a proxy table to ASE 15.0.x and run update statistics command on the proxy table on ASE 12.5.x, error 207 will occur.

Error Msg 553: "The number of parameters/variables in the FETCH INTO clause does

not match the number of columns in cursor 'new_densities' result set." is

raised. The output for 'Column Group' may be incorrect (show NULL) if multiple

columns are involved e.g. composite index.

for 15.0.1 esd1 binary when there is a subquery under OR and a

LIKE predicate under OR at the same time. Here is an example:

...

where (user_name(o.uid) like @uname

or

(@nouname = 1

and not exists (select 1

from sysobjects o2

where o2.name=o.name

and user_name(o2.uid)=@uname)

and user_name(o.uid) = "dbo"

)

)

datatype of TEXT, IMAGE or UNITEXT in the select list on one side of the UNION ALL and

a CONVERT builtin function to convert a literal string to a TEXT, IMAGE or UNITEXT

datatype will result in a segmentation violation with the function, sendtext in the

stack trace.

one of union/unionall projection is a column from an outer join inner

table.

and a stack trace that includes the function, CgGenLava::CGConsMJOp can occur when the

optimizer chooses a query plan with a MergeJoin operation that includes literal values in

the projection list of the inner stream of rows to the MergeJoin operator.

occurs in distinct queries with an order-by clause. Further, the target list must

contain an expression, and there must be a character type column in the target list

or order-by clause, and the number of columns in the target list plus order-by clause

must exceed 32 columns. If all of these conditions hold, then the query may return

the wrong number of rows.

if that fetch is from an empty table, 515 error might be raised from sp_helpartition.

1> set plan for show_execio_xml to client on

2> go

1> select showplan_in_xml(0)

2> go

The call stack will include the function, resetTraceObj.

may incorrectly identify query objects and raise spurious errors like

'Table not found' or 'Permission denied'

Adding a convert() is a good workaround.

But in fact they are 0 (off).

of stored procedures postponing the name resolution phase to

the stored procedure execution time.

If the query joins between the column on the derived table and a column on another

table using LEFT OUTER JOIN clause, the query will get error 325 (Adaptive Server finds

no legal query plan for this statement.).

reported in the error log when SP_SHOWPLAN is run or the configuration option

'plan text pipe active' is set and concurrent activity on table Syscolumns takes place.

Modules 'showplan', 'showKeys' and 'getcolname_level0_fromcache' will also be reported

in the stack trace.

Raise the value of the configuration parameter 'number of open objects'

for more descriptors." may be raised in general, after errors 1204 were

reported "ASE 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 ASE with more LOCKS." when ASE runs out of locks during the

instantiation phase of a descriptor's meta-data, resulting in leaking

free descriptors mistakenly.

One is the general OR optimization which requires a HASH operator to remove

duplicate RIDs and another which removes duplicate keys (e.g. sorts the IN list

to remove duplicates, which can be done at compile time if none of

the keys are variables). The general OR optimization is usually takes

more elapsed time, and the duplicate key approach should be preferred.

This change corrects costing that prefers the duplicate key approach

which could improve performance in some cases.

An example of the type of query that may be helped by this is the following.

select * from sysusers where name in ('dbo', 'sa_role', 'guest', 'probe')

go

" Cannot partition table xxx because your release level does not support partitioning.",

it shows up as a 4941 error whose message is,

" You cannot drop a column on which constraints are defined.

Drop the constraints before dropping this column.".

in the select list may give incorrect result.

been possible to obtain a temporary value to be used for

histogramming estimation. When this was not done then

default selectivities were used for estimates which could

result in incorrect join orders.

All possible constant expressions are now evaluated at compile

time, if it could help with producing a histogramming sarg.

for example, "r.a > <some constant case expression>" would

not be used for histogramming estimates, and instead the

default selectivity of 0.33 would be used for this sarg.

in the inner side of a nested-loop join strategy or beneath a subquery.

A relatively small amount of memory is temporarily lost for each outer or

correlated row in the NLJ strategy or subquery respectively. If there

are a lot of rows in the outer side of an NLJ strategy or in the outer

block of a subquery, then the cumulative amount of memory temporarily lost

can be quite large. Note that this memory is given back to the system

once the query completes.

order by clause may return results in the wrong order. This will

occur if one or more of the columns in the disjuncts always reference

the same literal value.

a query involving a DISTINCT view and the part of the DISTINCT columns

has a unique index on them.

within stored procedures are not initialized correctly during

execution of the stored procedure.

to a large buffer pool, and also have a reserved page count

significantly higher than the table pages. This was due to

a double costing of the page cluster ratio, in addition to

the sparseness of rows in the extent. Adjustments were made

to avoid the double costing so that PIO estimates for large IO

are more accurate

for a login event when a Kerberos principal name and mapped user login name

approaches 30 bytes in length.

chosen for the plan resulting in poor performance.

But "ja_JP.UTF-8" is not included in Linux section of locales.dat.

Please add the following lines to linux section of locales.dat.

locale = ja_JP.UTF-8, japanese, utf8

are corrected during upgrade. In this case, there were

range cells and yet the range density was zero, which

occurs occasionally due to a pre-15.0 problem in

update statistics in which the range cell only exists in the

last cell of the histogram and the remaining cells are

frequency cells. The inconsistency are now corrected during

upgrade to 15.0.

"Datatype numeric ( Column 'xxx' ) is not supported in a text index.".

This will happen only when there is datetype mismatch and implicit conversion is

need and also involve subquery in expression.

This Explanation is "cpu grace time."

<column>, the step value is not increasing" for variable length datatypes.

selectivities for sarg.

when the execution is performed from the cached plan.

@@opttimeout as a server global variable for the current optimization timeout limit.

'close_range' together with a stack trace which includes the

modules 's_estmt_loopend' and 's_execute' may be reported

in the error log if a DML statement is executed involving

a high number of tables, the table being modified has many

referential constraints and there is a trigger defined on it.

columns.

leak if it is interrupted by either a kill operation or reception of an ATTENTION sent by

the client application.

1) The implementation of RSA signature verification in SSL/TLS, or other application scenarios, may incorrectly verify forged signatures leading to security vulnerability. See http://www.sybase.com/detail?id=1047991 for more details.

2) U.S. Daylight Saving Time Changes in 2007, see http://www.sybase.com/detail?id=1048699 for more details.

so that the user can sort the columns in the same order as they were created.

compared with a scalar aggregate result under an OR predicate

and the same column appears under the OR predicate

multiple times. For example,

(A >= (select min(A) from t2) or A >= (select min(A) from t3)).

java.sql.SQLException: JZ006: Caught IOException: java.io.IOException: JZ0EM: End of data.

ETL does not allow multi-byte characters to be used in the meta-data

Resolution:

The ETL product has been modified to handle multi-byte meta-data.