QF-Test  –  The GUI Test Tool for Java and Web

Release notes This document lists changes to QF-Test since version 3.0. For background information about the QF-Test versioning scheme, please see The life-cycle of a QF-Test version. The latest version of QF-Test is available for download here.

Version 3.4.4 - January 30, 2012

New features:

	Support was added for current versions of Firefox 9 and 10.
	Eclipse/SWT support was added for Eclipse version 3.8M5 and experimental support for Eclipse version 4.2M5.
	New checks were introduced for SWT Decorations, small icons typically shown next to a text field. You can now check the visibility, icon and tool tip of a Decoration.
	Iteration ranges can now also be used when setting up data at script level via the databinder module or by calling one of the 'Procedures' qfs.databinder.bindList or qfs.databinder.bindSet in the standard library qfs.qft.
	The thumbnail images shown for screenshots in the HTML report are now of higher quality.

Bugs fixed:

	A small change in QF-Test version 3.4.3 regarding the RMI communication between QF-Test and the SUT was undone because of possible side-effects when the SUT was using RMI itself.
	When using 'Check image' nodes with an advanced algorithm and a timeout QF-Test always used up the full timeout even when the image matched immediately.
	When a 'Check' node with a timeout fails, QF-Test used to show the initial state of the target component in the run-log. Now it shows the state at the end of the timeout period instead.
	When using Drag&Drop to move nodes within QF-Test the tree could jump erratically in some cases.
	Trying to remove the same node twice from a search result dialog could lead to an exception.
	In the standard test-suite library qfs.qft the 'Procedure' qfs.shellutils.touch now correctly updates the timestamp for existing files and the 'Procedure' qfs.autowin.getURLFromIE now also works with Internet Explorer version 6.
	When recording a mouse click on an SWT ToolItem that opens a drop down menu QF-Test now records the coordinates so the event can replay correctly.
	If an alert dialog was shown after closing a modal browser window, the main browser window could remain blocked after closing the alert.
	Javascript errors in Internet Explorer were not logged correctly in QF-Test's run-log.
	In some cases the initial loading of a web page did not succeed.
	Highlighting the selected component in a browser on Linux has been improved.
	Web pages with a Flash element were sometimes not displayed correctly on Linux.
	The name recorded for file upload on Linux contained the Windows file separator character '\'
	There were problems with browser startup on special Windows versions if QF-Test was installed below a directory with blanks in its name, e.g. "C:\Program Files (x86)"

Version 3.4.3 - October 27, 2011

New features:

	Support was added for current versions of Firefox 8 and 9.
	The 'Excel data file' node now supports reading .xslx files in Excel 2010 format.
	The performance of some advanced image check algorithms was significantly improved.
	It is now possible to fetch option values in a script via the new method rc.getOption.

Bugs fixed:

	When switching test-suites by clicking a suite tab the keyboard focus was not correctly restored.
	The keyboard shortcut [Ctrl-D] for setting a bookmark did not work.
	Jumping from a node in the debugger window to the same node in the test-suite now works again.
	In some special cases QF-Test could kill the wrong process on Windows when trying to terminate the SUT.

Version 3.4.2 - September 15, 2011

New features:

	Full support was added for Firefox 6 and experimental support for Firefox 7.
	For easier integration with continuous integration tools like Hudson/Jenkins or TeamCity the exit code of a batch run can be forced to 0 even in case of warnings, errors or exceptions with the help of the command line arguments -exitcodeignorewarning, -exitcodeignoreerror and -exitcodeignoreexception.
	If the option "Replace no-break space with normal space" is activated, QF-Test now represents no-break space characters in a web page as normal space characters in checks and 'Component' node attributes to avoid confusion.
	QF-Test now supports capture and replay of "expand" and "collapse" events for SWT ExpandItems.

Bugs fixed:

	The integrated help viewer of Eclipse uses a separate SWT Display on Windows which caused problems during replay. It is now consistently represented with 'GUI engine' "swt1".
	Direct execution of Javascript code now also works in Internet Explorer 9 despite stronger security restrictions.
	The AJAX resolver for GWT has been updated to version 1.4.0 to fix filtering of <DIV> nodes from the component hierarchy.
	Opening a modal popup window could crash Firefox 6 on Windows.
	Double clicks are now recorded correctly in Internet Explorer.
	A 'Fetch text' node for a <SELECT> node now retrieves the text of the selected <OPTION> node and 'Fetch text' nodes for <OPTION> elements represented as items are working correctly.
	Printing unicode strings in Jython could cause problems when running in batch mode.
	The content of 'Data table' nodes was missing in the run-log.
	A 'Wait for client to connect' node with a very short timeout did not throw a ClientNotConnectedException when single-stepping in the debugger.
	The QF-Test daemon was leaking threads when starting a remote SUT process for use with the procedure qfs.daemon.startRemoteSUT.
	When updating failed image checks from the run-log the size of the image was not always updated correctly.
	When recording procedures, QF-Test failed to create a procedure in case the respective component had no value for a placeholder like <COMPFEATURE>.

Version 3.4.1 - July 15, 2011

New features:

	Full support was added for Firefox 5 and experimental support for Firefox 6.
	Eclipse/SWT support was updated for Eclipse version 3.7 Indigo.
	Groovy was updated to version 1.7.10.
	Error messages for errors in Groovy scripts now show proper context and line numbers.
	QF-Test can now handle custom SWT components that hide their child Widgets by overriding getChildren() to return an empty array.
	Support was added for ExtJS version 4 and for RichFaces version 4.
	The 'As regexp' attribute of 'Window' and 'Component' nodes can now be defined as a variable value.
	The startup sequence created by the quickstart wizard for web applications now has a smarter layout and is easier to adjust.
	On Windows systems QF-Test now makes use of the always-on-top attribute to force SUT windows to the front even if normally prevented by Windows. To disable this feature, deactivate the option 'When raising windows, try to force them to the top'.
	A new check was added to directly check the icon of a component instead of having to capture the whole component and define a check region.

