Sybase数据库技术，数据库恢复专家

Project execution might become unresponsive when database runs out of space.

Resolution:

When a database involved in a data transfer runs out of space, the execution of a project might stop responding. This is because the database driver does not return with an error message but waits until space is available to continue. Please ensure that sufficient space is available before running a project or job.

the engine-wide semaphore is not released, causing all the other kerberos logins on that engine to block.

or earlier, causing ASE crash or blocking all Kerberos logins when there are concurrent Kerberos

authentication attempts.

ASE now supports concurrent Kerberos authentication requests by default.

Trace flag 7844 is no longer needed and can be used if concurrent Kerberos authentications is not required.

When using the [CIFS] section of GridNode.ini to store the user and password for connect to a file server, projects that should store Textfiles on a UNIX server fail with the message "tree connect failed: NT_STATUS_ACCESS_DENIED".

Resolution:

On UNIX, installations of Sybase ETL Server ini file settings are not read by the GridNode application unless you have a GridNode.ini in place.

To specify settings for the GridNode application you need to: 1. Make a copy of the Default.ini file named GridNode.ini in the etc subdirectory of your installation path. 2. Specify general settings (to be shared by all applications) in Default.ini. 3. Specify exclusive settings for the GridNode application in GridNode.ini. 4. Disable (remove or comment) the keys for general settings in GridNode.ini.

has an ORDER BY on some expression.

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.

computed column will generate recompilation errors. In the case of a batch,

a 225 error is generated:

Cannot run query--referenced object (name NAME NOT RECOVERABLE) dropped during

query optimization.

error log if a GROUP BY used on a comupted column with sum and count aggregates including a distinct

clause.

two expressions (expr1=expr2) and through some other equijoin predicate(s),

all the columns referenced in the predicate can be completely mapped to

one joining table and at the same time also completely mapped to another

joining table. For example, "t1.c1+2=t2.c1*2 and t1.c1=t2.c1".

evaluate correctly having clause if it uses IS NULL operator.

File Content display in Text Data Provider.

Resolution:

When entering the configuration window of a text provider component, the file content is read with the current (default is ISO-8859-1) encoding setting. It is not reread when a different encoding is selected. The preview section of the configuration window displays data correctly (according to the current settings) and the functionality of the component is not affected.

To read and display the file content with a specific encoding, either

select the encoding in the configuration window and click Save

or select the encoding in the Property section

and reopen the configuration window.

Open Server Error: 16368.10.0.0 srv_senddone(): A message was initialized, but not completed.

Resolution: The install scripts were changed to display the correct numbers.

database from cache.

reported as an index scan in diagnostics (such as showplan) when there

is some predicate on the index column but not on the leading column(s).

0 - unset merge_join at server level.

1 - set merge_join at server level.

2 - set merge_join to default value (as per curent optgoal) at server level.

if the current database has any stored procedures depending on the table you are changing.

You might want to drop and re-create the stored procedures if necessary after changing table schema.

result value via the output parameters for the outermost procedure call.

When this option is OFF, only the return of the output parameters to the

client stream should be suppressed, but the output @variable should continue

to return the result value to the application.

(dirty reads) on an All Pages Locked table and a covered non-clustered index scan is

chosen by the optimizer as reported by the set showplan option with message

"Index contains all needed columns. Base table will not be read."

is larger than the maximum size (1964 bytes) allowed for this table." may be reported during

upgrade to 15*, during "upgrade step [ID 60], alter table (table sysstatistics)" when there are small histogram boundary values at the end of a sysstatistics row.

because it already exists" could be raised by the backup server. The

chances are higher when there are several dumps running simultaneously.

return too few rows. This can happen if the sum of unique distinct

(from the select-list) and order-by columns is less than 32 and there are

no character columns in the select list that are not also in the

order-by list. Taking the following query as an example:

select distinct a.a, a.c from A a, B b

where a.b=b.b order by c

In this case, "c" is the only character column in select-list and

it appears in the order-by list. So, the bug occurs. If "a" was a

character column, then the problem would not occur.

Adaptive Server Enterprise 15.0

System Administration Guide: Volume 1

Chapter 5: Setting Configuration Parameters

procedure cache size

The default value of procedure cache size of ASE 15.0.x is 7000.

Syntax:

alter table table_name drop partition partition_name [with skip_global_index_rebuild]

* If "with skip_global_index_rebuild" is provided, the global index will NOT be

rebuild and the index of the table will be labeled as suspect.

** If "with skip_global_index_rebuild" is not provided, the global index will be rebuild.

This is the default value.

Given this qualifier, the command marks any participating global index as 'suspect',

rather than rebuilding the index as would normally be done.

This can make the command complete more quickly, at the expense

of making an index unusable. To be made usable again, a suspect

index must be dropped and recreated.

may return incorrect results for COMPUTE predicates on

expressions involving columns from the inner table or

may raise a 325 error: "Adaptive Server finds no legal

query plan for this statement."

cannot be greater than 2GB.

Resolution: No workarounds exists - to be resolved in a future release.

plan is chosen where there is an Xchg operator above a Sequence

operator.

Resolution: All column numbers, lengths and sizes in the text source and target are counting Unicode code points and NOT bytes. This means that a code point/character will ALWAYS occupy the space of one code point/character, no matter what encoding is used and regardless of the space used by the representation on disk in the specified encoding. For example a Japanese character will always be counted as one code point, whether the encoding is shift_jis, euc_jis, UTF-8. UTF-16, or UTF-EBCDIC all of which will encode the code point with a different number of bytes.

Resolution: All column numbers, lengths and sizes in the text source and target are counting Unicode code points and NOT bytes. This means that a code point/character will ALWAYS occupy the space of one code point/character, no matter what encoding is used and regardless of the space used by the representation on disk in the specified encoding. For example a Japanese character will always be counted as one code point, whether the encoding is shift_jis, euc_jis, UTF-8. UTF-16, or UTF-EBCDIC all of which will encode the code point with a different number of bytes.

But the contents of a system management guide are not updated.

a deadlock occurs, ASE activates trace flag 3605 for the deadlock report

but forgets to unset it after the report is finished, causing additional

subsequent output to be written in the error log.

stacktrace could happen in CgpUnion::_cgpInit().

target process which is being probed is blocked before entering new LAVA engine, it could

cause stack trace in _LeOpShowOp() if plan use worktables.

has non-clustered indexes is very slow.

when a subquery from a view/derived table is referenced in the

parent query block's WHERE clause.

executing a query involving more than 2 tables

with outer join on a non equality clause and

using GROUP BY.

if a table has multiple foreign key constraints. A stored procedure using this table will

not automatically be recompiled yielding a 3601 internal error.

when encrypted columns is not enabled and an attempt to grant permission

on encrypted columns commands is made.

be reported when making a large increase to the size of a cache using

sp_cacheconfig, despite sufficient memory being available.

Resolution: The Character Mapper component requires that Unicode values are

specified as <u{Unicode Decimal}> or <0x{Unicode Hexadecimal}> e.g. <u8256> ir<0x20A4>. The User Interface now automatically translates Unicode characters to the <u{Unicode Decimal}> notation.

cache to run this procedure, trigger, or SQL batch. Retry later. etc." mistakenly, although

one or more engines have local procedure cache available.

Note that care must be taken to reset the SET command prior from

returning from the stored procedure, otherwise the execution

of subsequent stored procedures may be affected. If it is

intended that subsequent stored procedures should be affected,

then the export_options feature should also be used, since

the intention is for a later change to localize the following

SET commands unless export_options is used (such as in a

login trigger).

The following SET commands were affected:

DISTINCT_SORTED

DISTINCT_SORTING

DISTINCT_HASHING

GROUP_SORTED

GROUP_HASHING

BUSHY_SPACE_SEARCH

PARALLEL_QUERY

ORDER_SORTING

NL_JOIN

MERGE_JOIN

HASH_JOIN

APPEND_UNION_ALL

MERGE_UNION_ALL

MERGE_UNION_DISINCT

HASH_UNION_DISTINCT

STORE_INDEX

INDEX_INTERSECTION

INDEX_UNION

MULTI_TABLE_STORE_IND

OPPORTUNISTIC_DISTINCT_VIEW

ADVANCED_AGGREGATION

REPLICATED_PARTITION

GROUP_INSERTING

ASE125_PRIMED

AUTO_QUERY_TUNING

QUERY_TUNING_MEM_LIMIT

QUERY_TUNING_TIME_LIMIT

SET PLAN OPTGOAL

after an 11xx exception has occured. The modules

'ex_reraise_lasterror' and 'ex_raise' will be reported in the

stack trace.

Resolution: To use Unicode characters in a calculation rule, the uUniChr() function must be used for specifying the Unicode character.

literal auto parameterization are enabled.

For example: select b2 from t1 where a1 = 'TEST'order by b2

There is no blank between a literal value surrounded by single/double quote marks and a SQL keyword.

In a replication environment using coordinated dumps, a dump tran with standby_access that was successfuly run on the primary may fail when RS attemps to apply it into the replicated database as it may not find a quiesce point in the transaction log of the replicated database.

vector aggregate operaor (GroupBy) or union operaor under a three-way or more

nested loop join (NaryNlJoin).

whose's ON clause contains a joining expression and the expression

involves a column which also participates in another equijoin predicate.

For example, (t1.c1=t2.c1 and t1.c2=t2.c1+5).

min() aggregate or OrderBy).

CR 438735 - ODBC Driver by Sybase: Unexpected behavior in Crystal Reports when connecting using ODBC Driver, because SQLProcedureColumns returns ?NO? for IS_NULLABLE.

CR 416248 - ODBC Driver by Sybase: SQLColumns returns type name varchar for timestamp data type.

CR 416250 - ODBC Driver by Sybase: SQLColumns returns type names varchar and char for nvarchar and nchar data types respectively.

CR 438423 - ODBC Driver by Sybase: Accessing ASE using ODBC Driver from Microsoft Access 2003 causes reserved error (-7707).

CR 420176 - OLE DB Provider by Sybase: Unable to execute queries from Microsoft SQL Server using a Sybase ASE linked server.

CR 427244 - jConnect: Calling DatabaseMetaData.getProcedureColumns when auto-commit mode is off raises SybSQLException: SELECT INTO command not allowed within multi-statement transaction.

CR 442470 - jConnect: DatabaseMetaData.getTypeInfo call no longer returns ASE Local type of "double precision".

CR 443327 - jConnect: DatabaseMetaData.getTypeInfo returns internal ASE system user types (like longsysname) which should not be returned.

as well as equi-sargs on joining columns, and descending index columns.

