The current maintenance kit is Caché 5.2.3. This document is only part of the available maintenance kit documentation. The full set of documents is available in the \Docs subdirectory of your Caché installation as follows:

• Late-breaking information about the release: \relnotes\cache\readme.1st
• Release notes for Caché 5.2.3: \relnotes\cache\relnotes.htm
• Release notes for the previous maintenance kits, 5.2.1 and 5.2.2: \relnotes\cache\prenotes.htm (this document)
• Description of Caché 5.2 new features, prior to any maintenance kits: \GRN2.pdf
• Description of Caché 5.1 new features: \GRN1.pdf
• Description of Caché 5.0 new features: \GIC.pdf

This document is a historical archive of changes in Caché since the original 5.2 release, but prior to the current maintenance kit. As you can see from the above list, changes unique to the current maintenance kit are described in relnotes.htm. This document, prenotes.htm, lists changes between 5.2 release and the last maintenance kit before the current one.

The remaining topics in this document describe each modification in detail. The format for this description is a Change Report.

Each Change Report provides a table with the following information:

• Likelihood: The probability that a user on the affected platform will encounter this issue.
• Risk: An assessment of how likely this change is to introduce additional issues into the application environment.
• Ad Hoc: Whether this change has previously been deployed in an ad hoc release and successfully used by customers in production.
• Enhancement: Whether this change provides new features or capabilities for the product.

Previously, when you uploaded a file with CSP, and the filename contains non-ASCII characters, these were not being converted into the correct charset, unlike the other request parameters. This problem has been fixed.

If you tried to load CSP files on VMS with:

s tSC = $$LoadPageDir^%apiCSP("/csp/user","c+d")

Then it would not recurse through any subdirectories as it should. Also it would not report any errors even though it generated an 'invalid directory' error. Now the error reporting is fixed, and the code to recurse subdirectories is also fixed.

Increase timeout to one hour for the SOAP Client Wizard page that generates and compiles the classes.

Fixed a bug where if query's ROWSPEC column name contains a caption, then a table may not be shown or an error may occur.

As part of 5.2 ECP server optimization, the ECP server may not deliver the answer as soon it completes during light loads. The ECP server rarely may delay delivery of the answer for up to one second.

The daemon jobs types were not set in the proper order; daemons may hang around on failure (incorrect type may prevent proper cleanup due to incorrect job type). I.e. The ECP server reconnection daemon may not go away when the security subsystem disables the ECP service-connections from a host.

In rare conditions, the ECP client may not detect the connection is dead and it may not drop the connection and reconnect to the ECP server.

If you specify that a collection property of a class is read-only, Caché now enforces this in the collection object so any attempt to update this collection will return a %Status code saying the collection is read-only.

This change allows list or array collections to be sql computed. The SQL compute code for list collections must return the data in this form:

$listbuild(item1,item2,item3,...)

For array collections the SQL compute code must return the data in this form:

$listbuild($listbuild(key1,item1),$listbuild(key2,items2),...)

For example:

Class Test.Pointer Extends %Persistent
{

Property Data As %Integer;

Property B As list Of %String [ SqlComputeCode = { s {B}=$lb({Data}) }, SqlComputed, SqlComputeOnChange = Data ];

}

This defines a computed property 'B' which has a single element in it of value equal to the property 'Data'. Note that you cannot define a calculated collection because the collection would have no in-memory storage.

Caché 5.1 corrected the return type of the IsValidDT method to return a %Boolean rather than a %Integer which was its return type in previous releases. To promote compatibility with this older change, Caché 5.2.2 updates all user IsValidDT methods to a return type of %Boolean if they were %Integer or %Library.Integer. This change avoids errors reported when compiling these classes because the superclass IsValidDT return type does not match the subclass's IsValidDT.

When you call ..Collection.Clear() it now resets the modified bit on the collection 'oref' slot as well as the 'oid' slot. This is required by the bindings so they can detect the oref has changed.

If you insert a user defined document type into a project and then export the entire project then the user defined document will not be exported because if it was called 'test.abc' then the export code was trying to export 'test.abc.abc' rather than 'test.abc'.

When Clear() is called on a collection previously it would remove all the data and the oref and oid. Now it keeps the oref to avoid having to create a new oref if it is referenced again and incur the cost of the object creation. This also means that if you have a reference to this oref it is still valid.

When calling $system.OBJ.Load with the 'c' flag to compile the items imported any user defined document types were not compiled even though they were loaded correctly. This problem has been fixed.

This change modifies the CopyFrom method of %Stream.GlobalCharacter and %Stream.TmpCharacter to work like %GlobalCharacterStream and copy the LineTerminator property.