Bugs fixed:

	When loading the test-suites to restore the last session upon startup, QF-Test did not check for auto-save files.
	It was possible to start parallel tests in the same context by hitting [Return] at the right time. This could lead to confusing results in the run-log and even to exceptions.
	Swing applications started via Java WebStart have two independent application contexts, a user context and a security context. Unless the 'GUI engine' attribute of a 'SUT script' node explicitely defines the engine, the script is now always correctly executed in the user context and not in the security context.
	Special support for JSlider components was missing.
	The special variable ${qftest:runid} did not expand to the correct value if the runid was specified explicitly via -runid <id> in batch mode.
	A forced download triggered via a special PHP construct could cause Internet Explorer to crash.
	Clearing the cache for Firefox did not always work correctly.
	Trying to record a check for an XML document displayed in Internet Explorer could crash the browser.
	It is now possible to record checks for applets even when they are embedded in QF-Test's browser wrapper.
	The Groovy implementation of qf.logError was broken.
	In a workbench view with many open test-suites and some test-suite tabs scrolled out to the left, coordinates for dragging test-suite or node to different test-suites were off, leading to surprising results.
	QF-Test was leaking resources when taking Screenshots in an Eclipse/SWT SUT.
	Calling Display.asyncExec from a timer could crash Eclipse on Linux Gtk with 64bit.
	The order of SWT.Activation events could show very subtle differences between manual replay and automatic replay with soft events.
	A very long search string could mess up the size of the search and replace dialog.
	Markers in run-logs did not propagate beyond 'External log' nodes.
	Fallback to lazy binding did not work for not-yet-bound properties.

Main new features in QF-Test version 3.4

Following is an overview over the main new features that have been implemented for QF-Test version 3.4. For detailed information about the differences between QF-Test 3.2 and 3.4, please see the release notes below.

	Jython updated to version 2.5.1
	Groovy updated to version 1.7.6
	Support for Eclipse/SWT 3.6.2 and 3.7
	Web testing support for Firefox 4 and 5 as well as Internet Explorer 9
	Integration with Continuous Integration tool Hudson/Jenkins
	Fault-tolerant image checks
	Advanced search and replace
	Iteration ranges for data drivers
	Direct interaction with web servers
	Extended import functionality
	Debugger integrated into the workbench
	Ability to create a test-suite from a run-log
	Automatic splitting of run-logs and graceful handling of OutOfMemoryErrors
	Improved backwards/forwards navigation
	Improved pkgdoc and testdoc documentation
	Improved coordinate abstraction for mouse events
	Variables with default values and many additional special variables

Patch 3.4.0-p1 - June 27, 2011

New features:

Support was added for Firefox version 5 and Eclipse/SWT version 3.7 Indigo.

Version 3.4.0 - May 26, 2011

New features:

	Support for testing Eclipse/SWT version 3.7 was updated.
	All 'Data binder' nodes now have an attribute named 'Iteration ranges' that can be used to limit the iteration to a single element or one or more intervals, which is very convenient when implementing tests, continuing a test that had to be stopped or performing a retest with a limited subset.
	In workbench mode the current session with the currently open test-suites and the positions therein is now saved when QF-Test is closed and restored upon startup, provided the option "Restore last session on startup" is active. Also added to the workbench view were the ability to save all modified test-suites via the »File« menu and switching between recently active test-suites via [Control-F6].
	Moving backwards/forwards between selected nodes has been unified and should now work consistently even for special cases. After recording components you can jump directly to the recorded node by moving forwards, even if no node was actually recorded and the existing node is in a different test-suite.
	Filter capabilites were added to the search/replace result and other multi result dialogs.
	Marks can now also be set for nodes in a run-log, so multi-step search operations are now also available in run-logs.
	The quickstart wizard has improved support for launching applets in a browser and for setting the compatibility mode in internet explorer, backed by the procedures qfs.web.browser.general.startAsNewProcess and qfs.web.browser.settings.enableCompatibilityMode in the standard library qfs.qft.
	When started in batch mode with the command line argument -verbose [<level>], QF-Test now displays exception, error and warning counts for each 'Test-case'.
	The new package qfs.autowin.acrobat in the standard library qfs.qft contains methods to save a PDF document showing in an Acrobat window or to copy its text content to a file via the clipboard (Windows only). These are complemented by the qfs.utils.files package with procedures to verify the content of a text file.
	The new methods setProperty and getProperty in the generic scripting module qf can be used to associate arbitrary data with a GUI element, e.g. to cache some resolved feature or name values or other information that is too expensive to compute every time a resolver method is called.
	The TestRunListener API has been extended to provide the number of exceptions, errors and warnings that occurred below a given node in the TestRunEvent of a call to nodeExited.
	The AJAX resolver for GWT has been updated to version 1.2.0. It now filters irrelevant <DIV> nodes from the component hierarchy and has improved support for obfuscated GWT class names.

Bugs fixed:

	Early checking to handle OutOfMemoryErrors more gracefully could impact test execution speed significantly with JDK 1.5 on Windows. Tests should now run at full speed again with all JDK versions.
	Some of the advanced image check algorithms did not work correctly with JDK 1.5. All algorithms are now fully supported for JDK 1.5 and higher and return consistent results, independent of the JDK version, except for resizing, which is implemented differently in the various JDK versions and can lead to minor deviations in the resulting probability.
	The procedure qfs.shutil.copy in the standard library qfs.qft was broken. It now copies files and directories correctly, recursively if necessary and handles special cases like copying a directory into itself.
	'Check image' nodes now work correctly in Firefox 4.0.
	Scrolling an element visible was not always possible in IE9 and some special cases with a deeply nested hierarchy.
	In IE9 nodes were not highlighted properly and highlighting could interfere with recording image checks.
	Browser settings for IE9 did not always take effect.
	A download operation triggered from a popup window could crash the browser.
	A running test in the QF-Test daemon can now be interrupted and stopped cleanly via -stoprun even if the test is currently waiting for something.
	The structure of the JUnit XML report is now compatible with Atlassian Confluence.
	In Groovy the method rc.getProperties() was not implemented correctly.