The failure occurs due to incorrect handling of composite orderings

when a descending attribute is in the composite ordering, and an

equi-sarg exists on another attribute of the composite ordering.

predicate and one OR term appears in all the disjuncts of the OR

predicate (in other words, the OR term can be extracted out as the

common predicate and the origial OR predicate can be removed). For

example, ((t1a.c1a = 9 AND t1a.c1a > t1b.c1a) OR t1a.c1a > t1b.c1a).

predicate and one OR term appears in all the disjuncts of the OR

predicate (in other words, the OR term can be extracted out as the

common predicate and the origial OR predicate can be removed). For

example, ((t1a.c1a = t4a.c3 AND 0 > t4a.c3) OR t1a.c1a = t4a.c3 )).

and the query has GROUP BY on an expression.

raised when a DISTINCT view is used and one of the view projection is

a constant.

allocated IDES buffer returned by uboalloc() from the rsysides_mempool: > (ind__sysides_grab+0x4c),

siginfo (code, address) = (1, 0x0000000000000028)

expression and the expression involves a column which also participates

in another equijoin predicate. For example, (t1.c1=t2.c1 and t1.c1=t2.c2+t1.c2).

containing hash and/or sort operators with large estimated input sets.

Buffer requirements increase as the operator's estimated input set size

does. This had a secondary impact on contiguous memory requirements in

the procedure cache. Contiguous memory requirements to manage hash

and sort operator buffers has been eliminated.

is not a grantable command, Adaptive Server fails to report Msg. 4607.

So as to output a result to HTML or Excel, please choose an export menu in gui mode.

The user wishes to have the management tool of a Web base.

For example: If the target column is defined as CHAR(10) and you have 5 characters

that you believe are 2 byte characters but you notice some are missing. (2 bytes * 5

characters = 10 bytes).

Resolution: UTF-8 can represent 1-3 byte characters. If some characters are 3 byte characters the total length calculations will need to take this into account.

Unable to execute project: Unable to execute component: DB Data Sink Update: Unable to write data to destination: Executing prepared statement failed. Unknown error during code page conversion.

Resolution: The source data, when interpreted as the specified encoding windows-932, contains code points which are not convertible to the eucjis encoding used by an IQ server configured with the EUC_JAPAN collation. To resolve this, use code points that are mapped in both encodings.

id was specified. scan_context={objid <objid>, indid <indid>, cid <cid>}. Contact Technical Support for Assistance." with 2

assertions prior to it may be reported when creating global nonclustered index on partitioned APL table whose first non-empty

data partitions is not its first data partition.

as the "unshrinking memory" feature. It would be an advantage for future ASE design to

implement "shrinking memory" function and global shared memory pool for multiple ASE

instances to even out their flunctuating workloads through optimal load balancing.

SELECT to operate on every every row in the result set rather than just the last. This

behavior is currently available using traceflag 7703, or by using UPDATE SET (although

UPDATE will fire the update trigger).

table is joined with a proxy table and the WHERE clause contains IN or OR clauses on

'date' or 'time' columns.

As the 'sa' user cannot login, steps cannot be taken to terminate users who are consuming

too many locks, or attempt to increase the "number of locks" configuration parameter.

Workaround: Some Grid Nodes are only allowed to be used with the packaged Development UI. Do not register them as they will not be able to run jobs outside the appropriate license restrictions.

IDENTITY_INSERT' is OFF." displays the wrong table name in some cases.

Similarly Message 585 "Explicit value must be specified for identity field in table '%.*s' when IDENTITY_INSERT or IDENTITY_UPDATE is ON." displays the wrong table name in some cases.

number of buffers that can be allocated per lava operator. It is

called "maximum buffers per lava operator". It is analagous to the

"number of sort buffers" parameter. However, it is applicable to

hash operators and sort operators used in Lava query plans. The

"number of sort buffers" parameter is now only applicable to sorts

used in "create index" processing. Because the number of buffers

allocated can have a second order effect on the amount of contiguous

memory required to track them, this parameter can be lowered to

alleviate "out-of-memory" (701) errors. The default is 2000.

reference encrypted columns with the decrypt default property can cause

ASE to stacktrace with a "Worker Process internal plan copy error for

EVAL E_DECRYPT_PERM" message.

table; column does not allow NULLS. UPDATE fails) could be raised

when there is a join predicate with an expression and each column

in the expression is involved in another equijoin predicate with

some other column.

is a scalar aggregate view/derived table.

and recreated and 2) the stored procedure or cached statement is re-executed the following error may be raised: "Msg 15424, State 4 "Adaptive Server cannot access the encryption key with object id '<obj_id>' in database '<database_name>'.

of the joining tables is a proxy table and the join predicate

involves an expression on columns from both the proxy table and some local

table.

It's the master database instead.

query optimization can be caused by inconsistent naming such as a mix of owner.object vs object references. Workaround by rewriting to use consistent naming.

across databases, the destination table incorrectly retains the

user type information from the source table/database.

symbol not built in referencing model

directly by using dump and load, and the endian type of the source and target platform is different.

Workaround: Don't use these data types. They will be removed as options in a future release.

printing the stats information. As the task is not closing the database after

reporting the required information it will be accumulated in the task PSS and

will exceed the limit on the number databases used by the statement.

Resolution: The column properties were corrected for this release.

Resolution: All column names are now quoted by default.

Workaround: This can be avoided by not using the 'template' option. If migration projects are created by hand, using the 'Add destination table from input...' option, the tables get closer to what they need to be and the user can fine tune the column types.

Resolution: Added secondary LOB file loading support to the DB Bulk load Sybase IQ component. To use this new functionality, add the following to the 'iq_loader' section of the runtime INI file (GridNode.ini | Default.ini):

use_secondary_LOB_files=true

locallobdir=[directory location to stage the LOB files, relative to the ETL runtime]

remotelobdir=[directory location to stage the LOB files, relative to the IQ runtime]

Resolution: The Template wizard now contains a new "Preserve schema/owner" check box for specifying the destination table name. If the check box is set then the original owner and table name is used. The user must verify that the specified destination schema exists. The Template wizard additionally contains a new option to store SQL scripts to be

executed before the destination data model will be created.

Resolution: The load script and column of data are delimited by TAB(0x09), records are delimited by LF(0x0a).

Resolution The column delimiter was changed to '|@#&' to be unique.

Resolution: The engine can be added manually and it will work successfully.

"equals to" clauses with expressions containing

explicit NULL values, parameters or local variables

and it is executed inside a stored procedure could

return incorrect results in the rare case that

ANSINULL option changes after the stored procedure

was compiled, i.e. procedure is not recompiled.

returned when executing a query with a join on an

expression which involves local variables and ANSINULL

option is OFF.

scope of the ON clause, in which one of the operands is an

expressions and the other operand is a single attribute of the

table which is the inner table of the outer join, then there

is a possibility of an incorrect answer.

An example would be the following

select * from t2 left join t1 on c2<c1 where c1=c2-1

go

option for database name , table name in command line mode.

query on a table created in the same procedure. Similary timestamp returned after calling tsequal()

is also incorrect.

sp_locklogin attempts to lock a login account, an HA consistency

error is reported.

the connection does not succeed and following error is logged in the errorlog :

kernel -> JS: Failed to connect on socket.

And hence, the Job Scheduler fails to come up.

subquery under OR and the OR predicate's transformation could render

the subquery part of the always-TRUE conjunct (being "redundant").

For example, (a=1) or ((a=1) and (subq)) could

be transformed into the equivalent (a=1) without the subq evaluation.

syntax error if the table or column name following 'order by'

or 'group by' is 'range','list','hash', or 'roundrobin'.

The workaround is adding '()' around the table/column name

following 'order by' or 'group by'.

not to be stored in the statement cache.

thread is on remote instance, then we were reporting it as blocked by itself.

This erroneous self-deadlock reporting is corrected in this CR. If pss is

blocked by remote thread, then now it as reported as 'Blocked by spid: (Remote spid - Unknown)'

underneath a CASE expression and an implicit CNVT node is needed

there.

calls.

then sometimes a non-equisarg would get chosen over an equisarg, which is

usually not the best choice.

select * from r,s where (r.a = s.a OR r.b = s.b) and r.a >= 5

In the above example, or general OR optimization, we have a choice

between using "r.a = s.a" vs. "r.a > 5" and in some cases we

chose "r.a > 5" for the index lookup on "r.a", which is usually

not correct.

reported in module 'memfreeheap' during the execution of an ALTER TABLE

to change the lock scheme of a table from 'allpages' to 'datarows/datapages'.

The modules 'altertab' and 'alt_lock_change' would also be reported in the

stack trace in the error log.

'pchk_pt_sched_scan' together with a stack trace including the module

'checkstorage' may be reported in the error log when DBCC CHECKSTORAGE is run

for a database using a device whose vdevno is greater than 127. This applies

to ASE version 12.5.* on the IBM AIX platform only.

statement is executed in non-passthrough mode.

(for example with isql -X) may see error message: "ks_falloc() memory allocation failed"

in the errorlog. The current and subsequent incoming connection requests to

the server using this encryption can fail. This message may also appear

if "net password encryption" is enabled for server-to-server communication.

sort order and character columns that appear in the select list of

a query also appear in a join clause (or a where clause acting as a

equi-join condition), the column data returned may come from

any table where the data is considered "equal" under the

rules case or accent-insensitive sort order in effect.

Upgrade item 1559 is not executed when downgraded 1502 server is upgraded again.

select * from r,s where (r.a = s.a OR r.b = s.b) and r.a >= 5

In the above example, for general OR optimization, we have a choice between using "r.a = s.a" vs. "r.a >= 5" and in some cases we chose "r.a >= 5" for the index lookup on "r.a", which is usually not correct.

select abc = identity(9), col1 into newtab from oldtab

is much slower in ASE 15 than in ASE 12.5 because an optimized algorithm to generate the new identity values was not implemented in ASE 15. With this CR, the algorithm was implemented, resulting in performance improvement up to 30%, depending upon platform and hardware.

when extended password encryption is used to encrypt passwords on the network.

variable and the same column is also involved an IN list/OR predicate.

The variable could come from the literal autoparam or a stored

procedure local variable.

2. 'select * from syslogins' gives stack trace on downgraded 1501 server, if statistics on syslogins is updated.

procedure 'sp_hacmpaccessvrfy' due to mismatch on hostnames for hostnames with trailing domain names.

errors, due to an untranslated key column in a index on a system table, after loading a 15.0.x

database dump across platform where the endian type is different.

the new environnment variable SYBASE_JRE_RTDS has been defined. That new variable

is required for JMS option to work properly since ase1502 esd1. Variable SYBASE_JRE

is still used by other Sybase components, but is not anymore used in the context of

RTDS. RTDS 4.5 requires the new variable to be defined in the ASE environnement and

to point to a valid JRE distribution.

-- AC13: Error encountered during execution of installddlgen script

The customer is requesting an enhancement for SySAM to provide the option to use two servers for the configuration.

