The current maintenance kits are Caché and Ensemble 2008.2.5. This document is only part of the available maintenance kit documentation. The full set of documents is available in the docs subdirectory of the product installation directory. Release notes and related material are available as follows:

• Late-breaking information about the release: <install-dir>\docs\cache\en-us\relnotes\cache\readme.1st or <install-dir>\docs\cache\en-us\relnotes\ensemble\readme.1st
• Release notes for 2008.2.5: <install-dir>\docs\cache\en-us\relnotes\cache\relnotes.htm or <install-dir>\docs\cache\en-us\relnotes\ensemble\relnotes.htm
• Release notes for the previous maintenance kits since 2008.2: <install-dir>\docs\cache\en-us\relnotes\cache\prenotes.htm or <install-dir>\docs\cache\en-us\relnotes\ensemble\prenotes.htm (this document)

This document is a historical archive of changes since the original 2008.2 release, but prior to the current maintenance kit. Changes unique to the current maintenance kit are described in relnotes.htm.

This document includes the following sections:

• General Announcements
• InterSystems News, Alerts, and Advisories
• Online Documentation
• Change Reports for Previous Maintenance Kits

InterSystems advises all users to recompile their application classes as part of the upgrade process so as to take advantage of all the performance improvements as well as other changes that may affect them in this release.

Anyone upgrading from a Caché release at or before 5.0 will need a new license key.

From time to time, InterSystems publishes items of immediate importance to users of our software. These include alerts, mission critical issues, important updates, fixes, and release announcements.

The most current list can be obtained from the InterSystems Web site.

Users should check this list periodically to obtain the latest information on issues that may have an effect on the operation of their site.

As a convenience to our users, InterSystems provides online access to documentation for recent product versions at the InterSystems documentation Web site.

The following are the previous maintenance releases of 2008.2:

• 2008.2.4
• 2008.2.3
• 2008.2.2
• 2008.2.1

This correction fixes Ens.Rule.RuleDefinition:PurgeRuleCache to delete routine pointers. Without this change, it is possible to have two different rules pointing at the same routine in the rule cache.

This change makes the queries generated by the Message Browser use %ID instead of ID in naming the IDKEY field of message body classes. This allows the browser to correctly retrieve messages whose IDKEY field is named something other than "ID".

This change modifies %SQLGatewayConnection.Fetch to properly handle the end-of-data condition when using ODBC stored procedures that return dynamic result sets. In this case, the result set indicates whether more data is available; previously the adapter had been depending on SQLCODE to indicate this condition.

This change corrects the message trace display to function properly when the production is not running.

Before this, the display depended on the production being in a running state. When this was not true, events related to the production may have been omitted from the display.

In the details pane, the time logged for events was not being converted from UTC to local time. This change adds this conversion.

Users might experience performance problems using the Ensemble Message Trace when there is a high volume of Event Log entries. This problem has been resolved. In addition, there are now search filter criteria that allow users to have more control over the inclusion of events in the trace diagram.

When a message trace diagram is displayed, if the user clicks on the canvas a form containing fields with search criteria appears in the bottom frame. The control information provided is:

• IncludeEvents:

  Check the box to see Event Log items related to the session included in the trace, or uncheck to omit them. Default is checked.

• EventsWindow:

  Number of seconds, after latest time processed or EndTime field, whichever occurs first, to look for events to be added to the diagram. Default is 30 seconds.

• EventsMaxRows:

  Maximum number of table rows in table Ens.Util.Log that will be scanned, within the time limits defined plus EventsWindow, to look for events that match the session. Default is 1000 rows.

  The purpose of this limit is to contain scanning of the Event Log table. If the upper time limit is reached, the search stops; if not, in case there is a cluster of intense activity going into the Event Log, which is very unlikely, Ensemble will stop after searching this number of records.

If any of the above filter properties is changed, the change will remain for the duration of the Portal session. The other search fields affect the current diagram only. Clicking the Reset button reverts the properties related to events back to their defaults, reverts SessionId, and clears the other fields.

When running the management portal over HTTPS, the following SVG graphics would fail to render: Production Configuration diagram, Message Trace diagram, Dashboards and Business Process diagram. This problem has been fixed.