Version 3.4M2 - March 10, 2011

New features:

	Support was added for testing Eclipse/SWT versions 3.6.2 and 3.7 Milestone 5.
	The search and replace dialogs are now integrated so it is possible to start with a search and continue replacing, based on the same criteria. Also, the replace dialog is no longer modal so it can be kept open while editing.
	Integration of the debugger functionality into the workbench view is now complete. The current variable bindings are displayed automatically whenever the debugger is entered unless the option "Show variable bindings automatically" is turned off.
	The keyboard shortcuts [Alt-1] through [Alt-9] navigate directly to the respective test-suite tabs in the workbench view.
	Pressing [F2] in the test-suite tree now moves focus directly to the details field that is most relevant for the display name of the node.
	The new 'Procedure' qfs.client.renameClient in the standard library qfs.qft can be used to rename an SUT client after it connects. This is especially useful for an SUT that launches multiple sub-processes that connect as client:2, client:3 etc.
	The new special variable $_{qftest:skipnode} can be used to skip execution of a node.
	Hyperlinks in the Eclipse FormText widget are now recognized as sub-items.
	Advanced image checks are now also available to scripts via the new run-context method rc.checkImageAdvanced.
	The various rc.check* methods in the run-context now return the result of the evaluated condition.
	The image of a 'Check image' node in the test-suite and the run-log as well as screenshot images can now be opened in a separate window by double-clicking the image (run-log only) or clicking the respective button above the image.
	There is a new utility module for scripting called qf. It contains variants of commonly useful methods like isInstance and the ability to add messages to the run-log even when no run-context is available, e.g. in TestRunListener, Checker or Resolver implementations.
	Templates for procedure recording have become more powerful. Procedures can now be created conditionally by placing a @CONDITION tag with an arbitrary Jython or Groovy script in the comment.
	The 'Procedure' qfs.web.browser.settings.setProxy in the standard library qfs.qft now also supports passing a URL for automatic proxy configuration.

Bugs fixed:

	Switching execution to a different node in the debugger could unbalance the parent/child relationships of the associated nodes in the run-log.
	Setting and unsetting options at script level could mess up default option values.
	Improved double click replay for SWT on Windows.
	Some selection events for SWT now receive special treatment for focus handling.
	Empty lines at beginning of an SWT StyledText widget are now handled correctly on Linux.
	Errors in nested test-cases are now listed correctly in the report.
	Proxy server settings for Internet Explorer 9 are now working correctly.
	Image checks are now working correctly with Firefox 4.
	Javascript and authentication dialogs are now working correctly in Firefox 4.
	QF-Test now issues a warning each time you first try to modify a write-protected test-suite.
	Doctags for 'Dependency' nodes are now listed correctly in pkgdoc documentation.

Version 3.4M1 - February 3, 2011

New features:

	Groovy was updated to version 1.7.6.
	QF-Test now actively monitors the memory available in the Java VM for running tests. When available memory runs low, the test is aborted cleanly and the run-log is saved. This also prevents extreme slow-down when memory gets tight.
	To prevent running out of memory in the first place, QF-Test now splits run-logs automatically by default, based on the option 'Minimum size for automatic splitting (kB)'.
	Managing test-suites in the workbench has been improved. The tab for the active suite is now shown with a bold font, you can drag tabs to re-order test-suites and double-clicking to the right of existing tabs creates a new suite.
	The search dialog now has additional options for limiting the search and it is also possible to search all loaded test-suites at once. To prevent visual overload for the common case, the search dialog can now be toggled between simple and advanced mode.
	The debugger has been mostly integrated into the workbench, except for the variables view which still requires the separate debugger window.
	There is a new recording mode for web applications that tells QF-Test to ignore individual events and record entire GET or POST requests instead, which are then represented by the new 'HTTP Request' node. Upon replay QF-Test interacts directly with the HTTP/HTTPS server without requiring a browser. This can be useful for load testing or mass data entry.
	For Swing and SWT components with an icon QF-Test now records a hash value derived from the icon image as an 'Extra feature'. As long as the icon remains stable this should lead to improved recognition, especially for toolbar buttons without a label. The default state of the image hash extra feature is 'Should match'. You can change it to 'Ignore' for components where the icon is not reliable and to 'Must match' for components where the image is the best way to identify the component.
	There is a new special variable syntax that allows specifying a default value for a variable. This is extremely useful for things like generic components or really every place where a variable has a reasonable default, because that default value is then tightly coupled with the variable lookup itself and doesn't have to be specified at 'Sequence' or test-suite level. The syntax is ${default:varname:defaultvalue}.
	Additional special variables are now available to retrieve the name of the currently executing 'Test-set' or 'Test-step' in addition to the already available 'Test-case' name. Also, counters were added for the number of test-cases executed so far, failures etc.
	Where available, the dialog showing the methods available for a component now also displays the results of calling those methods that represent Java Bean values.
	The option 'Log successful advanced image checks' can be activated in order to store the results of a successful advanced image check in the run-log. This can be useful while tuning the parameters for an advanced image check algorithm.
	The new command line option -verbose [<level>] for batch execution with or without accessing a daemon causes QF-Test to print progress information during a test-run.
	The new 'Procedure' qfs.utils.dumpHeap in the standard library qfs.qft can be used to trigger a heap dump for the Java VM of the SUT when it uses too much memory. This can be detected with the help of the 'Procedure' qfs.utils.logMemory which now returns the amount of memory used.

Bugs fixed:

	When replaying mouse events without coordinates on items, QF-Test no longer targets the center of the item, but the position 5 pixels from the left, vertically centered. Apparently some events are sensitive to whether the click hits the text of an item and not just its bounding box, so short items in a broad column could cause problems.
	Calling stored procedures in a database could fail using the default Jython bindings.
	If the debugger was turned off for normal use, debugger-stepping after hitting a breakpoint could occasionally fail so the test simply continued.
	A positive search result for the type of an extra feature could cause an exception.
	Blank characters in names for saving external run-logs are now turned into an underscore because Windows doesn't handle trailing blanks in directory names gracefully.
	The input file for a 'CSV data file' node is now closed immediately after reading the data.
	Recursive procedure calls could cause problems when importing procedures or analyzing references.

Version 3.3.0 - November 30, 2010

Incompatible changes:

	Support for Java 1.4.2 had to be dropped in order to update Jython and Groovy. QF-Test now requires at least Java version 1.5.
	Interaction between Jython and Java has changed slightly between Jython 2.1 and 2.5 so occasionally scripts that worked in QF-Test may now fail or behave differently. We try to maintain compatibility where possible and provide a list of things to look out for.