This problem shall be fixed in PHSS_36345 and PHSS_36352 that are both scheduled to be released around the end of Sep/2007.

After Install Customer raised issue regarding permissions set by the Installshield

All dirs should be 755 or 750

All executables should be 755 or 750

All other files (datafiles / log / htlm / jar / language files,etc) should be 644 or 640.

Some files need special perms (eg ase_login_file, sybha) Some files need more restrictive perms (eg. files with passwords RUN_*) others need to be looked into, for example, data, jar or command text files have execute perms.

last log marker was not found for database '<name>'. The backward log

scan was started on marker (<value>, <value>)." may be reported when a

transaction dump that has been created with the STANDBY_ACCESS option is

loaded.

an OR predicate on an index column and the index column is

also participating in another equijoin predicate.

[SERVER_ENTRY]

master=NAMEPIPE,\\.\pipe\sybase\ase

query=NAMEPIPE,\\.\pipe\sybase\ase

master=NLWNSCK,server_host,5000

query=NLWNSCK,server_host,5000

In places where server entries were retrieved (such as Connect dialog), the above server entry would be skipped. Now the server entry will be shown ("server_host,5000"), with the named pipe service skipped.

(1) close a cursor implicitly during a cursor FETCH and after an

error was hit AND (2) dump expensive buffer allocation analysis . The first

behavior became critical due to another problem. The second

behavior was remapped to trace flag 9531

Resolution: The license management was changed to remove this option.

allocated." may be reported during the execution of a DELETE command when there is a

delete TRIGGER using the builtin DATALENGTH on a TEXT/IMAGE column.

column length mismatch error for DirectConnect/SDS servers.

when the configuration parameter "cpu accounting flush interval" or

"i/o accounting flush interval" is established with a small value such

that ASE needs to update syslogins table rather frequently.

return incorrect results if WHERE clause

references columns from one of the inner

outer joins.

CREATE VIEW v1 AS SELECT col1 = (SELECT name FROM t2 WHERE t2.id = t1.id) FROM t1

and this column appears on the left side of an IN-expression that contains a subquery on

it's right side, such as:

WHERE v1.col1 IN (SELECT ...)

Query execution may be aborted with a segmentation fault in the function, QpConstGetGc()

and a stack trace that contains GtqSingleSubqScan::GtInterestingTcIdBm().

- A view has an INSTEAD OF trigger for UPDATE or DELETE created on it,

- A cursor is declared to select a subset of columns of the view and no other database objects appear in the cursor's SELECT statement,

- The cursor's SELECT statement contains either DISTINCT or GROUP BY.

may produce a SIGBUS error. This is only likely to occur if the RESTRICT operator

evaluates an expression where one or more of it's operands are projected from

the UNION ALL operator and the operand requires alignment.

for a configurable period of time.

when the "show_abstract_plan" option is off.

all the databases looking for cross references to local encryption keys, even if no encryption keys existed in the affected database.

Resolution: Enabled Callback error handling to capture the error.

exceeding the row size limit. Such warning should be displayed only during procedure

creation not the procedure execution.

'Incorrect DBCC command: Please see the ASE Reference Manual for DBCC commands and options'

dbcc tune(maxwritedes)has been deprecated and replaced by a new sp-configure parameter: "i/o batch size".

dbcc tune(maxwritedes) needs to be removed from P&T guide

sig11 in ResRelOp() due to eager aggregation.

the cache (but it consumes the memory) if any of the user database is in 'single user mode'.

the COMPUTE clause of the query is referencing local variable or parameter.

Errorlog stack trace will show modules : 'match_compute' and 'paramchk'.

Since 15.x, we exploit available/useful orderings. So in 15.x we were not generating the "special OR strategy" when the IN list column has an useful ordering (as, for instance, when it is in an ORDER BY, GROUP BY, equi-join, etc.) unless the strategy did provide this ordering.

This fix relaxes the above constraint, by generating both:

- the index scan which provides the needed ordering but uses the IN list as a filter (and could be thus more expensive but could also be used by efficient parent operators which rely on the ordering);

- the "special OR strategy" which uses the IN list for index lookup but provides no ordering (and could be thus cheaper but would be used by less efficient parent operators or would need a sort to be used by the efficient ordering-needing parent operators).

Parent plans over both access methods are costed and the cheapest is chosen.

Turn on trace flag 15382 to disable this fix in the rare event that the cost-based decision proves to be suboptimal.

Transact-SQL extension of queries with grouped aggregates and columns in the

select list which are not in the GROUP BY clause.

is an expression join WHERE clause whose expression is referencing

a column from the right nested outerjoin inner table.

is still slower than 12.5.4.

set commands (such as "set forceplan on"). In particular, within a batch

or stored procedure, these new commands should be able to affect

optimization.

is effectively present on that queue. This occures when an other ASE spid having previously

already consumed a message from the same queue is waiting for a message from an other Queue

located on the same JMS Provider (same URL). That issue is only TIBCO JMS relative (EAS and

IBM MQ are not concerned).

error will occur.

Workaround: Colums of n-type are not accessible via OCS.

Workaround: Update the ODBC driver, including the ODBC stored procedures on the target ASE database, with the latest Sybase SDK ESD. Recommended minimum ESD level is ESD9.

a pre-15.0.2 ASE release, therefore error 1204 "ASE has run out of LOCKS ..."

may be reported.

Workaround: Applying SDK15 ESD #9 for the ODBC driver helps eliminate some but not all of the errors. This ODBC driver issue of returning the Catalog names if the catalog name passed in is "%" <ALL catalogs> is an enhancement request.

Current process (0x1b001b) infected with signal 11 (SIGSEGV)

The changes done in resolution 1 is backed out in resolution 2.

You have a table with a null column:

create table ntab (a int null)

insert into ntab values (null)

This parameterized query may or may not return the correct answer:

@p1 = 3

@p2 = null

select * from ntab where a + @p1 = @p2

The correct behavior is that this query should return one row. The "a + @p1" should evaluate to null, and the Sybase null matching

extension should cause the "=" to compare true against @p2. We see this correct answer with 12.5 and also whenever this

query is executed inside a stored procedure. However this comparison fails when the statement is sent as a ct-library program.

across platform, after the server is restarted.

and remote tables if the join clause has nullable datatypes, the proxy table is on the inside of

a nested loop join, and a datatype conversion is required.

is an OR predicate in the WHERE clause and one

OR term is <col>=NULL and <col> comes from the

outerjoin inner table.

local temporary database, that role cannot be completely dropped from another node.

Attempting to drop the role, results in error:

11178/16/1 "Role '%.*s' cannot be dropped because references to the role exist in one or more local temporary databases. In ASE Cluster, no privileges should be granted to the role in any local temporary database. This role can be dropped after a cluster reboot."

Workaround: Columns of n-type are not accessible via Open Client.

Workaround: SQL also uses square brackets to quote identifiers. Any variable defined by the user is replaced by its value in such a statement. The solution is to disable evaluation of the statement (right-click, remove check mark), with the drawback that dynamically generated statements are no longer possible. This is not the case for the index load component, where the [index] placeholder is treated specially by the GRID engine.

<col> and the <col> is also involved in another expression equijoin

join (<col> = <expr>).

between an outer join and other two or more tables

could return incorrect results if the plan chosen

to execute the query uses the Nary Nested Loop Join

mechanism.

statements may fail with error 225 (text) if the table is not qualified

with exactly the same prefix and name as in the CREATE TABLE statement.

Msg 969, Level 14, State 4:

Line 1:

You can access database 'local_tempdb1' only from its owner instance

'myserver1'. You cannot access local temporary databases from non-owner

instances except to use CREATE DATABASE and DROP DATABASE with local system

temporary databases.

Msg 15424, Level 16, State 3:

Line 1:

Adaptive Server cannot access the encryption key with object id '2036302387' in

database 'local_tempdb1'.

Similarly, the command 'sp_encryption help, <key>, display_cols' will fail in a multi-node cluster with the following error:

Msg 969, Level 14, State 4:

Procedure 'sp_encryption', Line 277:

You can access database 'ftdlinux10_3477_tdb_2' only from its owner instance

'ftdlinux10_3477_ns2'. You cannot access local temporary databases from

non-owner instances except to use CREATE DATABASE and DROP DATABASE with local

system temporary databases.

(return status = -4)

temporary database and the database is not accessible on the instance,

the following messages appear:

Msg 969, Level 14, State 4:

You can access database 'yellowsta0_13695_tdb_1' only from its owner instance

'yellowsta0_13695_ns1'. You cannot access local temporary databases from

non-owner instances except to use CREATE DATABASE and DROP DATABASE with local

system temporary databases.

Msg 4001, Level 11, State 1:

Cannot open default database 'yellowsta0_13695_tdb_1'.

after error 1105 is raised. The modules "dol_insert", "pg_getdatapage_forinsert"

and "ex_reraise_lasterror" will be shown in the stack trace.

select rand2(column1) from table1

performed on the database will be raised if the SELECT INTO EXISTING TABLE syntax

is used even if the SELECT INTO / BULK COPY dboption is turned off. The SELECT

INTO EXISTING syntax was meant for internal use and not intended to be documented

or used by end-users.

errors. This may happen if during the execution of the statement some serious error arrives.

Errorlog file will include stack trace dump with modules dbgrpc_control, sqldbg__execsql and ubffree.

Msg 969, Level 14, State 4:

Line 1:

You can access database 'local_tempdb1' only from its owner instance

'myserver1'. You cannot access local temporary databases from non-owner

instances except to use CREATE DATABASE and DROP DATABASE with local system

temporary databases.

Msg 436, Level 16, State 2:

Line 1:

The 'insert' statement will not be executed because the cross-database

referential constraint on table 'my_table' cannot be enforced. A table or

constraint definition is missing in database 'local_tempdb1'. Contact your

System Administrator.

Tables should not be allowed to reference data in the local tempdb of a cluster instance.

create index <name> on <table> [ column list ] LOCAL UNIQUE INDEX

instead of:

sp_chgattribute "table_name", "local_unique_index", 1

create unique index index <name> on <table> [ column list ] LOCAL INDEX

to create a local unique index.

2. Fix the issue of "select distinct but get duplicated result".

3. Old syntax is also supported.

4. sp_helpindex changes:

Printing out "Local Unique Index" for index_local field, instead of "Local Index" for index_local column and "unique" for index_description column.

is used inner to the outerjoin, such as a vector aggregate

view.

the view may result in timeslice or 701 error after an index on the table has been dropped

and re-created.

aud_getsdes() with a stack trace reported in the error log when auditing is active.

The modules " rvm__audit_table_view_access" and "aud_tabview_ref" will be also reported

in the stack trace.

table accesses is dropped and recreated.