In the Production Configuration page of the Ensemble Management Portal, after sliding the frame dividers, if the user exited the page and later returned to display another diagram, he would see just the title frame and nothing else but a blue background.

The reason is that IE8 would set an incorrect size for the title frame. When it restored the previous sizes it would acquire that wrong size. This wasn't a problem with IE7 and IE6; it has been corrected by this change.

This change consolidates all the fixes to the Message Trace related to performance issues when there is a high volume of Event Log entries.

This change modifies the SQL generation logic used in the Ensemble Message Browser. The improvements made with this change are:

• Do not use the TOP phrase when SearchTable terms are involved or when not using SessionStart
• For a simple equality test of a single SearchTable property condition, do not use an IN subquery
• For negation operators (!=, DoesNotContain and so on), generate a NOT IN of the inverse positive operator. This makes it possible to exclude messages that have several property values where one of them satisfies the negative condition.
• For = or != an empty-string, use logic that includes or excludes messages that have no virtual document property at all of the specified type.

TAP833 modified the queries generated by the Message Browser page when there is only one simple equality SearchTable condition. This change fixes that modification to work when there is only one *enabled* SearchTable condition (and it is a simple equality condition). Disabled conditions no longer count.

This fix corrects an error in the platform detection logic of InstallShield that would detect a Windows 2003 PDC as Windows XP.

If a breakpoint was set on the line immediately following a ;;comment, the breakpoint would be ignored.

This is now fixed.

This change corrects an error where an attempt to start a remote database in ECP via the Management Portal on Solaris systems caused the ECP server to hang.

This change improves the checking of IP addresses to account for the fact that IPV6 addresses do not support wild card characters as IPV4 address do (for example, 192.168.*.*).

Prior to this change, output parameters in ODBC linked stored procedures were not accessible in the case where dynamic result sets were returned. This change also modifies the return value of %SQLGatewayConnection.Fetch. It no longer reports an error if SQLCODE=100. The user should check SQLCODE for errors when the end of data is reached.

These changes correct an issue where certain activities in Studio caused temporary information to be left in CacheTemp. Repetition of these activities resulted in expansion of the CacheTemp database to the point where it degraded system performance.

When a user disables syntax checking (Options->Editor->Syntax Check and Assist->Enable Syntax Checking "unchecked"), this change also disables all syntax-check parser-related actions, including studio assist, tooltips and context sensitive menu for particular syntax element will be disabled.

This change corrects a problem where the output of Japanese characters is garbled in Studio Debugger output window. Attempting to write

Japanese : <ASCII(12354)>

would result in

Japanese : B0

instead of

Japanese : <Japanese symbol>

This change assures that the currently active namespace is properly re-established when a routine search follows mappings to other databases.

With this change, Caché now allows the definition of custom i/o tables at the locale level that use built-in translations for input, output or both sides. Previously, attempting to use such tables would cause an access violation.

Application extensions to the Caché SNMP MIB were not getting registered on UNIX (or OpenVMS) when the Caché ^SNMP sub-agent started, and no instances of the user's MIB could be queried. This is now corrected.

This change corrects a problem where certain Unicode strings with sequences of repeated digits was stored in a compact form on the disk, but then decoded incorrectly in a situation where the encoded repeat sequence crossed a disk block boundary on disk.

This change works around an issue on Solaris with Zones configured. If LC_COLLATE is set, translation of strings in scripts fails. Because of this, an attempt to start a remote database in ECP via the Management Portal causes the ECP servier to hang.

Under some infrequent circumstances, a race condition could occur between a queue being deleted and the insertion of an entry into that queue resulting in an <UNDEFINED> error.

This has been corrected.

This change corrects a problem where a request for job summary information could result in a <STORE> error when there were many (>1000) unfinished business processes.

This change increases the amount of time allowed for a lock attempt to succeed. It also improves the job cleanup when the lock attempt fails which assure that future jobs will always have properly initialized global data.

This change assures that the various parts of the Ensemble scheduler use the same time resolution. A failure to do so in prior releases caused some Alarms to be ignored as already having been acted on because they were being measured against a clock with lower resolution.

This change fixes a condition that would cause an invocation of Queue:EnQueue to exit without properly releasing a lock.