New features:

	Jython was updated to version 2.5.1. Most scripts should continue to run without problems. Incompatibilities may arise due to the changes in character encoding and because Java strings are now uniformly treated as unicode strings in Jython.
	Groovy was updated to version 1.7.2.
	Support was added for Eclipse/SWT version 3.6.1.
	Experimental support was added for Firefox 4 and Internet Explorer 9.
	QF-Test now integrates well with the increasingly popular Continuous Integration system Hudson CI. This includes JUnit as an additional target format for QF-Test reports.
	Image checks are no longer restricted to testing for exact pixel-by-pixel equality. Several flexible algorithms now allow image checks with a configurable degree of deviation from the expected image.
	Besides components you can now also import test-cases and procedures from one test-suite into another.
	When coordinates for mouse events are irrelevant, QF-Test now removes them altogether instead of setting them to (1,1). Upon replay the event is then aimed at the center of the target component or item.
	By combining search and replace operations with marks it is now possible to narrow searches in several steps, giving very fine-grained control over where to perform replacements.
	Moving backwards or forwards between selected nodes now works across test-suites and even run-logs. Thus a single mouse click or keystroke is now sufficient to return when following a procedure call to its target in a different test-suite or when jumping from a run-log to the respective node in the suite.
	Several new toolbar buttons were added and all the buttons now look much better. The menu structure also was cleaned up with operations on tree nodes now collected under the new »Operations« menu, among them many new node transformations. Also, several new keyboard shortcuts were introduced.
	The layout of pkgdoc and testdoc documentation now equals that of reports, including node icons and support for test-steps. These can be configured via the command line arguments -testdoc.teststeps, -testdoc.nodeicons and -pkgdoc.nodeicons or the respective options in the dialog for interactive creation of those documents.
	'Skipped' and 'Not implemented' tests can now be ignored in the report via the command line argument -report.ignoreskipped or the respective option in the interactive dialog.
	Daemon tests can now be stopped at the command line via the new argument -stoprun.
	The comment of a 'Procedure', 'Test-case' or 'Dependency' node can now be filled automatically with doctags for parameters, characteristic variables and author by right-clicking the node and selecting »Additional node transactions«-»Fill comment« from the context menu.
	The TestRunListener API now also provides information about whether a test-run was terminated normally (with or without exceptions) or aborted.
	The new method rc.callTest in the run context can be used to call a 'Test-case' or 'Test-set'. In contrast to callProcedure, callTest can only be called from 'Server script' nodes. It returns the exit status of the called test.
	The content of QF-Test configuration files is now sorted which significantly improves handling of those files in a version control system.
	Several procedures were added to or refined in the standard library qfs.qft. Its documentation is now also available online and indexed by the full text search engine on the QFS website. A link to the online text search covering all technical material is provided via the menu »Help«-»Online search...«.
	It is now possible to create a test-suite from a run-log.
	It is now possible to pass GUI elements from an SUT script as individual targets to QF-Test's component recognition mechanism via the new run context method overrideElement.

Bugs fixed:

	When started from QF-Test, Internet Explorer 8 always ran in IE7 compatibility mode. This is no longer the case. Compatibility mode can now be actively selected in the quickstart wizard or via the procedure qfs.web.browser.settings.doStartupSettings in the standard library qfs.qft.
	On Linux QF-Test could sometimes crash SWT/Eclipse applications with disposed Combo popups.
	Test replay for Eclipse/SWT on Linux is now much more robust, especially when handling menus.
	Trailing empty newlines in StyledText widgets are now handled correctly on Linux.
	Creating reports with many screenshots now requires much less memory.
	Downloads in Internet Explorer should now work consistently.
	In case of overlapping nodes with transparent nodes on top QF-Test could pick the wrong node for check recording. It now lets the browser make the decision instead. Switching back to the old algorithm can be done via the option 'Let the browser determine the target element for check recording'.
	A memory leak in QF-Test resulting from SUT clients printing a lot of output to the shared terminal in workbench view was plugged.

Version 3.2.2 - July 15, 2010

New features:

	Support was added for testing Eclipse/SWT version 3.6 "Helios".
	Multiple file selection in an SWT FileDialog can now be recorded and replayed.

Bugs fixed:

	Replay performance could suffer severely if the user's home directory was located on a network drive.
	A workaround for an AWT weakness introduced in QF-Test 3.2.1 could impact replay performance for AWT/Swing.
	Yet another performance problem resulting from very large hierarchies of Swing menus has also been fixed.
	Several special cases for uploading files with Internet Explorer are now handled correctly.
	Downloading files with Internet Explorer now works correctly in case of HTTP redirection.

Version 3.2.1 - May 11, 2010

Bugs fixed:

	With a 1.4 JDK in interactive mode QF-Test would only start if given a file to open.
	Web-testing now works again on Windows 2000.
	The QF-Test debugger used to show duplicate entries on the variable stack for 'Procedure call' and 'Test call' nodes.
	There was a memory leak in connection with interactive editing of test-suites with 'Check image' nodes for large images.
	In very special cases an import of components into their own suite could result in incorrectly updated references.
	Saving entries for zipped run-logs could occasionally fail on Windows due to a problem in the underlying library. QF-Test now works around that.
	Two problems were fixed for SUTs with more than one GUI engine, e.g. a WebStart application or combined SWT/Swing. In such a case QF-Test used to log more than one screenshot of the main screen after an error and sometimes connections from the second engine could fail.
	The upload dialog for web pages was not handled correctly for all browser versions and the procedure qfs.web.input.fileUpload in the standard library qfs.qft did not handle all cases correctly.
	In recent QF-Test versions there was an inofficial global Jython run-context rc available in addition to the local one. In special cases this could cause subtle namespace problems, so the (still inofficial) global version was renamed to __rc.
	Several minor problems and occasional exceptions in connection with the result list dialogs were fixed.

New features in QF-Test version 3.2

Following is an overview over the main new features that have been implemented for QF-Test version 3.2. For detailed information about the differences between QF-Test 3.0 and 3.2, please see the release notes for versions 3.2.0 and 3.1.x.