When a new listener is started with SSL protocol after "sp_ssladmin setcipher" has changed SSL cipher suite preferences, only listeners started on the local instance will use the new cipher suite preferences. Listeners started on other instances will not see the new preferences until after cluster restart.

is set, the password is not changed on all concurrent connections on other

instances in the cluster. This can result in authentication failures when concurrent

connections attempt to establish external connections, use proxy tables, or remote

procedure calls.

D00: SYBMULTBUF ERROR: Wait for i/o completion failed: State=zzz, Msg=Attempted to transfer xxxx bytes, but yyyyy bytes were transferred by operating system call read/write.

Backup Server: 4.145.2.22: [zzz] Error for device '<device>'. Attempted to transfer xxxx bytes, but yyyyy bytes were transferred by operating system call read/write.

defined java objects which internally run nested queries

using the internal JDBC driver, some errors like 622 error,

"Opentable was passed a varno of <id>. Object '<object_name>'

in database '<dbname>' already has that session descriptor

in use." or 6103 error, "Unable to do cleanup for the killed

process" may be reported in the errorlog.

Workaround: Colums of n-type are not accessible via Open Client.

- A local table has a column that has a foreign key referential integrity constraint to a primary key column in a proxy table,

- An INSERT or UPDATE query adds or modifies the foreign key column in the local table,

- The result of the INSERT or UPDATE causes a integrity constraint violation,

a signal 11 will be raised when the query is rolled back.

- Rank functions

- Distribution functions

- Inverse distribution functions (percentiles)

- Hypothetical set functions

- Cumulative and other forms of moving aggregates

in the module 'LeAPScanContext::LeRequiresMDAUpdate' together with a

stack trace which includes the modules 'LeScanContext::updateMDAStats'

and 'exec_lava' may be reported in the error log if the configuration

option 'enable monitoring' is changed from 0 to 1 while a query

is executing concurrently.

full-passthrough mode. These statements were being executed in non-passthrough mode.

created in the same procedure having an IDENTITY column

could wrongly increment the IDENTITY column even though

it is not affected by the UPDATE.

The following XML request that was generated by Microsoft IIS will not work with our product:

<?xml version="1.0" encoding="utf-8"?>

<soap:Envelope xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xsd="http://www.w3.org/2001/XMLSchema" xmlns:soapenc="http://schemas.xmlsoap.org/soap/encoding/" xmlns:tns="http://localhost/testservices" xmlns:types="http://localhost/testservices/encodedTypes" xmlns:soap="http://schemas.xmlsoap.org/soap/envelope/">

<soap:Body soap:encodingStyle="http://schemas.xmlsoap.org/soap/encoding/">

<tns:Echo1 />

</soap:Body>

</soap:Envelope>

Customer would like all MS IIS generated requests to be supported.

in page 177 through 179 of ASE 15.0 System Administration Guide Volume 1'.

with the following error if ASE sends it LONGCHAR data:

DO NOT SUPPORT CS_LONGCHAR_TYPE

in the ASE log file:

SSL or Crypto Error Message: 'Error reading the private key. The specified buffer

content is not in the expected format.'.

system table before saving the SQL text of the procedure, view or trigger to syscomments.

End comment mark (*/) may not be added to syscomments on non-English environment.

"blank", ":", "/", "\".

Resolution: Change the value you entered in box "name" of the "windows schedule" section. Replace characters like ":","/" with "_" or remove the blank in the name value.

Resolution: ETL license files were updated to support 'SQL Anywhere', version '10.00.0000'.

and the aggregate is on some column from outerjoin inner table and there

is also an equijoin predicate between that column and some column from

outerjoin outer table. This is related to eager aggregate feature.

it was set for the given number of days.

table that has been renamed, it runs with the old plan that references the old

object name.

will be cached. This does not make the best use of the statement cache and literal autoparameters.

will report in the stack trace modules like LeRun,ptngetrowcount,open_nextvarno.

in the SELECT list is referenced at another outerjoin view inner

side and the outerjoin view is later degenerated into an inner

join when being queried.

at least one of the non-decrypt_default columns and withheld from the column with the

decrypt_default value, then any select of the column with the decrypt_default will

correctly return the decrypt default value and also return (incorrectly) Msg 10332.

e.g. sp_tempdb "bind", "ap", "ABC%", "DB", "tempdb" where '%' is a wildcard character

to char type, error 530 will be raised.

The default unit of disk init is not MB.

disk init

size

...

If you do not provide a unit specifier, the value provided is presumed to be in megabytes.

password. This resolution fixes the problem to update the password in

ha_companion.<servername> file when sa(or user in ha_companion file) password is

changed.

Workaround: ETL currently has the restriction that you need to connect to the repository using the same connection type with which the connection was created. If it was created with ODBC, connect with ODBC. If it was created ctlib, connect with ctlib.

held if multiple caches are configured.

.SQL file. This problem was very rare, timing-dependent, and more likely to

occur on faster machines. This has been fixed.

-- [QTS-479086 (OTW)] Previously, when importing data into an existing

table, the "Preview" table in the Import Wizard would show values

in BINARY columns as "[B@" followed by up to 8 numbers and letters.

Now, the value from the file is displayed.

-- [QTS-476693] Previously, you could not use the INPUT statement

to populate a GLOBAL TEMPORARY table that did not preserve

rows on commit. The INPUT statement had the unintentional side-

effect of commiting changes. This has been fixed; INPUT no longer

commits changes.

-- [QTS-475259] Previously, if you selected the "@@rowcount"

variable, it was always displayed "1". Now, the actual value

is displayed.

-- If DBISQL is configured to display result sets using text,

and you execute a statement which causes a JConnect "JZ0P9"

error, that error was not being reported. Now it is.

-- If DBISQL is configured to display results in grid mode,

and you execute a statement in a batch which throws an

unrecoverable error, you used to get a "JZ0CU error" message,

but the text of the underlying error was not displayed.

Now, you see the text of the underlying error, and the

"JZ0CU" text is not not displayed.

To reproduce this:

1. Connect to an ASE server using a non-DBA account

2. Execute: use sybsystemproces

2. Execute: sp_hidetext sp_jdbc_getcatalogs

You should see an error message which reads "Could not

execute statement. Object does not exist in this

database."

-- If DBISQL is configured to display results in text mode,

and you execute a statement in a batch which throws a

recoverable error, you used to get a "010P4 error" message,

but the underlying cause of the error was not displayed.

Now the cause is displayed, and you don't see the cryptic

"010P4" message.

To reproduce this, connect to an ASE server using a non-DBA

account, and execute the following as a single batch:

exec sp_helpdb

select * from sysobjects

-- Previously, if connected to a slow server, double-clicking

the database drop-down in the toolbar could crash DBISQL.

This has been fixed.

-- [DBISQL-054] Previously, when executing statements that returned

one or more large result sets, DBISQL could become unresponsive

or could exhibit undefined behavior (e.g. it could close the

database connection, or be unable to stop the statement).

This has been fixed.

-- [DBISQL-054] Previously, if you tried to stop a statement which

returned a large result set, DBISQL could enter a state where

it was unable to execute any further statements this has been fixed.

-- Previously, when executing a batch of statements, DBISQL would

ignore any results returned from statements that followed the

first statement in the batch that caused an error. This has

been fixed for both "text" and "grid" display modes.

-- Previously, when displaying results in "text" mode, it was

possible that DBISQL would not display the results at all due

to an internal race condition. This has been fixed.

-- Previously, when DBISQL executed a batch of statements, it

would not report an error raised by one of the statements if

the preceding statement raised a warning. This has been fixed.

This bug had been reported by JPMC as "dbisql-056".

-- Previously, error messages reported via jConnect did not

include the severity level, (ASE) state, or the transaction

state. Now they do. Also, if the JDBC driver state is "ZZZZZ"

(meaning "no information"), it is now not displayed at all.

This bug had been reported by JPMC as "dbisql-033".

-- OEMs can now disable the "SQL/Execute" & F5 accelerator key by

adding the following line to the ".isqlPreferences9" file:

disableExecuteAll=true

This feature was demanded by J.P. Morgan-Chase.

-- [QTS-472617 (OTW)] Previously, the ISQL window could stop painting

when executing a long-running query. This has been fixed.

workspace are configured to the same object name value. The stored

procedure sp_dbcc_configreport allows the same object name value to

be used for both mistakenly.

will mistakenly result in error 12901 being reported "This command is not allowed on

an offline database '<database_name>'."

being killed when an UPDATE command is run in the 'deferred_varcol' update mode as shown by SET SHOWPLAN and a data page split occurs. The modules

'xact_endupdate', 'def_insert' and 'linksplit' will also be reported

in the stack trace found in the error log.

incorporated into ASE to improve query performance on queries involving Simplified Chinese (GB)

characters. This essentially replaces the default string comparison processing which uses the

Unilib APIs. Streamlining the Unilib comparison removes the processing overheads hence improves

performance. The new implementation supports the existing GBPINYIN and GBPINYINNOCS sortorders

for UTF-8, CP936, GB18030 and EUCGB charsets. A boot-time Traceflag 2411 is available to revert

back to the default behaviour.

Workaround: ETL currently has the restriction that you need to connect to the repository using the same connection type with which it was created. If it was created with ODBC, connect with ODBC. If it was created with ctlib, connect with ctlib.

Workaround: ETL currently has the restriction that you need to connect to the repository using the same connection type with which it was created. If it was created with ODBC, connect with ODBC. If it was created with ctlib, connect with ctlib.

the 'display'/'report' action, the error message "Duplicate

position specifier in format string of PRINT command" is

raised.

and fails with errors 10317 and 17720.

Server terminated abnormally or DUMP DATABASE was done at a time when a

(fast) bcp command was being used to import data into that table.

Workaround: Set the Database option "Auto Commit" to "1" will insure the transaction is committed.

command. When dump database command is in progress in one of the other instances,

startup or shutdown of instance waits until the dump database command completes.

maybe incorrectly reported, resulting in the command

being aborted when running a INSERT SELECT statement,

the target table has a check constraint involving more

than a single column and a NULL value from source table

was being inserted into the target table.

current session option settings.

'print deadlock information' is active has been enhanced such

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

cause a stacktrace.

sp_configure "optimization goal", 0, "allrows_oltp"

can disable replication agent threads on some platforms.

1. referencing a materialized outerjoin view (worktable), and

2. at the same time another view is merged into the parent outerjoin query

as the inner joining table, and

3. the SELECT list is referencing some column from that merged view, and

4. that column from the merged view is an CASE expression.

clusters compared to single node.

table may be reported by SP_LOCK, despite them being compatible. The same may be seen

with 'Sh_intent' locks.

Added a "Cancel Plan" button in planviewer.