This change adds an option to do SearchTable RemoveIndex() operations with the %NOLOCK SQL option. This prevents the table lock threshhold from being reached on the SearchTable when a virtual document instance is deleted that has many SearchTable entries associated with it.

When the table lock threshhold is exceeded during an SQL DELETE operation, the whole table is locked. For HL7 services, this prevents new SearchTable entries from being stored during the time it takes for the DELETE to complete.

In order to ensure that new SearchTable entries don't fail to insert when a Purge operation is executed during normal operation of the Production, this change also makes the delete triggers in the document classes invoke the SearchTable's RemoveIndex method using this new option

The File Passthrough Service (unlike the FTP passthrough service) moves the file to the ArchivePath (if any) after ProcessInput() returns. If ProcessInput does an Async call forwarding the filename, then a race condition would occur where the target Operation would look for the file in the WorkPath while the Service was moving it to the ArchivePath.

This change fixes the behavior of the EnsLib.File.PassthroughService to use the correct determination of when to use an Async call and when to use a Sync call, in order to avoid this race condition.

This change also modifies the File Inbound Adapter to make it skip renaming the file to ArchivePath if ArchivePath=FilePath. This avoids a problem where the renamed archive file is picked up by the adapter as if it were a new file because it appears in the FilePath input directory.

This change also adds the stream size to the trace messages in both the FTP and File Passthrough Services.

This change removes an expensive internal check for specific properties in the DTL target class. The check has been replaced with an error trap that is more efficient than the check.

This change allows the method EnsLib.EDI.X12.Document.buildRuntimeIndexFromPath() to successfully parse segment paths that include a required key field (e.g. HL!3=22-12) for content-sensitive segment map parsing.

In particular, this affects field 3 of HL (Hierarchical Level) segments. Depending on the value in field 3, an HL segment can identify any of several different possible loop structures within a document schema. When attempting to assign or save documents that are constructed in "PathMap" mode, e.g. by an ordinary DTL transform, this allows documents that contain HL segments to be created.

This change also strips the 'key field' specifier (e.g. !3=22) out of the segment type declaration stored as part of the node data in the segment map array that gets created when a new X12 schema is imported from a SEF file into Ensemble.

An X12 HL segment (Hierarchical Level) is used to distinguish nested loops that may contain similar types of segments. To do this it requires a fixed value in field 3 (Hierarchical Level Code). However, there is no ambiguity when an X12 document structure contains only a single loop with an HL segment; therefore, it is OK for a SEF file to have an HL segment specified without specifying a specific value for HL:3.

This change also prevents a possible crash should this circumstance occur.

This change fixes the behavior of the X12 Service classes when ReplyMode=Immediate. This mode requires that child documents not be forwarded until the entire inbound document is received and parsed. Otherwise, it could be rejected while portions of it have already entered the Ensemble system.

This change also fixes a condition where this class would lose the assignments of DocType to each child document based on the value of the DocSchemaCategory setting.

This change corrects and simplifies the logic used to determine when to show a non-separator character in field ISA:11 (defined as InterchangeControlStandardsIdentifier/"U" under Control Versions before 00403, Repetition Separator in later versions).

When the value of field 11 is a character or number, that value is displayed. Otherwise, the display value is the repeat separator in effect for the current display of the segment. The control version is not taken explicitly into account in this determination; any character that is not a number or a digit or a different separator is a valid Repeat Separator.

For documents input with a "U" or other alphanumeric character in field 11, there is an implied Repeat Separator equal to the default X12 repeat separator which is Ascii 30 (RS/Record Separator), an invisible control character.

This change fixes mapping of X12 schemas that use HL (Hierarchical Level) segments containing key values to distinguish different loops in the document structure. The schema importer generates special decorated segment names of the form such as "HL!3=22-12" where HL is the segment name, !3=22 means that field 3 of this HL segment is required to contain the value 22, and -12 denotes its ordinal position in the schema.

This change allows the segment type name to match in the buildRuntimePath... method by stripping out the ! clause from the segment path name.

This change fixes ASTM parsing of M (Manufacturer) message types, that is, message types that contain manufacturer-defined M segments with field 2 containing the type of manufacturer segment (and by implication the type of manufacturer message). Formerly, this type of message would get a schema error.