	Support for Web testing in combination with Java applets
	JDK instrumentation is no longer required
	Workbench view for a modern user interface
	'Test-step' nodes for detailed reports
	Refactoring support with automatic update of references
	Integration with the open source test management tool TestLink
	New 'Best Practices' part in the manual
	Eclipse/SWT support updated for the latest versions including GEF
	Incremental search in all of QF-Test's text fields
	Result list dialogs for complex operations
	Component method spy
	Options can be set at script level
	Relative and test-suite-local procedures
	New-style 'Check' nodes and support for custom Checkers
	Static validation of references
	Automatic parameterization of procedures
	Persistent bookmarks and transient marks
	New APIs: ItemResolver, Checker, ClassNameResolver and TestRunListener

Version 3.2.0 - March 25, 2010

New features:

	Support for Eclipse/SWT has been brought up-to-date with the current versions 3.5.2 and 3.6M6.
	Swing components are no longer highlighted by exchanging their foreground and background color but by painting a rectangle, similar to SWT and Web. Target sub-items are now highlighted as well.
	The new 'Boolean check' node with 'Check type identifier' attribute 'visible' is similar to 'Wait for component to appear' node, but in contrast to the latter is is listed in the report as a check and it has different default settings for the result handling.
	The most recently accessed component and possibly item in the SUT can now be retrieved in Jython and Groovy 'SUT script' nodes via rc.getLastComponent() and rc.getLastItem().
	Selecting a file for upload in a web form can now be recorded and replayed directly on Windows. For a cross-platform implementation it is still necessary to call the procedure qfs.web.input.fileUpload from the standard library qfs.qft.
	Additional features were added to the result lists for complex operations. For example, you can now perform actual replacements in the result list for the replace dialog or you can update multiple failed checks at once from the error list in a run-log. Besides, the result list for an operation that takes time is now filled continuously and the operation can be aborted if necessary.
	Recent search expressions are now saved in the user configuration.
	'Dependency' nodes inside 'Packages' are now listed in the pkgdoc documentation unless disabled via the command line argument -pkgdoc.dependencies or the respective interactive option.

Bugs fixed:

	Right-clicks for Javascript-based popup menus no longer need to be replayed as hard events in order to trigger the popup reliably.
	When recording checks the z-order of DOM nodes is now properly taken into account so that QF-Test no longer records nodes that are actually in the background, covered by other nodes.

Version 3.1.4 - February 9, 2010

New features:

	It is no longer necessary to instrument standard JDKs or JREs, though it doesn't hurt to keep existing instrumentation.
	Support was added for the latest Eclipse/SWT version 3.6.
	QF-Test now supports Firefox 3.6.
	The new part II 'Best Practices' in the manual describes recommended procedures based on lessons learned from several customer projects and user feedback that will hopefully help you to find the best strategy for using QF-Test in your own projects.
	Reports have been restructured and further extended. 'Check' nodes now have an optional 'Name' attribute for their representation in the report. Besides, the performance of report creation has been drastically improved and its memory impact reduced.
	The tree display in run-logs now shows expanded values instead of variables. This can be toggled via the »View« menu.
	Compactification for run-logs is now more efficient and special placeholder log entries show where nodes were removed during compactification.
	Support for the Eclipse Graphical Editing Framework (GEF) is now included by default and works correctly with multiple concurrent GEF editors.
	The new procedure qfs.web.input.fileUpload in the standard library qfs.qft handles INPUT:FILE nodes for the various browser variants.
	TestLink integration has been simplified.
	For Swing and SWT tables and multi-column trees it is now possible to check a whole row at once.
	ExtraFeatureResolvers can now also be used for SWT and Swing.
	The standard library qfs.qft now includes wrapper procedures for the autowin module in the package qfs.autowin. These can be used to check for the existence of a window and to close it on Windows systems.
	The runid of the current test-run now is now available via ${qftest:runid}.
	Special variables or properties of type group:name can now be used for 'Characteristic variables'.
	For nodes like 'Test-step' or 'Test-set' with several tabs for variable bindings the number of bound variables is now displayed in the tab which makes it much easier to locate those bindings.

Bugs fixed:

	Several causes for IE crashes were fixed.
	Wait for absence now works correctly for invisible nodes.
	QF-Test doesn't get blocked as easily when the SUT generates a lot of output and the JTextArea for the shared terminal cannot keep up.
	The internal variable stack of QF-Test could get messed up in some special cases.
	The procedure builder now correctly handles '_' characters in package or class names.

Version 3.1.3 - December 1, 2009

New features:

	The new 'Test-step' node can be used to provide further structure and document tests below 'Test-cases'. 'Test-steps' are listed individually in the report unless deactivated with -report.teststeps. Similar effects can be achieved by marking arbitrary nodes with the doctag @teststep, optionally followed by a name for the step.
	The report now also can list checks (off by default) and warnings (excluding component recognition warnings, on by default) integrated into the test-step listing and also shows icons for nodes. The respective settings can be changed via -report.checks, -report.warnings and -report.nodeicons in batch mode or the respective options in the dialog for interactive report creation.
	In the workbench view there is now a context menu for the test-suite tabs.
	All 'Comment' attributes can now be edited in an external editor similar to scripts.
	There are new toolbar buttons in the run-log window for navigating to the next/previous error or warning.
	The incremental search feature now has more intelligent scrolling for hits and also allows scrolling the component being searched via the [Up], [Down], [Page up] and [Page down] keys.
	'Dependencies' can now optionally be resolved in a namespace which allows having different sets of dependencies for different parts of a test so that the dependencies don't interfere with each other. See the 'Dependency namespace' attribute for details.

Bugs fixed:

	Due to the JIDE libraries used, the new workbench view does not work with JDK 1.4. It now gets deactivated automatically when starting QF-Test with a 1.4 JDK.
	Several exceptions caused by the new workbench view and the multi result dialogs were fixed.
	Text checks for SWT Group widgets were missing.
	Fixed automatic scrolling of web pages for image checks and replay of hard events.
	The popup sub-menu for setting and removing marks was sometimes empty.

Patch 3.1.2-p1 - October 20, 2009

Bugs fixed:

	In the new workbench view QF-Test could lose track about which suites are currently loaded, causing a suite to get loaded more than once, leading to confusing behavior and sometimes an exception.
	The new mechanism for on-the-fly SWT instrumentation only handled one SWT plugin, so that in case of a multi-platform Eclipse based application with several SWT plugins the correct plugin for the current platform could remain uninstrumented.
	Calling stopRun could cause the shared run-context in the QF-Test daemon to become unusable.