Previously, if you have a parent/child relationship and then some 'dependson' dependencies, the compiler was not making sure that both the parent and the child class were compiled together. This meant it would fully compile the parent class and only then start compiling the child class which is not possible and so led to another compiler error. Now Caché ensures that the parent and child are compiled together, and if this can not be done it reports a class compiler dependency error.

With this change, buttons are now enabled only when something is selected that they can operate on. The exception is "Add" for zipping or jar, since you can always add a new zip or jar regardless of mode.

A problem with cursor movement via the TAB key has been corrected.

Previously, the key 'd' gave a state dump when typed in the Console. Caché no longer allows this. It does allow the key 'c' to clear the console.

Previously, Caché saved the properties file upon a "Next" event from the login panel. It no longer does this.

When connection fails due to a null pointer exception Caché now show it rather than hiding it.

Caché now displays, below the namespace list box, a text message about whether or not configuration file was read.

Attempting to emulate many-to-many relationship using an intermediate table in Jalapeño caused some stale objects to remain around. This problem has been fixed.

Previously, there were problems with character streams in Jalapeño when the connection was in full mode. Also, ClassParameter annotation was ignored during schema building. These problems have been fixed.

Jalapeño left some ghost objects inside ObjectManager when objects were first deleted and then resaved. The same happened if an object was deleted and then its clone saved. These problems have been fixed.

This change adds a method to use IEEE doubles in the class com.intersys.objects.SList (Java-side representation of Caché $List). By default they are not used, because old servers do not support them. To send doubles inside $list in IEEE format when using SList use this call:

useIEEEDouble (true);

It is also possible to call this static method:

SList.supportIEEEDouble (true);

Which causes all subsequently created instances of SList to use IEEE format for double.

Almost no other related functionality is changed, with the exception of the SysListProxy class. The latter is changed so that the user must always provide a ConnectionInfo object to create an instance of SysList.

When using a zero value for milliseconds and nonoseconds in a Java timestamp value, the incorrect representation was sent to the server. This caused datatype validation errors. The problem is now fixed.

Caché now has the ability to inspect the server-side $zobjval of an object in the Java debugger for the Full Java Binding client. For object obj try to evaluate the following expression:

obj.dumpZobjVal()

or:

obj.dumpZobjVal(true)

The former variant displays both server-side and client-side $zobjvalues while the latter skips the client-side and only displays the server-side. This change also adds a System Property that runs a permanent assertion that client and server values are exactly the same. The property is:

com.intersys.cache.assert

with valid values true and false. The default is false.

If "Generate" is pressed when there is no classes to generate, Caché now throws an exception in SchemaBuilderWizard.

Having no classes to generate can arise either because there are no classes picked in the right-hand side generation panel, or because the directory picked to generate is empty of classes (it and all of its subdirectories have no classes).

The generate panel contains the classes, dirs, zips or jars for which Caché will be generating schemas. When it is empty, Caché now disables (grays out) the "Generate" button.

A %Double scale can be 0-15. Previously, an incorrect %Double scale value did not cause a compilation error. This problem has been fixed.

A problem has been fixed that was introduced in Caché 5.2 where an $c(1) value embedded within the MAC routine source text was stripped by the macro preprocessor. Now, if you have a line in a routine that contains every ASCII value from 0 through 255, a warning will be displayed and the routine map used for debugging will not be built.

Caché now performs a validity check on the value of XMLNAME to see that it is a valid XML name.

In some cases, the user has a predefined client for which Caché needs to create a web service. To do this the user needs to be able to further customize the request and response messages. Specifying SoapAction for a WebMethod method specifies the SoapAction header for this method. If SoapAction is specified and not "", then SoapAction must be unique for this web service. Specifying SoapMessageName for a Webmethod method specifies the outer wrapping tag for the response message. The XMLNAME property parameter on the return type specifies the inner wrapper of the return value.

This change modifies the SOAP web client to optionally use SAX's built-in Http Requester which does the entire HTTP request in the SAX parser. To turn on the use of this feature, change the value of HttpRequester property of the web client class to "SAX". The default is "Cache" which uses the Caché %Net.HttpRequest class to issue the Http request.

The features available through the SAX Http Requester are somewhat limited. Thus, if any of the following properties are set %Net.HttpRequest will be used no matter what the value of the HttpRequester property:

• Attachments
• HttpProxyServer
• HttpProxyPort
• HttpProxyHTTPS
• HttpRequestHeaderCharset
• HttpUsername
• HttpPassword

In addition, if the message is a one way message (i.e. one without a return value), then %Net.HttpRequest will always be used.

A new parameter, XMLFORMAT, controls the generation of the XMLExport method for XML enabled classes to include code for only literal or only encoded format. This allows the generated routines to be significantly smaller since usually both formats are not needed.