This change makes sure that numeric values are converted to $DECIMAL before being stored. In prior releases, it was possible to store a double value in a %Numeric field.

This change makes it possible for someone with read access to the CACHESYS database to run %SS. In previous versions, write access or %Admin_Manage:USE were required.

A problem has been corrected where a single-field IDKEY field defined in a Map Subscript expression, for example,

$$Tag^Routine({MyIDKeyField})

specified for the map subscript expression in a child table would cause an <UNDEFINED> error.

This change corrects a problem where, in certain rare cases, a syntax error would be generated for a SELECT query. The occurrence of the error depended on what names were used in the variable list of the INTO phrase.

This change fixes a problem in cstat in Caché 2008.2.x versions where the list of processes waiting for a resource would not be shown.

This change corrects a situation where multiple processes examining the JOB queue at the same time could cause system degradation, and continually get NOJOB errors.

This change corrects a situation where a simple routine to exhaust memory, for example,

f i=1:1 s a(i)=$zu(188)

will cause a JobServer to loop forever.

Addressed an issue where $ZDATEH() with date format 3 (ODBC format) and a two-digit year would return an error result instead of the correct value.

A problem has been corrected where in certain rare cases a code generated syntax error would occur on a SELECT query depending on what names were used in the INTO variable list.

InterSystems internal testing uncovered a small window in OpenVMS clusters where the failure of a cluster member within a minute or two of a new member joining could result in a cluster crash. This change resolves that issue.

The problem was introduced in 2007.1 and can be recognized because there will be a message in the console log of the cluster master that the wdpass values in the write-image journal files differed by more than 1, or there were more than 2 distinct values found.

This change corrects an error that could cause the write daemon to hang. The circumstances triggering the error include frequent journal file switches, heavy journal activity, and a system interruption.

This changes corrects a situation where routines that need to start a transaction (for example, to modify an object or execute an SQL query) would hang if called via %ZSTOP during shutdown.

A problem has been corrected that could cause starting a Caché instance from the cube to fail with an "Access Denied" error message. The problem occurred on Windows Vista and possibly Windows Server 2008 with a non-English locale. The problem was reported with Brazilian Windows Vista.

This problem was present on all Windows Vista capable versions of Caché (2007.1 forward).

This change corrects an issue during an upgrade where the upgrade code was referencing the wrong version value.

This change corrects three issues with upgrades from Caché version 5.0:

1. The code was checking for version 1.4 not 1.3
2. The code was not returning a %Status value
3. The code was not properly handling an old bug where tasks could get miscategorized

Upgrades from 5.0 to 2008.2 require that the Task Manager be upgraded before new tasks can be created. This change ensures that the Task Manager is upgraded before the installation creates the system Inventory task.

Version 2008.2.2 of Caché includes the following changes:

This change avoids problem with the IIS V7 Native Module Interface that prevented chunked CSP responses displaying correctly (that is, responses sent using HTTP chunked transfer).

This correction only applies to x64 platforms.

Also, this build strengthens the code that checks for systems management forms: if the PATH_TRANSLATED field for the request is empty then the Gateway will use the value of SCRIPT_NAME instead to determine the extension of the requested file.

This change corrects a problem in the code responsible for sizing the buffer requirements for configurations containing huge numbers of server and application path definitions. The issue could potentially result in UNIX and OpenVMS based Gateway installations not loading all the configurations listed in the CSP.ini file at start-up time. This, in turn, would result in applications not being able to access the resources that had failed to load (for example, access to particular Caché servers).

This change resolves a security problem that prevented the use of Shared Memory under Windows 2003 (and later).

The availability of shared memory is not critical to the Gateway and CSP can function normally without it. Shared memory is currently used as means for communicating between the Systems Management and Run-time module. For example, it is used to cache the security tokens used to protect access to the Gateway management forms for configurations where the administrator chooses to password-protect them. As a result, one visible consequence of an inability to access shared memory is the need to repeatedly log in to the Gateway management forms.

The Gateway, on failing to secure shared memory under Windows writes the following message to the Event Log:

cspSHMinit : Startup error
Unable to create a handle to a shared memory block - code: 5 - Access is denied.