Version 3.1.2 - October 15, 2009

New features:

	Support was added for the latest Eclipse/SWT version 3.5.1.
	We have started work on special support for several AJAX web toolkits, including Ext GWT, Ext JS, GWT, qooxdoo and RAP. Instead of deeply nested hierarchies of <DIV> nodes QF-Test now operates in terms of the actual rich interface of buttons, labels, trees etc. For information on how to activate this feature in its current state, please get in touch with QFS.
	QF-Test now comes with support for interacting with the open source test management tool TestLink.
	By default QF-Test now starts with a "Workbench View" where all open test-suites are arranged as tabs in a single window. This feature is work in progress. In case it causes problems or if you prefer the old style, you can switch back to multiple windows via the »View« menu.
	The new function »Analyze references« in the context menu of a node can be used to statically validate the correctness of a test-suite by verifying the existence of the target nodes of 'Procedure calls', 'Test calls', 'Dependency references' and references to 'Components'.
	Operations that search for or result in changes to multiple nodes now bring up a mass result dialog listing the affected nodes. You can either navigate to specific nodes directly from that dialog or use it to set marks on some or all of the resulting nodes.
	Nodes in a test-suite can now be marked with markers in four different colors for easier reference and quick navigation. Marks can be set either individually via the menu »Edit«-»Marks« or the context menu, or from a mass result dialog.
	All text fields, terminals etc. in QF-Test's user interface can now be searched locally by pressing [Ctrl-F] when the field owns the keyboard focus.
	The new Component Method Spy retrieves and displays fields and methods of Swing or SWT components as well as the Javascript properties of a DOM node in the web client.
	After recording a new sequence and transforming it into a 'Procedure' you can select »Parameterize node« from its context menu to tell QF-Test to automatically extract parameters for it from the contained nodes.
	The Eclipse/SWT instrumentation performed by the 'Procedure' qfs.qft#qfs.swt.instrument.setup now operates on-the-fly, building a new plugin from an existing plugin and the QF-Test specific add-ons. Thus it is now possible to instrument custom SWT plugins or plugins for unofficial Eclipse versions for which QF-Test doesn't provide an exactly matching instrumented plugin.
	Event nodes can now be transformed into a 'Wait for component to appear' node.
	The keyboard shortcut for the 'Redo' operation was changed from [Ctrl-R] to [Ctrl-Y]. Over the last years the latter has become a de facto standard.
	The warning about nested 'Test-case' execution can now be suppressed via the option "Log warning for nested test-cases".
	Where scrolling a web page is required to perform a hard event or an image check, QF-Test now tries to scroll a web page itself via Javascript calls instead of relying on the browser's internal mechanism which often completely messes up the display.
	Component recognition based on geometry can now be suppressed by setting the coordinate and size attributes of a 'Component' to '-'.
	It is now possible to specify just the 'Class count' or 'Class index' of a 'Component' node and ignore the other.

Bugs fixed:

	Using the replace dialog to replace a regular expression with an empty string caused an exception.
	The function to make 'Procedure calls' or 'Dependency references' relative did not work reliably.
	When moving several 'Procedures', the automatic correction of 'Procedure calls' could fail in some special cases.
	Procedure recording now works correctly for custom classes with a special ClassLoader.
	When creating reports, warnings can be suppressed entirely by specifying -report.warnings=false for batch mode or deactivating the respective option in the dialog for interactive report generation.
	Recording a click on an editable SWT table cell could cause the cell editor to get activated in the wrong table row.
	'Fetch index' nodes were not implemented for SWT.
	Removing or marking unused components, importing and other complex component operations could fail if the test-suite contained a 'Data table' node with no columns or empty table cells.

Add-on for Eclipse/SWT 3.5.1 - September 30, 2009

New features:

This add-on adds support for Eclipse/SWT 3.5.1 "Galileo".

Version 3.1.1 - July 28, 2009

Incompatible changes:

The changes to the check system described below are fully backwards compatible. However, once a test-suite has been saved with QF-Test version 3.1.1 it cannot be run correctly with an older QF-Test version except for QF-Test version 3.0.3 which has explicit bidirectional compatibility support for this feature. Once you introduce QF-Test version 3.1.1 you can safely continue to work with QF-Test 3.0.3, but you should not use older versions like 3.1.0 or 3.0.2.

New features:

	The transition to the new check mechanism is now complete. Custom Checkers are now fully documented and can be created for all check data types. All 'Check' nodes now have a 'Check type identifier' attribute and the old 'Check editable', 'Check enabled' and 'Check selected' nodes were replaced with 'Boolean check' nodes with a matching 'Check type identifier'.
	It is now possible to bookmark test-suites, run-logs and individual nodes in a test-suite so they are directly accessible via the new sub-menu »File«-»Bookmarks«.
	The new ClassNameResolver - though also useful for SWT and Swing applications - is an important milestone for AJAX web testing. With its help and the new pseudo class hierarchy for DOM nodes a deeply nested hierarchy of DIV nodes can be turned into a concise component hierarchy with expressive class names that vastly improve component recognition.
	'Procedures' can now automatically be parameterized via the sub-menu »Parameterize node« in the popup menu for the 'Procedure' node. This feature is work in progress and will need more fine-tuning so we hope for your feedback.
	When creating testdoc documentation it is now possible to follow 'Test call' nodes and include the target 'Test-case', 'Test-set' or test-suite into the generated documentation by activating the respective option in the dialog for interactive use or via the command line argument -testdoc.followcalls in batch mode.
	Several more variants for converting and (un)packing nodes were added.

Bugs fixed:

	Several problems with split run-logs were fixed. Log entries could get lost in compact logs, when the split run-log was not saved as a zip file or when split run-logs were disabled but the attribute 'Name for separate run-log' was defined.
	DaemonTestRunListeners did not work correctly in asynchronous mode.
	When creating a report, screenshots for failed 'Check image' nodes were always saved regardless of whether the failed check actually represented an error or was just a part of a conditional construct.
	Use of the new Java regular expressions in QF-Test's replace dialog was not correct so that some replacements based on regexps were not possible.
	When determining uniqueness of "id" attributes for use as names, QF-Test was too strict so that in some cases potential names were not assigned.
	Instrumenting an OpenJDK directory now works.
	Waiting for a new document did not handle the window name attribute correctly.
	Waiting for a download to finish always waited for the full timeout.
	In very rare cases replaying text input could crash an SWT application on Linux.
	Option settings from rc.setOption() in the SUT were reset whenever options were changed in the option dialog.
	Event replay for SWT MenuItems with Gtk 2.4 on RedHat Enterprise Linux 4 was not reliable.
	The ktable and gef Jython modules for the respective SWT components were not correctly updated for the new ItemResolver and Checker mechanisms.
	QF-Test now properly discerns the most modal SWT Shell in case multiple modal Shells are opened simultaneously.