Resolution: Download Adobe Flash Player from the www.adobe.com web page (http://www.adobe.com/shockwave/download/download.cgi?P1_Prod_Version=ShockwaveFlash)

FYI: User is likely to be prompted by Windows Vista with this information - after shutting down ETL Development - if they have attempted to run the flash demos during their ETL Development session.

which might cause differ query behavior.

with other aggregates in a query containing a group-by clause, the process can be terminated

prematurely.

can be improved by implementing a new operator to echo a row, without

any cache set up required.

index of parent table when performing materialization for view/derived table involving

outer join.

str_replace is documented under string functions, but does not show up in the list of strings functions in the above url.

The flag '-P' is not supported by Backup Server.

Unable to parse '/opt/sybase/ASE-12_5/install/RUN_r1254_back'

sp_passwordpolicy 'expire role password' raises error 17678, "Illegal role name '%1!' specified." even if the role exists.

This only affects logins/roles having upper case characters.

one server certificate to be added so that if the current one expires it can move to the

next.

while its transaction log is undergoing truncation triggered either by an explicit

'DUMP TRANSACTION' command or by the 'CHECKPOINT SLEEP' process.

out-of-date fairly soon after update statistics is completed, this

feature will assign some selectivity to out-of-range search clause

accesses. This will only be done for search clauses that select values

that are greater than all values in the domain described by the histogram.

In the "set option show_lio_costing on" tracing, there will be messages

associated with the selectivity of the column that indicate an

"Out of range histogram adjustment" has been made. The selectivity

assigned to each out of range search clause will be slightly less than the

selectivity of one domain value i.e. if you have 100 unique values in the

column, then 1/(100+1) is assigned to the out of range search clause, and

then weights are stolen from the remainder of the cells in the histograms,

including the newly created cell, so that all the cells in the

normalized histogram still add up to 1.0. Thus, the actual weight in this

example will be (1/101)/(1.0 + 1/101) after normalization to 1.0

The feature can be turned off by trace flag 15355.

I am able to see non-zero values for I/O busy in sysmon report with 15.0.2 ESD#2 binary. Hence it is not the case that always I/O busy value is reported as '0'. Also the value of "I/O ticks", the counter used by sp_sysmon is incremented only if the CPU has no task to run(i.e. it is idle) and some I/O is pending. If there are tasks to be run then only

"cpu ticks" counter is incremented. Hence it is not a bug if the I/O busy value is '0' unless the customer observes any performance drop. Hence closing the CR as not a bug.

datachange across all histogramed columns) could

result in a SIGSEGV or SIGBUS error. This could occur if there

are a lot of columns in the table with histogram statistics.

The problem is more likely to occur if no partition id (check

the column's datachange across all partitions) is provided and

the table has many partitions.

first branch of the nary nlj, may return too few rows.

create index "AA_KK" on "AA" ("last_modified")

and before you set quoted_identifier on then the index, when looking at it should display AA_KK .

single-byte non-binary servers or in utf8 binary servers in allrows_oltp mode.

appear inserted data is missing.

For example: If target column is defined as CHAR(10) and you have 5 characters

that you believe are 2 byte characters and some are missing. (2 bytes * 5

characters = 10 bytes).

Resolution: UTF-8 can represent 1-3 byte characters. If some characters are 3

byte characters the total length calculations will need to take this into

account.

Object 'table_name' in database 'db_name' already has that session descriptor in use"

during recompilation.

Resolution: This is a known issue. The source of the problem seems to lie in either the Vista ODBC driver manager or Vista memory management. If possible, use native database connectivity instead of ODBC connectivity.

query when either the ON clause predicate involves correlated subquery

or there is a nested join inner to the outerjoin.

messages when the regeneration of a keypair fails.

with different errors related to the sort operation. Sometimes the stack trace in the

errorlog may indicate the module startscan.

on different tables then sp_helpindex on a specific table will report wrong index type i.e.

global index as local index.

'dfl_dmp_initflushlist' or 5205 error may be returned during the execution of a DUMP DATABASE command,

The modules 'cdfl_dmp_initflushlist' and 'dpdb_main' would also be reported in the stack trace in

the error log.

of the runtime compilation.

"GW_TOKEN, gwSession: <session id>, vs_streamGetToken called too many time for stream <stream id>, key <doc key>". Looping is caused by repeated calls to the stream call back function.

"Warning: command 'addcert' adds certificate path <certifcate path> with its common

name field different from server name."

seen by the presence of a RIDJoin operator in the showplan output), the updates to the qualifying rows may be done in deferred mode when direct mode is possible (also seen in the

showplan output). This could result in performance degradation in highly concurrent systems.

On a very busy server with a busy tempdb it can deadlock or cause

contention for other tempdb users. So Customer would like the

ability for sp_object_stats to create tables in a user defined

tempdb instead.

for the Sybsecurity..Sysaudits tables.

due to view materialization.

stored procedures.

in the statement cache. This is not currently expected or supported.

which uses execute immediate for creating other stored procedures

and had to be recompiled maybe be aborted reporting in the

errorlog the message "current process infected with 11" in

the module 'set_cast' including a stack trace which includes

the modules 's_renormalize' and 's_recompile'.

The costing of EXISTS flattened queries or implicit EXISTS queries

(as was the case in this query, where the query was converted into

and EXISTS query) was overly optimistic in cases where more than

one table participates in the existence joins. In particular,

the fast first rows costing used for EXISTS queries, underestimated

the number of outer rows for some of the nested loop join operators,

and underestimated some of the scan costs for the inner tables/indices

of a nested loop join involved in a EXISTS (or semi-join). The

change puts some lower bounds on the fast first row costing so that

the outer row counts are consistent with the overall cardinality estimate

of the semi-join group (a semi join group is a set of 2 or more tables

that together are transformed into a multi-table semi-join in the

show plan output).

Also fixed a possible hostname comparison failure when one hostname obtained

is short form of the hostname, and another is a fully qualified hostname with

tailing domain name.

could raise Assertion (_SrtTcIdBm.VbmDisjoint(*pConstSargs)) failure. Also

one further SELECT DISTINCT on such DISTINCT view could lead to some stacktrace during

code generation.

This issue is reproduced by ASE SmallBusinessEdition and Developer Edition.

Resolution: This problem is a result of improperly configured ODBC.ini configuration file.

The CommLinks property needs specific syntax:

Incorrect Property syntax:

CommLinks=tcpip(host=host-machine-name,port=4454)

There are 2 choices for Correct Property syntax:

first: CommLinks=tcpip(host=host-machine-name;port=4454)

second: CommLinks=tcpip(host=host-machine-name,4454)

a particular optgoal is set through the abstract plan "plan" clause

versus at the session or server level.

For example: If the target column is defined as CHAR(10) and you have 5 characters

that you believe are 2 byte characters but you notice some are missing. (2 bytes * 5

characters = 10 bytes).

Resolution: UTF-8 can represent 1-3 byte characters. If some characters are 3 byte characters the total length calculations will need to take this into account.

Workaround: The installation directory can be manually changed to another directory if desired.

SELECT * FROM mytable

would be editable, but the result set generated by:

SELECT * FROM dba.mytable

was not.)

Previously, when a statement returns more than 10 result sets, DBISQL displayed only the first 10 silently in grid mode. Now DBISQL will display a dialog with the following message:

"The statement returned more than 10 results set, but only the first 10 have been displayed."

of the query plans.

properly initialized after the command 'LOAD DATABASE' has been executed.

Workaround: Upgrading from 4.2 to 4.2 ESD #1 or from 4.2.1 to 4.2.1 ESD #2 changes the location of the default local MS Access ETL repository file. The ETL 4.2 uses a local MS Access repository (ETLDEMO_REP.mdb)in the Windows users ..\Documents and Settings\<USER> directory. After upgrading to the ESD the user must copy this ETLDEMO_REP.mdb file to the ESD installation\Demodata directory to access previously stored jobs and projects.

contains the following lines:

| |SCAN Operator (VA = 0)

| | FROM TABLE

| | rv3

| | [ Using Dynamic Partition Elimination ]

| | Table Scan.

| | Forward Scan.

If the query plan is re-executed (i.e. it's in a stored procdedure), the showplan output

for the second and subsequent executions will show extraneous lines of output beginning

with "{ Eliminated Partitons", such as:

| |SCAN Operator (VA = 0)

| | FROM TABLE

| | rv3

| | [ Eliminated Partitions : 1 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 ]

| | [ Eliminated Partitions : 1 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 [ Using Dynamic Partition

Elimination ]

| | Table Scan.

| | Forward Scan.

a declarative or bound default, the bcp-in operation may fail. The ASE

error log may contain a stacktrace containing: infected with signal 11 in col_encrypt.

Unfotunatly this error is not hit until a large amount of processing has already taken place on a dump and a large amount of time therefore gets wasted just to find out a dump is useless.

Customer would like an enhancement to the dump command to make sure that the dump produced will load correctly under XPDL without giving a 3151 (or any other) error.

Customer would also like a corresponding enhancement to load database to make sure that a load can be succesfully loaded via XPDL without having to go through the lengthy processing to find this out.

Customer is currently using 12.5.4 and would like these enhancements in the 12.5.x codeline.

the recursive call of SeHyPermStrat::_SeHyPsPrime() when stack size

is not large enough.

source text is executed after database is loaded across platform where the byte order(endian type)

is different.

- A cursor is declared with a SELECT statement that contains either an IN-list or OR-clauses on the same column, such as

DECLARE curs1 CURSOR FOR SELECT id FROM sysobjects WHERE id IN (1,2,3); and

- The optimizer chooses a special OR strategy for the query plan, as can be seen in the showplan output by the presence of 'FROM OR List'; and

- The cursor is OPEN'd and FETCH'd at least once; and

- The row that the cursor is currently positioned on is deleted, either through a searched DELETE command or a searched UPDATE command that causes the row to move.

A 582 error will be raised if another FETCH command is issued on the cursor.

index id <value> and there is no entry in the OAM for this page." when running DBCC CHECKALLOC,

TABLEALLOC, INDEXALLOC or TEXTALLOC with the FIX option. This error is seen when OAM entries

with zero used/unused page counts have been deleted.

involves a local variable, did not perform partition elimination when the statement is executed

inside execute-immediate.

With this fix, partition elimination is enabled, resulting in increased performance.

When using raw devices, the quorum device and data devices do not need to be removed.

1 Stop the srvbuildres or dataserver utilities, if either are running.

2 Stop the UAF agents on all nodes.

3 Remove the directory with the name of the cluster you tried to create

during the wizard session in the

$SYBASE_UA/nodes/node_name/plugins/cluster_name directory.

4 Remove the interfaces file entries for the cluster you tried to create.

5 Restart the UAF agents on all nodes.

needed for optimization of simple queries.

The Assertion (!_pGcojLegacyRgInfo || pOjRgInfo == _pGcojLegacyRgInfo ||

pGt->GtIsValues()) might fail for the diagserver. For non-diagserver,

search engine could generate wrong outerjoin plan or gets into

excessive compilation time and memory consumption for large join.