This change resolves a race condition caused by multiple frames accessing the Gateway data block responsible for registering state-transitions (preserve mode changes) against incoming sessions IDs.

The functionality responsible for managing the transmission (and caching of) response data between the Cache/CSP engine and the hosting web server has been redesigned in order to resolve occasional memory management issues when GZIP compression is used.

The new design greatly simplifies (without introducing any extra overhead) the management of memory and the relationship between the response streams and the Gateway cache. It also simplifies the processing of response data that must be locally cached for the purpose of generating a content length field for the HTTP headers.

A fault in the Apache API module which could occasionally result in a failure to dispatch HTTP response headers for which there was no following content has been corrected. This fault could affect HTTP redirect headers. A failure to dispatch the headers under these circumstances would typically occur when KeepAlive connectivity was in use.

This change corrects an issue that resulted in wildcard requests (*) for all CGI environment variables (as specified in the Gateway configuration) not being honoured by the CSPnsd module.

This change corrects a situation where updating a production could cause extra, duplicate alarms to be set in the scheduler for the production. Actively modified productions could accumulate many, many such alarms.

ScheduleSpec is to be specified in $ZDATETIME($H, 3) format; it can include wildcards ("*") for all fields and "WEEK" for year value.

ParseScheduleSpec looks at all the ScheduleSpec values, resolves the wildcards and returns the next closest event in absolute time. In ParseScheduleSpec, there is code that parses the ScheduleSpec looking for the special values and copies the non-special values to the output which is in $ZDATETIME format.

However, when the values in ScheduleSpec are malformed, it could cause the $ZDATETIME to be malformed. For example, "2008-11-04T*:00:00S" is such a value with an extra "S" at the end. This would yield a malformed $ZDATETIME value "2008-11-04T19:00:00S".

This change forces a normalization to numeric value of each field when constructing the output $ZDATETIME.

Before this change, the week-count value was not normalized. Now it is.

There was a circumstance where the attempt to identify dead jobs created ghost job nodes with increasingly long names that eventually led to generating a <SUBSCRIPT> error.

This has been corrected.

This change corrects an internal error where Ens.MonitorService:Purge was using the wrong global to search for jobs to terminate.

This change corrects a situation where the use of the WEEK syntax in scheduling events could cause a tight loop that wasted processor time.

This corrects issues in constructing reply messages that could cause an infinite loop while attempting to contruct a message reply.

This change fixes an <UNDEFINED> error in the code for HL7 and X12 'AckOut' Operations.

This change fixes an issue where EnsLib.HL7.Segment.OutputToIOStream causes the segment terminator characters that may optionally be appended to the pSeparators argument to appear in the MSH:2 field where the field separator characters should appear.

This change fixes an error with IN conditions used in the Message Browser search criteria. An empty IN condition would always be generated regardless of the user's specification.

This change corrects the segment references used in "leftoversegs" PathMap entries. Leftover segments are the remaining ones in an HL7 message after BuildMap() has parsed all the segments it can parse. Typically they are Z segments, but they can be any kind if the schema does not match the document.

The silent upgrade of an instance which does not have "Caché Engine" feature installed will no longer result in "dialogs out of order" error.

This change causes Caché to release its cached line tables when internal free memory is low.

This change causes Caché to examine successively higher-level cached routines to free when it is about to run out of memory in the routine cache.

This change corrects an error that could cause an <ACCVIO> when merging two arrays that had only integer subscripts.

This corrects a situation where $ZDATEH() could return an error when the date ended with a trailing space character.

If a routine has two or more references to a label that is within a dotted section of code or procedure block, all but the first reference would be compiled incorrectly, causing them to branch to an incorrect location in the code. This has been fixed.

To support ECP recovery the database server syncs its activities in the journal before the client can drop them. This change modifies ECP to not perform the synchronization when the request does not insert any journal entry.

In rare conditions in low load environments, the ECP client may delay flushing pending requests. This modifies the client to flush as soon as possible.

This change forces the ActiveX binding to use %Library.Decimal for numeric return values.

Before this change, CacheActiveX and the dynamic interface ofthe C++ binding projected %Numeric and %Decimal values as doubles. Now they are projected as decimal values, so the server and the client receive the data without any conversion.