Patch 3.1.0-p1 - June 25, 2009

New features:

The primary cause for this patch release is its support for Eclipse 3.5 "Galileo".

Version 3.1.0 - May 19, 2009

Incompatible changes:

QF-Test now uses standard Java regular expressions by default which are much more powerful than the old GNU regexp package. One major difference is that '{' and '}' are special characters for Java regexps and may need to be escaped with a '\'. If necessary, you can switch back to the old GNU regexps via the option "Use old-style GNU regexps (from before QF-Test version 3.1)".

In contrast to Mozilla, Internet Explorer reports a local navigation to an anchor on the current page as a normal navigation event, so up to now a 'Wait for document to load' node was recorded which failed when replaying the same test on Mozilla. QF-Test now treats this situation uniformly by ignoring the extra navigation from IE so no 'Wait for document to load' node is recorded with either browser. Consequently, old 'Wait for document to load' nodes for local navigation will now fail and need to be removed from your tests.

New features:

	QF-Test's embedded browser for web testing now supports Java applets with JDK 1.6.0_10 and up, so QF-Test can now test the interaction of a web application with its embedded Java applet.
	Many of QF-Test's options, specifically those with an impact on test execution, can now be changed at run-time via the new run-context method rc.setOption.
	When moving 'Procedure', 'Dependency' or 'Test-case' nodes, all references to these nodes can now be updated automatically. Similar updates are now performed when changing the name of such nodes or 'Component' IDs via the replace dialog. As an additional feature the new special Variable group ${id:...} and the run-context method rc.id can be used to mark arbitrary text as the definition of a component id that will be updated whenever the referenced component changes.
	'Procedure calls' from one 'Procedure' to another 'Procedure' in the same 'Package' hierarchy can now use a relative procedure name. Also, 'Procedures' and 'Packages' can now be made local to their test-suite by prefixing the name with an '_'.
	The selection dialog for a 'Procedure', 'Dependency' or 'Test-case' node now also shows the comment of the selected node. Also, it is now possible to select multiple nodes in order to insert multiple calls at once.
	A 'Procedure call' to a specific 'Procedure' can now be created by copying the 'Procedure' node and pasting it to a place where insertion of the 'Procedure' is not allowed, which is basically everywhere a 'Procedure call' is allowed except for the 'Extras' node. Dragging and dropping a 'Procedure' with the same effect is also supported.
	It is now possible to interact directly with the Javascript engine of the browser and execute arbitrary scripts via the methods DomNode.toJs, DomNode.evalJs and DocumentNode.evalJs.
	When an error or exception occurs, the run-log now contains a stacktrace node, showing all nodes on the execution stack including their variable bindings, similar to the variables view in the debugger window.
	There is a new API for various kinds of resolvers and checkers, including an ItemResolver interface for implementing your own sub-items of custom components and a Checker interface for custom checks that can be captured and replayed just like QF-Test's standard checks.
	The new TestRunListener API can be used to monitor a test-run from a script. Its variant DaemonTestRunListener supports the same remotely via QF-Test's daemon API.
	Support for Eclipse/SWT now includes the latest Eclipse milestone 3.5M7.
	The run-context method rc.callProcedure now also works from an 'SUT script' in a web application.
	Execution of a 'Test-case' or 'Test-set' can now be aborted or skipped via the run-context methods rc.skipTestCase, rc.skipTestSet, rc.stopTestCase and rc.stopTestSet, e.g. to prevent wasted time and irrelevant error messages in case a dependency cannot be resolved.
	The name of the currently executing 'Test-case' is now available via the special variables ${qftest:testcase.name}, ${qftest:testcase.qname} and ${qftest:testcase.reportname}.
	Via the new methods setProperty and getProperty in the class Node from QF-Test's pseudo DOM API for web applications, arbitrary data can be associated with DOM nodes without affecting the web application.
	The new command line argument -ping for qftest -batch -calldaemon can be used to check whether a QF-Test daemon is up and running.

Bugs fixed:

	For SUTs running on Mac OS X QF-Test disables use of the screen menu bar to allow proper capture/replay of menu actions.
	Activating an engine license in the option dialog now works as expected.
	When recording procedures, a new root 'Package' is always created instead of merging the procedures into a possibly existing 'Package' named "procbuilder".

Version 3.0.3 - July 28, 2009

New features:

Handling of 'Check' nodes is reverse-compatible with changes in QF-Test version 3.1.1 in which checks underwent a major overhaul. QF-Test version 3.0.3 can load, execute and save test-suites created with version 3.1.1 without losing information.

Bugs fixed:

	Several problems with split run-logs were fixed. Log entries could get lost in compact logs, when the split run-log was not saved as a zip file or when split run-logs were disabled but the attribute 'Name for separate run-log' was defined.
	When determining uniqueness of "id" attributes for use as names, QF-Test was too strict so that in some cases potential names were not assigned.
	JDK instrumentation now also works for an OpenJDK directory.
	Waiting for a new document did not handle the 'Name of the browser window' attribute correctly.
	Waiting for a download to finish always consumed the full timeout.
	Activating an engine license in the option dialog now works as expected.
	In very rare cases replaying text input could crash an SWT application on Linux.
	For SUTs running on Mac OS X QF-Test disables use of the screen menu bar to allow proper capture/replay of menu actions.

Patch 3.0.2-p2 - June 25, 2009

New features:

The primary cause for this patch release is its support for Eclipse 3.5 "Galileo".

Bugs fixed:

	Event replay for SWT MenuItems with Gtk 2.4 on RedHat Enterprise Linux 4 was not reliable.
	QF-Test now properly discerns the most modal SWT Shell in case multiple modal Shells are opened simultaneously.

Patch 3.0.2-p1 - April 14, 2009

Bugs fixed:

	In QF-Test version 3.0.2 hard TAB characters in Jython or Groovy scripts were replaced with blanks when showing the script, immediately causing the respective node to become modified. This change has been taken back.
	With JDK 1.6 some of QF-Test's dialogs were not displayed with the correct size.

Version 3.0.2 - March 27, 2009

New features:

	Support was added for Eclipse/SWT versions 3.5M6 and 3.4.2.
	The method callProcedure was added to the class DaemonRunContext of QF-Test's daemon API. This method and the existing method runTest now also support direct passing of parameters to this call.