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

and "des_idtnewval" when an INSERT command attempts to create a new identity column value.

experience excessive procedure cache usage, that could eventually result

in 701 errors. This change allows more efficient

reuse of memory associated with histograms.

configuration parameters has been set:

metrics elap max

metrics exec max

metrics lio max

metrics pio max

However, if there is any among the above 4 thresholds set, only the non-zero values will be effective

the thresholds for the QP Metrics.

inconsistency will lead to Signal-11 for further reference to the system catalog.

subquery under an OR predicate and during the actual execution the

subquery evaluation is mostly skipped due to shortcircuit logics of

the OR predicate.

Stream/BLOB interface.

the statement may fail with error 15432, 15467 or 15469 if all the following conditions

hold:

- The statement is executed in the context of a nested stored procedure.

- The nested procedure exists in a different database from the parent procedure.

- The parent stored procedure has been created by a user with sa_role.

- The parent procedure is invoked by a user who does not have sa_role and is not the dbo of the nested procedure's database.

> kernel Current process (0x180018) infected with signal 11 (SIGSEGV)

> kernel Error at address 0x4000000000db6c33 (pg_fill_supergam+0xfb)

1) table <tablename> has an encrypted column c1 with decrypt_default;

2) alter table <tablename> modify c1 ... decrypt_default <default value>;

3) sp_help <tablename>.

Reported: Adaptive Server Enterprise/15.0.2/EBF 14331/P/Linux Intel/Linux 2.4.21-47.ELsmpi686/ase1502/2486/32-bit/FBO/Thu May 24 08:15:50 2007

Reproduced on same platform/ build and Adaptive Server Enterprise/15.0.2/EBF 14332/P/NT (IX86)/Windows 2000/ase1502/2486/32-bit/OPT/Thu May 24 04:10:36 2007

Not Reproducible on Adaptive Server Enterprise/12.5.4/EBF 13387/P/NT (IX86)/OS 4.0/ase1254/2006/32-bit/OPT/Sat May 20 00:54:28 2006

Create a database create a table and a view. Dump database then load and on-line using standby_access. try to access view.. Error:

Msg 225, Level 16, State 1:

Server 'V1502B32', Line 1:

Cannot run query--referenced object (name NAME NOT RECOVERABLE) dropped during

query optimization.

If database in on-lined normally then no problem accessing the view.

may not be able to decode and print the system catalog rows.

The message "Could not locate SYSCOL array for catalog

<catalog_ID> (<name>) at log version <log_version_ID>" is

printed instead.

may return less rows than expected if MERGE JOIN is chosen. It could happen

if one equijoin column is part of the index keys and the same column

has an equisarg (col=<constant>) in the derived table/view definition.

a segmentation fault in the function, s_execute().

Customer would like to be able to have an option to be able to atomaticly open a new

window in the session that only has the result set just like SQL Advantage did.

database '<value>'. Aborting the transaction." may be reported by the HK GC system task

while trying to garbage collect a page that is in the process of being deallocated.

correctly for scans beneath XCHG operators.

Adaptive Server accepts, but does not require, the disk init vdevno (virtual disk number) parameter. If you specify a vdevno, you may choose any currently unused identifier from 1 to 2,147,483,647 (virtual device ID 0 is used by the master device).

In sp_sysmon, Result space is given three characters width only.

When the given value for vdevno > 999, value is exceeding three character width hence causing the error.

partition elimination. When runtime partition elimination will be attempted, the

showplan_in_xml output will contain:

<DynamicPartitionIdentification>Yes</DynamicPartitionIdentification>

and when runtime partition elimination will not be attempted, the showplan_in_axml output

will contain:

<DynamicPartitionIdentification>No</DynamicPartitionIdentification>

"Msg 1504 Internal Sort Error. Empty Index Page found.". This may happen if the optimizer

decides to use the Dynamic Index as an inner table in the join and the parallel sort

is applied to the Dynamic Index worktable.

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

and "drop_with_retry" following error 3702 "Cannot drop the table '<n>' because

it is currently in use." when a session tries to drop a temporary table while

the system task HK CHORES is flushing statistics for the given table in systabstats.

also contains another correlated subquery.

the diagserver will shutdown due to an error condition during decryption which will cause

an assertion to fire due to an invalid length.

dbcc checkdb command should lock one table at a time and release the lock after it finishes checking that table.

as SELECT ... (SELECT COUNT(DISTINCT col) FROM tab)... can lead to a segmentation fault in

the function a_remapqblockvar.

under the default as allrows_mix plan.

As a result future cache bind operations, requiring single user mode operation, may see 924 errors

on other nodes.

a column not in the SELECT list, infected with 11 may occur in OptGlobal::GblSetInterestingTcIds().

on allocation <pageno2> has objid <objid>, indid <indid>, and used bit on,

but reference bit off", may be reported by DBCC CHECKALLOC if a sort was

active on an instance just prior to the node being shut down abnormally.

for 100011 errors than just Dropping/recreating the table/index or loading

from clean backups. There are other alternative fixes for text pointer error

100011, such as running dbcc textalloc, sp_rebuild_text, or dbcc rebuild_text.

of the following roles in order to perform this operation: 'sa_role,

oper_role'" may be reported when a database owner who does not have sa_role executes load with listonly or headeronly command.

during the execution of a DUMP DATABASE command or the DUMP command may hang, when non-logged

operation, like reorg rebuild, is heavily executed on the remote node during the DUMP command.

The modules 'dn_rdfl_add' and 'dn_rdfl_handler' would also be reported in the stack trace in the

error log.

binding to a named cache of an index whose index id value is zero,

resulting in overwriting the table's binding if any.

will be raised when trying to dump a database that is in single user mode.

wrong results and/or assertion failure may happen

administrative tools to concurrently reconfigure the cluster, overwritting each others changes.

restricts the upperlimit of procedure group number to 32767 ( the max value of 'smallint')

around in the old cache. subsequent requests for physical locks from another cache raise an 891

error if the buffer in the old cache is found to be dirty.

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

the modules "dmap_merge_diskmaps" and "db__sysaltusages_map", when the

DBCC CHECKSTORAGE command is run on an archive database.

cfg_syscacheinfo will be raised when executing the stored procedure sp_configure

if at boot time one of the options is not present in sysconfigures causing

error 5864, 'Invalid option encountered '.

might hit div-by-zero execution error or return wrong result when

there is an equijoin under the aggregate and the join column has an

equality sarg on it (such as t1.c1=t2.c2 and t2.c2=5).

exceeds the maximum allowable limit of xxx." may appear if the query has 6 or more joining tables and GROUP BY clause.

references a column of that view in a predicate in a CASE or NULLIF expression and that

same view-column is referenced in at least one additional predicate, a segmentation fault

can occur during excution in the function, CgEvals::copyEvalsToVirCol.

This will cause the login name to remain in sysobjects even after the login is dropped.

such as:

WHERE nullif(nullif(4, case d2.d2a when 4 then d2.d2a else 4 end), v1a.v1b) = v1a.v1a

and also has the column being joined to (v1a.v1a in this example) under at least one

additional OR predicate, a segmentation fault (SIGSEGV) can occur during execution within

the function, LeRun().

may consume more system time on machines with limited resources and running alongside other processes besides the server.

"Internal error occurred while attempting to perform post-commit cleanup".

minimum row length of xxx." together with a stack trace which includes the modules 'collocate',

'makeindrow' and 'sobld_presortindbuild' may be reported in the error log during CREATE CLUSTERED

INDEX WITH SORTED_DATA on an APL table in a database that has been loaded from a dump of a platform

with different byte order.

literal auto parameterization are enabled.

Example: SELECT b2 FROM t1 WHERE a1 BETWEEN'xxxx' AND 'xxxx'

There is no blank between a SQL keyword and a literal value surrounded by single/double quote marks.

partition condition objects and also gives repeated linenumbers in partition condition.

"current process ... infected with 11" may be reported in module 'lddb__unscr_log'

when database is loaded from 11.9.3 version of ASE across platform where the byte order

(endian type) is different.

the column from another table and the first new value is a NULL, a segmentation fault can

occur. The resulting stack trace will contain the function, inserttext.

If the query is executed by the diagserver, an assertion will preceed the segmentation fault: Assertion ((s_sdes->sdesp)) failed in file /remote/heartland_nbld2/ase1502/sunsparc32/porttree/svr/sql/generic/source/manager/text.c line 2268

handling of the identity column on each side of the UNION. These kinds of queries are no

longer allowed in ASE. A new error message, 16306, will be raised when such a query is

parsed. An example of such a query is:

1> select a=id*10000, n=identity(2) into #t from sysobjects where id < 6

2> union all

3> select 999, 999

Msg 16306, Level 15, State 1:

Line 2:

You cannot use an identity function in a SELECT INTO statement with the UNION operation.

The list of the commands associated with the structures is:

TMPOBJ

CSS

JOINDATA

J_FILTER

SARG

SARG_ONDISK

COST

RRANGE

PLIMIT

E_CACHEPROT

ST_TABLE

ST_IND

ST_PTN

ST_SYSTABSTATS

ST_ALLCOLUMNS

ST_COLSTAT_LIST

ST_COLSTAT

ST_COLINFO

SYB_COLDESC_ARRAY

SYBCOLDESC

EXECTXT

partitions were created on Solaris64) can cause a stack overflow error.

stale information and might result in propagating wrong values like ptnroot, ptnrow etc.

steps on the column, the compilation could hit the

Assertion ((*ppFreeBase >= *ppBoundary + ElementSize) || (*ppFreeBase +

*pFreeSize <= *ppBoundary)) failure under diagserver. Under

non-diagserver, it might lead to memory corruption and signal 11

stacktrace in the optimizer statistics handling code area.

column(s) with unique index (in other words, the usage of DISTINCT

in the view is redundant).

"Backup Server: 1.70.1.2: Configuring the shared memory per stripe to 0 bytes."

PIO/LIO for multiple executions of a subquery, (via set statistcs plancost diagnostics).

Previously, only one execution of a subquery was costed, and the total cost was

multiplied by the number of times a subquery was expected to be executed. The

new approach is to initially cost 100 executions of a correlated subquery, and

subsequent cost the subquery plan, at the attachment point, with the correct

number of executions, which will not change the subquery plan, but will affect

the cost reported, including taking advantage of any buffer manager optimization

that can be gleamed by multiple executions of a subquery. Also, the reformat

strategy will be chosen in cases similar to 12.5, which assumes 100 executions

of a subquery in order to decide whether reformatting is appropriate.