This corrects a situation where an Export qualifier had an incorrect entry in its list which caused an error when used.

This change corrects a problem in which, when the Light C++ Binding stored an index entry for an index on a property of type %Timestamp, the month was 1 less than it should be (for example, it stored 0 instead of 1 for January). Timestamp values were (and continue to be) stored correctly in the data, it was only the index entries that were stored incorrectly.

This change fixes a problem in which Light C++ Binding failed to lock when checking for duplicates on a unique index, which could lead to incorrectly violating a unique index constraint if multiple threads near-simultaneously attempt to store objects with the same index value.

This change affects Light C++ Binding C++ code generation, so after upgrading to a version containing this change, any Light C++ Binding generated C++ code must be re-generated and recompiled.

This change makes the ReadLine() method of class %IO.StringStream work properly when more than one LineTerminator character is specified. That is, it should break the line at whichever terminator appears first, and if the terminators are CR/LF it should return them both as a single detected-terminator string.

Use of an XDEV in CacheSAX or CacheXSLT requires a callout to return a string representation of a pointer value. The printf specifier that was being used would truncate a pointer on a 64-bit system. This has been corrected.

This change notes the rebuild of the Xerces Libraries for x64 in 2008.2.

A correction has been made to the SQL privilege checking. Before this change, it was possible to CREATE, DROP, or ALTER an index even if you did not have %Alter_Table privileges.

A problem has been corrected where a DDL statement added a UNIQUE constraint or PRIMARY KEY constraint to a table with existing data, and the data for the constraint was not unique, an error was not returned.

This change corrects a problem with linked tables. If an application contained a query that was over 4000 characters long, an extra character would be added to the SQL that could result in an invalid statement and an <UNDEFINED> error at runtime.

This change corrects an SQLError that results when ODBC attempted to set up an environment header, if the connection has not been established.

This corrects an error caused by incorrect version checking in the Caché server. The error message that resulted was "class '%Studio.Debugger', method 'Attach': too many arguments". It is present in 2007.1 server when the maintenance version is higher then 5.

Under some circumstances, the current location shown by the debugger could be one line before the correct line. This is now corrected.

When searching for a routine to load, if any namespace is reconfigured, the unused namespace is released. Without this fix, an incorrect and freed namespace is associated with the current routine context. If the freed memory is reused for something else and if the routine namespace context is required for $TEXT or error trap, the consequences are unpredictable.

This change modifies the number of routines cached to be dependent on the amount of shared memory to be allocated per process. Prior to this, attempts to explicitly set the number of routines could result in <STORE> errors.

Previous changes to the viewbuffer message made the defined length of the viewbuffer too small, causing overwrite which corrupted the message queue pointers. Users might see <ACCVIO> in the message queues during view operations.

This has been corrected.

This change allocates the storage needed for monitoring or debugging a routine in the routine buffer itself.

A minor issue where the control process might have suspended the write daemon for 1 second after Backup.General.ExternalThaw() is executed has been resolved.

This situation could have occurred any time ExternalFreeze() was in force for more than 10 minutes. In this case, when ExternalThaw() is called there was a window while the write daemon was waking up where the control process does not see the extended time window set as part of ExternalFreeze() and tries to suspend the system. The write daemon starts running and the control process unfreezes the system one second later.

While this could have been seen on Windows or VMS systems, it was most likely to occur on a Unix system.

This change fixed a ExternalThaw() hang when the write daemon is suspended by ExternalFreeze() and no global buffers are available.

Previously, the ExternalThaw() would hang when it is started from a command line of a different Caché job from the job called ExternalFreee().

This change addresses an issue where the write image journal info was not kept up-to-date when journal files were being switched many times in a short period of time (for example, if the maximum size of a journal file was configured too small, and there was heavy journal activity).

If the system crashed during this time, the journal recovery at next startup may take a long time due to the outdated journal info.

This change corrects an issue that caused cluster journal restore to abort with an <UNDEFINED> error.

Version 2008.2.1 of Caché includes the following changes:

When CSP caches a page returned from Caché it is usually the case that the page is cached with the session ID (either as a CSPSESSIONID cookie or CSPCHD form variable) of the user that first requested the page. On subsequent retrievals of this page from the cache for other users, CSP must replace all instances of the old session ID with the ID of the current user accessing the page.