Bugs fixed:

	'Wait for document to load' nodes for web clients did not work reliably in case of multiple browser windows.
	Checking the enabled/disabled state of components is now also implemented for web clients.
	Capture and replay of MOVED and SIZED events is now also supported for browser windows of web clients.
	QF-Test now performs regular explicit garbage collections to prevent running out of memory in the PermGen Space, a special Java memory region that causes trouble for a lot of Java applications. For the SUT these extra garbage collections can be turned off via the option 'Automatically perform garbage collection in the SUT'.
	Disposal and recreation of an SWT Display is now handled correctly.
	Custom SWT Widgets implemented as an anonymous local class were not treated correctly, possibly leading to a ComponentNotFoundException.
	For checks with a very long timeout deadlock detection is now adjusted accordingly.
	The demonstration test-suite for integration with the Imbus TestBench was updated to the latest TestBench version and now handles split run-logs correctly.

Version 3.0.1 - January 23, 2009

New features:

	The Jython and Groovy run context rc now provides the convenience methods getBool, getInt, getNum and getString for simplifying lookup of typed variable values in scripts.
	It is now possible to specify proxy settings for a web SUT directly in the quickstart wizard or via the procedure qfs.web.browser.settings.setProxy in the standard library test-suite qfs.qft.
	An entry for Swing JInternalFrame components has been added to the template file for automatic procedure creation.

Bugs fixed:

	Web-testing with Mozilla on Linux no longer requires using a 1.4 JDK.
	Using the "Update component" feature for web pages with deeply nested tables could lead to duplicate component IDs.
	Left-to-right binding in 'Data table' and other data nodes has been fixed.
	SQL statements in 'Database' nodes did not work with Oracle databases, due to an automatically appended ';' character. This has been fixed. If your database engine requires that SQL statements are terminated with a ';' you have to specify it yourself.
	The procedure qfs.swt.instrument.setup now always picks the SWT plugin with an exact version match for instrumentation, if available. This is important for SWT 3.4, where versions 3.4.0 and 3.4.1 are mutually incompatible.
	When recording procedures, a '$' character displayed in one of the fields of the SUT could lead to an exception.
	Groovy class loading and script parsing in the SUT should now be faster, in some cases - specifically applets - drastically.
	A caught exception in the 'Finally' node of a 'Try' node could mess up the state of an exception caught earlier in a 'Catch' node.
	The demo test-suite swt_addressbook.qft was still based on SWT 3.1, support for which is no longer distributed with QF-Test. The suite is now based on SWT 3.4.

Version 3.0.0 - December 5, 2008

Major new features:

	Cross-platform and cross-browser web testing with web2.0/AJAX support.
	Integration with HP/Mercury Quality Center and Imbus TestBench.
	Groovy as a second scripting language.
	New Jython version 2.2.1.
	Improved Jython and Groovy script editor with syntax-highlighting.
	Automated creation of basic procedures.
	Extended and unified functionality for 'Check...' and 'Wait...' nodes.
	Direct support for negated checks.
	Split run-logs enable long-running tests without memory problems.
	64-bit support for SWT on Windows and Linux.
	Extensions and an external driver for the daemon API.
	New concept: "Expected failure" for 'Test-case' nodes.
	Directing and documenting manual tests.
	Extended undo support.
	Immediate vs. lazy variable bindings.
	Image API for creating and saving screenshots.
	Crypting of text input for password fields.
	Recursive sorting and transformation of nodes.
	Revised search with highlighting of results.
	Extensions to the standard library test-suite qfs.qft.
	New tutorial chapters for advanced features.

The life-cycle of a QF-Test version

There are different ways of handling the life-cycle of software that, like QF-Test, is widely used in production environments while continuously being maintained, improved and extended. In the past, QF-Test versions were separated into development branches (with odd medium numbers) and production branches (with even medium numbers), but this approach was difficult to comprehend and sometimes led to misunderstandings. Starting with QF-Test version 3.4 the life-cycle for a QF-Test version is as follows:

Milestones

Development never stops. While the latest version is still being maintained, a new version evolves. When a number of new features has matured, the first "Milestone" is released, for example 3.4M1. More milestone releases like 3.4M2 with additional features may follow.

The version of QF-Test currently under development is the one that undergoes the most testing and generally is the most stable. Thus a milestone is not a "use-at-your-own-risk" beta version but a fully supported release. The main difference between a milestone and the "dot-zero" release is that a milestone is not yet feature-complete. Consequently, the release of the first milestone marks the release of a new medium or major version with respect to maintenance and upgrade conditions.

A special feature of milestone releases is that, during a limited time span, they can be used with a license valid only for the previous release, giving users without maintenance a chance to evaluate the new version without committing to anything.

"Dot-zero" and maintenance releases

When the main feature set for a version is complete, a "dot-zero" version is released, 3.4.0 in this example. Having been available in the form of milestones for some time, QF-Test "dot-zero" releases are quite mature ane the reasoning above about testing and stability of the latest development version also applies.

Maintenance releases like version 3.4.1 are released at irregular intervals, mostly to fix bugs but with the occasional new feature. In case of a severe bug a maintenance version will be released as soon as the bug is fixed.

Patches

Sometimes small but annoying bugs are discovered soon after a release. They don't warrant a new release but are important enough to take care of. In such a case patch versions like 3.4.0-p1 are released to provide the latest changes to those who need them. Patches are announced on the QF-Test mailing list only (more about the QF-Test mailing list).

During development of new features or bug-fixes it is often necessary to solicit feedback from users. To that end, experimental patches are sometimes released to individual customers or, if of general interest, via the mailing list.

A patch is a very small set of changes relative to a given QF-Test version. It is provided in the form of a ZIP archive which is extracted over an existing QF-Test installation. There is only one ZIP file for all operating systems and installing a patch causes far less overhead than upgrading to a new QF-Test release.

A patch is always a temporary solution. The next full maintenance release will include all working fixes from patches since the last release and thus supersede previous patches.

We always run a full set of tests against official patches. However, due to the shorter release cycles patches cannot be quite as well tested as full releases. If you encounter problems after installing a patch, you can usually disable the patch by removing the file .../qftest/qftest-x.y.z/qflib/patch.jar. Please report all problems with patches to qftest-bug@qfs.de or to the mailing list.