A user who is authenticated with LDAP server will get "username/NONE" in audit record instead of "username/LDAP" for LDAP state change auditing event.

may be generated in LeStoreOp::_LeOpClose(). This will only

occur if the session is recovering from an 1105 error while

trying to initially create the work table. Note, that this session

typically will not be holding very much tempdb space. Analysis

should still be done to determine why there is no tempdb space

available.

executing a query using "advanced_aggregation" and having

equality or non-equalitiy WHERE clauses involving columns

which store NULL values.

on the new coordinator instance encounters a stack trace.

process that is doing a DUMP DATABASE and a recovery process that is doing node failover recovery.

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

and 'd_usedextents' when DBCC USEDEXTENTS command has been previously run

against an archive database.

"Can't change the owner of the master, model, tempdb or sybsystemprocs database."

is printed which doesn't include local system temporary databases.

system tables, with other tasks executing DDLs and utility command on the same database.

stacktrace with sig11 (in CgpPop::cgpInit()) could happen.

gets the new values of the LOB column from a UNION ALL query (either a derived table or a

union all view) that has a literal LOB on one side of the UNION ALL and a LOB column on

the other side of the UNION ALL can change the target LOB column to garbage values instead

of null when the source LOB column is NULL.

timestamp column into a binary null column gets the current timestamp

value rather than the appropriate value from the source table.

after node failover recovery has been performed while a 'REORG REBUILD'

command for an index was active on the failing node.

instead of raising error.

buffer from the cache, on the pending validation status. The task itself

might be holding a latch on allocation page while going to sleep and that

can potentially cause an undetected deadlock.

spids involved in hang will contain function 'cm_bindremote_cache'.

as VARBINARY type instead of correct column type after BCP OUT with -C option is executed.

Shared Disk Clusters.

routed to the wrong logical cluster upon failover.

may be reported in the error log while doing a checkpoint.

return null-supplied rows. This will only occur if the merge-join

outer-join op's inner stream returns no rows.

Local file contents can be transferred into remote file/directory/table and vice versa.

recompilation. This can occur if there are not enough

worker threads initially available and the plan

cannot be runtime adjusted. Some plans simply cannot

be adjusted to use fewer threads and in other cases

there may still not be enough threads available to perform

RTA. In these cases, the query must go through RTA

recompilation (serialization). This will generate a SEGV

when the initial parallel plan is released if the user

has enabled query metrics, statistics io, or statistics plancost.

noaccents sortorder servers.

produced by doing a full scan instead of a direct access of the grouping

worktable produced by vector aggregation in a non-correlated nested

subquery.

since the statement will be placed into the statement cache, with a

stored procedure wrapper. The parameter is defined as the max length

for a datatype e.g. for an integer it would be 4. However for columns

with smaller lengths than the max length for a datatypes (e.g. tinyint)

there will be a mismatch on the search argument for an index lookup

which could result in suboptimial plans for the greedy algorithm.

The change will allow the search argument for an index to be considered.

It may affect queries with greater than 6 tables in the query, in

which indices have columns whose length is less than the max length

of the datatype. If timeout occurs in the optimizer, then a suboptimial

plan may occur.

search engine timeout may not happen even when the procedure cache consumption

by the search engine has reached the timeout limit.

varchar/varbinary column which is less than 255, the query may hang.

'login__ldapauth' may be reported in the error log if an engine is onlined after offlining it. No user

connections will be accepted on this newly onlined engine when LDAP authentication is enabled.

result in unnecessary renormalization of the stored procedure.

Similary if the procedure is executed by different users the object name resolution may not be done

correctly raising permission errors or breaking procedure execution with 'too many recompilations'

warning message.

cursor on the target table exists, subsequent sp_bindcache or utility

execution will fail since Adaptive Server will complain that there is a

level0 scan query active on the target table. Alternatively, a level0

scan query will fail since Adaptive Server will complain that there

is a utility execution active on the target table.

rows if the inner stream is empty and returns LRET_ENDOFSCAN

when opened. Note that the rows will be returned if the inner

stream returns LRET_SUCCESS on open and LRET_ENDOFSCAN on the

first next call. The later scenario is more prevalent.

concurrently with a badly partitioned application that is being run on more than one instance.

The stack trace of the DBCC thread will show the routines pchk_oam_work() and getpage_remote().

qualify a single row, ASE was incorrectly choosing to update the row in deferred mode

instead of in direct mode. Direct mode results in improved performance.

For internal use only: Included consistency checks for cache configuration operations.

Cache configuration will fail if consistency checks fail.

error 821 while executing database binding using sp_bindcache.

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

'rec_onlinetime' may be reported in the error log during ONLINE DATABASE on the HPUX 64 bits

platform.

optimizer might choose a sub-optimal plan.

heavy load environment.

The reason it is marked as 'internal only' is because this problem is seen in SMP version that is built off asecluster3 branch and that product is never released to the market (at least so far).

"ind__itagsetup".

configuration parameters.

For now, we disallow 'set ciphertext on/off' command within an instead-of trigger.

If the ciphertext setting when an instead-of trigger is executed is different from

the time when it is compiled, the instead-of trigger will be recompiled.

STEP 1

The type of query is EXECUTE.

Executing a newly cached statement (ssql_id=12345678912345678)

STEP 1

The type of query is EXECUTE.

Executing a previously cached statement (ssql_id=12345678912345678)

sessions; and in such situations, ORPHANTABLES will leave the temp tables orphaned without

reporting that there was a problem dropping them.

return null-supplied rows.

alter database to same log device (using override option) more

than once, the extended device fragments (except the first

extended fragment) are marked as "data and log".

spids involved in hang will contain function 'cm_bindremote_cache'.

the form "tempdb..#<temporary_table_name>".

parameter/datatype, 706 error might be raised.

where a table scan is choosen instead of index scan.

SELECT-INTO) statement is executed in parallel, LOB datatype

(text, binary, unitext) is involved and other session is

concurrently updating the source table.

system table. This may happen after the "LOAD DATABASE" command has been

executed for a database being greater than 63Gb.

when a session that was running an ALTER TABLE command to partition a table

has been chosen as a deadlock victim due to concurrent activity on system

catalogs caused by other DDL commands.

1) A query has one or more expressions in an IN-list or two or more OR clauses on the same column that contain expressions, such as:

select count(*)

from opl o , bals b

where (substring(o.zb,1,4)=b.bals or substring(o.za,1,4)= b.bals )

2) The expressions contain references to substituted columns (o.zb and o.za in the example)

3) The optimizer chooses a query plan using the special OR strategy: Showplan will contain a 'SCAN Operator FROM OR List' output

4) The OR-SCAN Operator is under the right hand side of a Nested-Loop Join operator in the query plan

5) The expressions in the IN-list or OR clauses return the same result value for the first row processed,

Then too few rows may be returned.

can take place twice and the OAM counts can possibly go negative.

or wrong result.

counts are configured for the remote server in the interfaces file.

while accessing the faults recorded in the dbccdb database' in a heavily loaded

server due to an internal but unreported deadlock.

When you add a char(3) column to this clustered index, it completes in an hour.

The where clause does not contain the char(3), it only contains the datetime column.

you may find approximately 3000 bytes memory leak in Network Buffers.

a control character at the end, less rows might be returned. Similar, if a like

pattern is a variable or stored procedure parameter that has control characters

at the end, less rows might be matched.

upto 5 joining tables involving cartesian products

and/or theta-only joins.

created in a caller procedure may result in signal 11 if the indexes

created on the temporary table changes between executions of the

caller procedure. This change of indexes on the temporary table

could be caused by index creation failure due to resource issues.

when a database is loaded using a database dump that was taken on a platform where the byte

order (endian type) is different.

sp_cacheconfig 'default data cache','mixed' operation succeeds instead of failing.

1) There is a parallel query plan with at least one correlated subquery,

2) At least one correlation column or the result value must be variable

length datatypes (i.e. INTN, not INT2 or INT4, or VARCHAR, not CHAR).

3) The query plan is being reused (it must be a stored procedure),

4) The query plan is runtime adjusted from a parallel to a serial

plan when the plan is reused after the first execution.

A signal 11 can occur during execution, which includes the function sq_add_to_cache in the stack trace.

message that the database must be in single user mode - and proceeds to display the results

of the check. The database does not need to be in single user mode to run the 'check'

option, however, spurious errors may occur due to concurrent activity.

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

'ptnc_replace_ptncond' when an ALTER TABLE DROP PARTITION command is run

while concurrent activity on the target table takes place.

that are considered invalid in the client character set.

in e_eval*LeEvals::LeEvGetFirstEval() during execution.

2) Kerberos logins upgraded from pre 12.5.4 to 12.5.4 having status value of 224 can not login.

3) No upgrade action automatically fixes status values 224 and 480 to be 0. sso should execute new stored procedure sp_update_authmech_value to fix the status.

assignment in a DISTINCT query with outerjoin.

Or the message "ks_falloc: Cannot allocate using ubfalloc(rnetmempool, 8192)." may be

printed out from HK CHORES task in ASE errorlog.

"current process infected with 11 in module 'cmcc_bufsearch'" may

be reported in the errorlog while doing cache configuration.

it tries to add log records to Syslogs.

and "cluster vote timeout" are incorrectly allowed to be set at the instance level.

for complex join query using DISTINCT view or DISTINCT derived table on expression(s).

was expected to be 1. Instead 2 was found." when SP_FIXINDEX is run against the system

table systabstats.

'dfl_dmp_initflushlist' or 5205 error may be returned during the execution of a DUMP DATABASE command,

The modules 'cdfl_dmp_initflushlist' and 'dpdb_main' would also be reported in the stack trace in

the error log.

not be used in the final plan even when it would be better.

Trace flag 467 can be used to turn off this functionality. So when 467 is on, such statements are not cached.

ct misses a clear documentation about the Repositories supported by each Version and Platform of ETL.

E.g. Where can a one find out if for IQ EEE on WIN Terminal Server an ASE native Repository is supported? And what for AIX?.....

Same is true for all ETL Versions.

Information needed is:

ETL Product, ETL Version, Database Name, Database Version, connect via: Native or ODBC, Driver Version supported.

SYB_BACKUP srvnetname is SYB_BACKUP, where it should be {SERVERNAME}_BS, as is in the sql.ini

The result is an inability to connect to the backupserver from isql.

for DISTINCT on a scalar aggregate result together with another expression result.

subquery.

reported in the errlog instead of message 3708.

@@cmpstate is >= 0 in HA-env, and it has no meaningful value in SDC or in non-HA env (it will be < 0).