This change corrects a problem where the replacements were done even in places where they should not be, such as embedded Javascript in the page.

In the 2008.2 releases of Caché, the optimizations for (a+1) and (a-1) do not work on 32-bit Power PC systems when A is an expression of the IEEE double-precision binary floating-point type. The compiler special cases increment-by-1 and decrement-by-1 and evaluates these operations using special routines written in assembly language. The 32-bit Power PC version of these assembly language routines do not properly implement increment/decrement of $DOUBLE values.

An argumentless ZLOAD with an immediate end of file leaves the partition empty. In this situation an access violation could occur on a subsequent DO command. This is now fixed.

This change corrects a problem where $LISTSAME could give erroneous results for lists involving strings and negative numbers.

On the Task Manager wizard, new prompts have been added for Start Date and End Date for Daily definition, similar to that of Monthly and Monthly by Day.

The default for the Start Date is today.

On the Task Manager wizard, new prompts have been added for Start Date and End Date for Weekly definition, similar to that of Monthly and Monthly by Day.

The default for the Start Date is today.

The ECP server shutdown has been changed to cleanup any queued pending pre-fetch requests.

If not cleared properly, in very, very rare conditions, the ECP server may receive a msg while the same session's write daemon is shutting down, and leave the background pre-fetcher in a confused state.

To provide its guarantees, the ECP server depends on the durability of its activities. Durability tracking was dependent on the journal file number, which is not updated and tracked in parallel with the journal offset. This change modifies the durability logic not to depend on the journal file number.

This change corrects an error in exchanging strings between the client and server. When receiving strings from the server, the C++ binding now converts them to Unicode unless the client has the same locale. When sending strings to the server, the binding now converts strings to Unicode and then to the server locale, unless the client and the server have the same locale.

This change converts strings back to numbers when they are returned from methods. Previously, this was not done in all circumstances.

The class compiler had a dependency on '~' collating in a certain order which was not true on a specific Czech locale. This caused an error like:

ERROR #5355: Method generator dependency unresolved: ,,,,,

ERROR #5030: An error occurred while compiling class User.Test

when compiling a class.

This has been changed so that the code now uses an integer value which will always collate the same way regardless of locale.

When the class compiler processes something like ##class(Name).Method(), it normalizes this classname based on the import list or the origin of this method, that is, the class where this method was defined.

However, generator methods are treated as if they were implemented in each subclass. So this changes the compiler to normalize names in this case relative to the current class rather than the class where this generator was initially defined.

The class keywords, Dynamic and SqlRoutinePrefix, are no longer used, but this change still allows them to be imported in XML files.

When sending the cached output in a redirection, an error could occur if the redirection was executed in a context that made bindserver variables unavailable. This has been fixed.

A problem has been corrected where an include file was stored in a precompiled state but the include code contained a #execute MPP directive that needed to be executed for each routine where the include file was used.

If a class defined a property of type %Stream.GlobalCharacter or %Stream.GlobalBinary and then called the <prop>GetObject method to return the stream OID, the OID was not correctly formed. This caused the delete of a stream to silently fail in those cases where it used the OID from <prop>GetObject to call %Delete on the stream class.

This has been corrected.

This change makes sure that the element prefix is always defined and thus fixes an error in the Canonicalize method of %XML.Writer that lost the element prefix when there were multiply-defined prefixes.

This change corrects a signing error with elements that have a prefix different from that of the default namespace defined in that element. The symptom of the error was extra xmlns="..." in child elements.

This changes fixes a problem with a <SUBSCRIPT> error when calling Security.Applications.Modify(). The error occurred when the list of target roles for the application definitions exceeded 255 characters in length.

A problem has been corrected where "FOR SOME %ELEMENT" on a MV multivalue field might encounter a <LIST> error at execution time.

This change corrects the behavior of package mapping. For example, assume

• namespaces exist called N1 and N2;
• there is a package PK in namespace N1, and it has been mapped to N2 via package mapping;
• a package FK exists in namespace N2.

If there is a foreign key in N2.FK.Class that references N2.PK.Class, the application must make sure package FK is also mapped to namespace N1. Failure to do so will cause any SQL INSERT/UPDATE/DELETE or Object saves/delete on class table PK.Class in namespace N1 to fail with a <NOROUTINE> error.

