The habitual Desktop Environment (...

The habitual Desktop Environment (CDE) test team was given the challenge of designing and organizing the unfolding of an automated regression ordeal suite for CDE 1.0 composings This project contained all the usual question s involved in designing and creating a exhibition suite for a large and mixed desktop environment, plus a number of challenges that came up because the disentanglement was a joint project between four different companies. The article forward page 50 provides some background about the creation of the CDE frame and the four companies involved.

Several goals for the proofs were developed early in the plot that influenced the design of the experiment suite, the choice of software testing tools, and the implementation of the testing proces The rationale for about of these objectives will be discloseed in more detail in later parts of this article.

In setting the goals for the proof suites, we determined that they had to be:

* Easy to disclose There wasn't a lot of time in the CDE schedule. * Robust. proofs shouldn't start failing because of a minor visual change, a different baptismal vessel selection, and so on. * Reliable. proofs should find the defects located in the digest but they should not report a destitution when there isn't one. * Consistent operation. equable though four companies would be developing proofs individuals at each company had to be able to haste the entire test suite, not just the parts written at avow site. It was not acceptable to make someone learn four different ways to scamper tests. * Consistent design and implementation. At the period of the joint development, personnel at each site would secure engineering responsibility for the whole suite, including those portions written at other companies. It was important that the proofs be written such that an experienced criterion engineer at one company could easily understand the internal workings of the proofs written at other sites. * Portable. The proof suite not only had to go proceed on each of four regard platforms (one from each company), yet also had to be easily ported to other nonreference platforms. * Maintainable. The touchstone suite was not just for the CDE 10 sample implementation, still was to be the basis for company works or later versions of CDE* It had to be relatively painless to update the criterions if they had defects, enhance the proof suite for new functionality, and in the way that on.

CDE Components

The CDE composings were the software under exhibition (SUT) for this project. From a testing point of view, CDE constitutings can be divided into three types: CDE API (application programming interface) constituents CDE GUI (graphical user interface) elements and graphical API components. CDE API constituents have no effect on the desktop, that is, no visual impact. An example of a CDE API composing is the ToolTalk[R] API.** CDE GUI constituents present desktop graphics that can be manipulated by way of the user, resulting in graphical changes onward the desktop. Examples of CDE GUI ingredients are the file manager and the icon editor. Graphical API constitutings consist of a library of functions like a standard API, do not include that calls to these functions usually do conclusion in visual changes on the desktop. Examples of this prototype of CDE component include the DtHelp library and the DtWidget library.

Tools, Utilities, and Environment

This section describes the tools choiceed for the test suite, the utilities created to augment the tools, and the arrangement of the test implementation and operating environment.

Synlib

The Synlib API was the same of the most important tools used in the progression in a continuously ascending gradation of the test suite. Synlib is a C-language interface that provides a means to simulate programmatically the actions of a user with a GUI. It also contains features that allow the touchstone program to monitor the state of the desktop, so as watching for windows to appear or disappear, checking the title of a window, or checking other desktop features.

Synlib was used to exhibit all the tests that required either manipulation of correlates on the desktop or verification by way of checking the state of destination; recipients on the desktop. This turn rounded out to be the majority of the examples The CDE test team chose Synlib as the GUI trial tool because: * Synlib is portable. The simply requirement for using Synlib is that the server be equipped with either the XTEST or the XTestExtension1 extensions,(*) which are used from Synlib to do such things as simulate keyboard and mouse marked occurrences Synlib was functional on all partner arrangements * Synlib is unencumbered. Synlib was cause to growed at HP and was made available in source digest form to each of the CD partners without charge. * example development could begin immediately. Engineers could begin writing proofs based on the components' specification Since Synlib is not a record-and-playback process a functioning CDE component is not required for initial example development. * Synlib reduces staff upon image capture and comparison. Many of the earlier tools for testing GUI composings use a method that hangs heavily on capturing correct shield images during an initial phase, then comparing that image to the guard in a later test stream Experience shows that this process is quite fragile and likely to about many false failure reports. With Synlib, other ways of checking for cloak state make the need for image capture and comparison greatly less important. * Synlib contains a feature that allows position independent manipulation of desktop items. A special data file called a focus map is created that defines the keyboard traversal for setting focus to items within a window (Fig. 1)(**) With the focus map, the proof program can set the keyboard focus to a particular button or paragraph field without needing to know its physical location in the window. Since the focus map is a separate data file that is read by the agency of the test program, changes in the component part that result in changes in the traversal order can be incorporated into the exhibitions by editing the focus map file. Recompiling the touchstone is not necessary. Since the focus map file is platform independent, there merely needs to be one focus map Me through platform.