1) lock scheme is APL (All Page Locking),

2) table has at least one encrypted column,

3) table has at least one variable length column,

4) a row has more than 255 bytes of fixed length data,

then BCP IN on those tables will fail to insert rows.

requested step count and the histogram tuning factor, is restricted to the maximum of 10% of the

number of rows in the table and the requested step count. However, if the histogram tuning factor is at the default value of 20 and the proposed number of extended steps is greater than 400, then the actual number of steps in the histogram is restricted to a maximum of the requested step count and 400.

power() is an integer datatype, a 3606 error (arithmetic overflow) can be incorrectly

raised.

column in the select list the same column alias will cause error 7348. If the same query

is executed without assigning column aliases, the query executes successfully. An example:

> select col1 as 'column1', col1 as 'column1' from tab1

> go

Msg 7348,

Level 15, State 1: Select expression results in more than one column having same name.

Column name 'column1' is specified more than once.

an OUTERJOIN query whose inner table is a VIEW with a CASE expression in the SELECT list and the CASE expression from the VIEW is also referenced in a correlated subquery above the OUTERJOIN.

configuration file format error will be seen during server boot-up. The handle

for this special case is added in cfg__read_section().

take effect when server is rebooted. To make changes in the same instance increase the value of

'max memory' config parameter and retry command." may be reported while configuring a named cache

with 64 cache partitions using sp_cacheconfig.

and DUMP DATABASE was running on the crashed node when the node crash happened.

for a join query could lead to signal 11 stacktrace in CgEvals::_CgCompVcol() when

a reformatting (store_index) plan is chosen.

request sent to the node.

while explicitly inserting max precision identity value to identity column.

This configuration value can be in the range of 16 - 100 with 16 being the default value.

to 15.0.1 CE, the system view sysquerymetrics will be missing. When installmaster

script is run it will hit error 208 while installing the sp_show_metrics system

stored procedure.

2. After successfully upgrading from an Adaptive Server installation version older

than 15.0 to 15.0.1 CE, the config option 'enable metrics capture' will be missing

in the sysconfigures system table.

We then correctly note that this change will not be permamnent (which is true) and that the command should also be entered into either the /etc/rc.d/rc.local file or /etc/init.d/boot.local (depending on platform) to make the change permamnet.

Whilst both these options work, the published method to make such changes permanent is to add the commands to /etc/sysctl.conf which should be supported on all Linux platforms.

For redhat confirmation see

https://www.redhat.com/docs/manuals/enterprise/RHEL-4-Manual/pdf/Reference_Guide-450.pdf

or for SuSe see

https://secure-support.novell.com/KanisaPlatform/Publishing/495/3664772_f.SAL_Public.html

recovery after an instance shuts down, if the instance that shuts down was the instance running

the Job Scheduler. Also, the shutdown of the cluster or instance when Job Scheduler is enabled

result in server hang. This happens when the cluster is configured with only one engine.

In the following scenarios, the server may hang :

o A Cluster enabled with Job Scheduler may hang during shutdown if

shutdown cluster was issued from the instance that was not running Job

Scheduler

o A Cluster enabled with Job Scheduler may hang during shutdown if

shutdown instance was issued for the instance that was currently running

the Job Scheduler from the other instance that was not running Job

Scheduler.

and mount the commands that will permit customer to apply the transaction again or to revert some transaction. There was CR 315186 but it was closed.

1) A left semi-join using the hash join algorithm and

2) Filter predicates in addition to the join clause that can

eliminate rows after they have matched on the joining column and

3) Rows with duplicate joining keys in the outer input stream

to the hash join op and

4) At least two rows that match in the inner input stream, the

second of which must fail the additional predicates the first time

it matches the joining column and pass the additional predicate

on a subsequent match of the joining column,

too few result rows may be returned.

Query plans with left semi joins usually appear for queries that have subqueries in an

IN-predicate.

unpartitioned might result in a 14108/20/4 error message "Could not find partition

descriptor for objid xxxx, indid x, ptnid xxxxx in dbid x" followed that by a stacktrace

that includes the following routines: ptn_get_pdes(), ptn_pdes_set(), open_partition(),

update_statistics_col()

CT-LIBRARY error: ct_connect(): network packet layer: external error: There is a protocol packet error. An illegal length was received

At present, the PLC discards are also displayed under the column "by Other" in the PLC flushes section. This leads to erroneous counter value for PLC flushes.

could happen raising messages like "current process infected

with 11" with routine names like 'mjutil_sdes_update' and

'memcpy' when a query joining two or more tables is executed

in parallel inside a stored procedure, the configuration

parameter "enable sort-merge join and JTC" is ON and the

query uses one or more of the stored procedure parameters

in the WHERE clause.

there are no concurrent database updates on the instance on which the checkpoint

is executed, and many concurrent database updates on one or more other instances.

1) It has correlated subquery that does not get flattened (i.e. NOT EXISTS)

2) Subquery caching on (it's on by default)

3) It has two or more correlation columns that come from at least two

different tables (i.e. a join under the left-most child of

the LeSQFilterOp.

4) The values of correlation column(s) comming from the inner

scan of the join must have varying lengths (i.e. some NULL

values and some not NULL or different VARCHAR string lengths).

5) There must be multiple rows from the inner scan of the join

that match the joining column of an outer scan row.

5) The ordering of the columns in the rows of the subquery cache

has to have the corrrelation column from the inner side of the

join placed before the column from the outer side.

6) You have to get some matches in the subquery cache.

Also, this bug is most likely to happen after the size of the

subquery cache has been adjusted down to a single row - but this

is not a prerequisite.

first time on an allocation unit, we might ask for an EX_LATCH while already

holding it. The task will keep waiting indefinitely for the latch to be granted.

This task can block other tasks too which tries to take latch on same allocation page.

in use) and literal parameterization is on, stack trace may occur from

opt_like_and() -> find_ep() when executing a statement with like clause against

a literal value.

fail over recovery process.

may run into stack trace in opt_like() -> find_ep() if the query is run under 'set noexec on'.

case insensitive server.

partially sorted may do many useless comparisons. This can

use a significant amount of CPU and slow down the query

significantly as well. We now check to see if previous

group checking is not effective (low hit rate). If not,

the last group comparisons are disabled.

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

'ind__get_sysindrow' when the DBCC CHECKVERIFY command has been previously run.

a timeslice error may be reported in the error log. In this case, the timeslice error

will be accompanied by a stack trace which includes the modules 'online_database' and 'th__recover_altered_logsegment'.

1> primary node fails, processing moves to secondary (already works)

2> power fails to the secondary node

3a> power is restored and only the secondary reboots (primary hardware still offline)

3b> power is restored and both hardware nodes recover

alter table lock change may run into 1142 error:

"Invalid OAM Page 0. ..."

back from the last statement in the data frame for grid mode. For text mode, you do not get any row counts at all.

select * from publishers

select * from authors

select * from authors where au_id = 'white'

Even though you get all the data back in data windows that are tabs.

You only get the last statement row count in grid mode.

Text mode returns no row counts at all.

Customer would like to get the row count listed for each select statement

executed. in each tab with the results set./ Or, in text mode, at the end of each output or in a seperate tab with the messages.

or sig 11 from com__subexctnumetochar().

LISTOAM.

join query (more than 6 joining tables) with correlated subquery in the SELECT list.

an EXECUTE IMMEDIATE command to be added to the statement cache. Without trace flag 456,

all SQL statements that appear within EXECUTE IMMEDIATE commands are excluded from the

statement cache. With trace flag 456, SQL statements within EXECUTE IMMEDIATE comands that

appear in SQL batches and in SQL stored procedures that have a not-dynamic ownership chain

will be eligible for the statement cache.

CREATE TABLE command using column defaults and an ALTER TABLE DROP CONSTRAINT command.

on allocation <pageno2> has objid <objid>, indid <indid>, and used bit on,

but reference bit off", may be reported by DBCC CHECKALLOC if a sort was

active just prior to an instance or a server being shut down abnormally.

An 'remote descriptors' is required for every ct-lib connection used by CIS

Valid values are documented to be between 1 and 100 inclusively.

Using sp_configure to attempt to set "max resource granularity" to

0 will now generate an error.

of them validating a data page at a node and another node cannot give-up

the lock on allocation page as it is blocked during destroy of a buffer.

function to run this query." maybe be raised when "statement cache" is configured and a

parameterized statement (containing '?' parameters) is executed without using the "dynamic prepare"

option and it sends a NULL value for one or more of the parameters and the datatype specified for

the NULL value is incompatible (no implicit conversion available) with the target column affected

by the statement. Command line traceflag 7741 is added to allow the execution of those statements

with success when "statement cache" is configured but those statements sending NULL values for

parameters will not be cached.

time and the Lava context has been cleaned up. That is, if a

second exception occurs while cleaning up for an initial exception.

In these cases, the stack will show that there is some intervening

handler before le_execerr(). In the typcial case, le_execerr()

is the first handler dispatched below the exec_lava() routine and

there is no problem.

view materialization.

extension of a large database when the database has been extended over

a boundary that is a multiple of 63 GB.

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

'buf_online_database' and 'cm_bufunhash' when an offlined database is in

the process of being onlined while a concurrent session is trying

to run a DBCC command in the given database.

columns of a table.

may be encountered while starting a new listener on the abnormally terminated listener port.

The status of that listener is listed as active. However, no logins can be made to ASE using that listener port.

lowest cost. There are certain circumstances, such as the use of Dynamic

Partition Elimination which result in the true cost of a query being lower

than the estimated cost as the value of the dynamic partition elimination

cannot be considered in costing to plan before execution.

error log when using encrypted column in a WHERE clause in a UNION query

to return the plan in the plan tab. In some case past the five min mark.

Seems to be performance degregation going on.

indid <value> in dbid <value>" together with a stack trace which

includes the modules 'ind_getrow' and 'startscan' maybe reported

when a stored procedure which uses temporary tables (or tables

create inside) and conditionally creates different indexes on

them is executed reusing a query plan from a previous execution.

Current Documentation on Served Licenses from the PowerBuilder 11 Installation Guide

Installing a served license

If you are installing a served license, your organization needs a license server. The license server allows you to share and control licenses within a pool of users. It controls license availability and allows you to use licenses as needed. The license server also blocks use when all licenses are being used.

after error 1105 is raised. The modules "dol_insert", "pg_getdatapage_forinsert"

and "ex_reraise_lasterror" will be shown in the stack trace.

and the RAND builtin appears in a SQL WHILE loop, the same value of RAND will be used in

each iteration of the WHILE loop. An example is:

delcare @v int

select @v = 1

while @v < 10

begin

if (rand() < .5)

select 1

select @v = @v + 1

end

go

The result of rand() in the IF statement will not change for each of the 9 iterations of

the WHILE loop.

clause such as:

declare @v varchar(20)

select 1 where

@v IN ('abc', 'def')

OR (@v in ('data') and NOT exists (select 1 from sysobjects where name = 'foo'))

can return incorrect results. If the query is executed with the diagserver, and assertion

will fire:

Assertion (pNode->sym.node.type == AND) failed in file

.../generic/source/preprocess/resrelop.cpp