A problem has been corrected where some literal values used in a SELECT query via xDBC may not report the correct datatype value if literal values appear in both the SELECT clause and the FROM clause of the query (such as in a sub-query in the FROM clause of the main query).

This change optimizes the processing of queries involving arrays of elements. The processing for the elements was moved to be inline rather than processed in routine calls.

This change corrects the handling of a condition of the form:

(F=? OR F=? OR ...)

to avoid causing a <FRAMESTACK> error when it contains hundreds of terms.

This change corrects an error in the handling of queries in the style generated by MS/Access. Such queries contained conditions in the WHERE clause that occurred in an order different from the field order of the row. Under some circumstances, this caused incorrect code to be generated for the query.

This change corrects an error that can occur with FROM clauses and subqueries. Under some circumstances, queries involving nested IN queries within WHERE clauses in conjunction with HAVING clauses could disable important optimizations.

Studio was calculating version based on MAJORVERSION and MINORVERSION. This caused a problem in connecting to 2008.2.1.

The calculation has been corrected so that it now satisfies the requirement that the client should have a version not lower than the server.

This change fixes the methods AsAString() and AsXString() used by Studio to determine the length of a string to be converted. Prior to this, under some circumstances, calculation could result in an access violation.

Beginning with 2008.1, Studio printing added an extra blank page at the end of a print job.

This is now fixed.

This change re-enables support for float-point that was disabled accidentally for 2K databases. The symptom of the problem was that errors occurred when attempting to KILL a floating-point node in a 2K database.

This change adds two missing cp1250 table descriptors that prevented 8-bit locales based on this character set (for example, csw8, plw8, huw8) to be installed.

This change avoids a potential <WIDE CHAR> error during the NLS phase of Caché upgrades running on 8-bit locales that use non-Latin1 character sets such as CP1251-based Russian (ruw8).

This change allows digits in the middle characters of locale names (for instance, "yt8w"). These were prohibited by ^NLS, but could be produced by cnls.exe in older versions and therefore are valid.

This change corrects an internal stack management problem when loading a routine larger than 65311 bytes. Before the change this situation could lead to the stack overwriting other memory structures which could, in turn, lead to memory exceptions and process dumps.

This issue occurs on any system where a process loads a routine larger than 65311 bytes more than 255 times.

The 'margin' parameter of the terminal USE command (e.g. USE 0:(270) ) was handling values >255 differently on OpenVMS Alpha and Itanium. The documentation says "Values from 1 to 255 set the right margin for output; any other value disables the right margin".

However, on OpenVMS Alpha, Caché handled values up to 511. On OpenVMS Itanium, Caché internally truncated the value to one byte, essentially acting as a #256 on values between 255 and 512. This change makes OpenVMS Itanium act like OpenVMS Alpha.

A callin program that tries to start an instance before the instance is fully initialized locks the instance and prevents startup while the callin program exists.

This change corrects that situation.

This change fixed a database security hole that allowed a user who did not have access to a global to successfully execute a $ORDER on such a global.

This change fixes a possible lock table inconsistency between client and server when an ECP server shuts down and is restarted. In this situation, some "cancelled" locks might stay in the client's lock table even though they did not exist in the ECP server.

This problem exists on big endian platforms.

This change corrects a problem where a missing "^" before a reference to rINDEX caused Caché to rebuild the routine index incorrectly with a bad length.

This change corrects a situation where the default answer of "NO" might be interpreted as "YES" in %ER.

This change allows Backup.General.ExternalThaw() to complete in an OpenVMS cluster environment. Prior to this, it would hang.

This change corrects an error where extra columns may be displayed in Zen table panes under certain circumstances. This can happen when (1) dataSource is set to "columns", (2) there are more columns in the source data than there are columns defined by the tablePane, and (3) table uses "snapshot" mode.

If the current day is the 31st of the month, and you select a date in a month with less than 30 days, then the calendar will display the wrong month.

This is now corrected.

This corrects an error where client is using daylight savings time, close to the end of the day the date used by calendar may be off by one day.

© 2009, InterSystems Corporation. All rights reserved.