If XMLFORMAT="Literal", then only support for literal format import and export is generated. If XMLFORMAT="Encoded", then only support for SOAP encoded format import and export is generated. The default is to generate support for both literal and encoded format.

Add the XMLIGNOREINVALIDATTRIBUTE parameter to XML enabled classes to control error reporting for unexpected attributes in an XML document when reading the document with %XML.reader. The default value for XMLIGNOREINVALIDATTRIBUTE is 1 (true) which is the current behavior. If XMLIGNOREINVALIDATTRIBUTE = 0, then an error will be reported if an unexpected attribute is encountered.

This change renames the %Net.SSL.Configuration class to Security.SSLConfigs. It also adds 2 new properties: description and enabled. It renames a previous property called Role to Type (Role is an SQL reserved word, and couldn't be used in queries.) It also exposes several API's to manage the SSL/TLS configurations. %NET.SSL.Configuration is deprecated and will be removed in the next version. Note that the name argument to %OpenID(name) needs to be lowercased with $zcvt(name,"L") before being called.

^SECURITY now has an additional option to import or export All security information into one file. Previously each type of security record had to be exported into different files. Prompting in ^SECURITY for IMPORT/EXPORT has changed.

A change has been made to SQL INSERT, UPDATE, and DELETE statement atomicity first introduced in Caché 5.1. Beginning in 5.1, the SQL Engine used a transaction internally to implement statement atomicity, which means the entire INSERT, UPDATE, or DELETE statement either occurs or does not occur. If any row failed to file for any reason, the entire statement fails. If the process is in NO_AUTO_TRANSACTION mode ($zu(115,1)=0), Caché will no longer start this internal transaction to implement statement atomicity. If you INSERT, UPDATE, or DELETE in this transaction mode and some row fails to update, other rows in the table may be modified by the statement.

This change converts floating numbers represented as a string (e.g. if IEEE double can not be used) to Caché Canonical form.

Previously there was a problem in that the following constructs have identical semantics in SQL:

x <> ALL (SELECT y ...) <=> x NOT IN (SELECT y ...)

<=> NOT (x = ANY (SELECT y ...))

However, treatment of NULL was not the same in these two cases. In particular, if either x or y are nullable and NOTs had been fully pushed down prior to the transformation, then the equivalent expression in both cases was:

NOT EXISTS (SELECT 1 ... WHERE x=y OR x IS NULL OR y IS NULL)

The semantics of x NOT IN (SELECT y ...) and x <> ALL (SELECT y ...) are now equivalent, under the condition that NOTs have been fully pushed down. To achieve this equivalence, the negation state is now maintained at parse time so that the appropriate NULL handling of ALL or NOT ANY or NOT IN inside the subquery can be generated properly.

SQL queries on a property with LANGUAGECLASS = "%Text.English" triggered unexpected <ILLEGAL VALUE> errors when run from the System Management Portal. This problem has been fixed.

In all versions prior to Caché SQL 5.2, processing of NOT IN (SELECT ...), NOT EXISTS, etc., subqueries did not strictly follow the SQL standard with respect to NULL handling. This problem was addressed by JLF090 by transforming:

x NOT IN (SELECT y ... WHERE p)

into:

NOT EXISTS (SELECT 1 ... WHERE p AND (x = y OR x IS NULL OR y IS NULL))

This strictly follows the SQL standard, and if either x is not nullable or y is not nullable, then one or both of the IS NULL tests is removed. If both IS NULL tests are removed, then JLF090 would have no net effect.

However, if one or both of x or y is nullable, then the presence of the OR makes the query harder to optimize. Fortunately, with extensions provided in this devlog (JLF111), another optimization provided in previous Caché releases and known as the "Set Valued Subquery Optimization" (JLF005) can be used in this case to provide excellent performance.

When importing XML documents with user defined documents in them remember the error count before calling the user defined document and restore it afterwards including any errors returned from the user defined document method.

If you import an XML document with a user defined document in it, the import notices the location of the user defined document and then after importing everything else constructs a stream containing just this embedded document and passes this to the user defined document class to import it. This worked correctly on Windows and Unix but OpenVMS when reading lines of data will not report the line terminator (unlike the other platforms) so you do not know if you read an entire line or if the line was >32k of data and so you need to keep reading to get the full line. This OpenVMS problem has been fixed.

If you are running a collation where 'a' collates before 'A' then no user defined documents would show up in Studio because it was searching for the user defined documents which always have the extension in lowercase internally as collating after upper case letters. Now the code no longer relies on this collation at all.

The distributed mgstat utility has been updated.

A comma was missed in the csv output format. This problem has been fixed.