LIME Interface Test Bench User Guide

Xi Chen, Janne Kauttio, Olli Saarikivi and Kari Khknen a o October 6, 2009

Contents
1 Introduction 2 The example program 3 Installation and starting up 4 Monitoring tools 4.1 Opening and compiling the example program 4.2 Running the example program . . . . . . . . 4.3 Other monitoring tools . . . . . . . . . . . . . 4.4 Command-line usage . . . . . . . . . . . . . . 5 Testing tools 5.1 Using Lime Concolic Tester . 5.2 Limitations of LIME Concolic 5.3 Using JUnit Tools . . . . . . 5.4 Command-line usage . . . . . 6 Other utilities 2 2 6 8 8 10 12 14 15 15 20 20 23 23

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . . Tester . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

Introduction

LIME (LightweIght formal Methods for distributed component-based Embedded systems) Interface Test Bench (LimeTB from now on) is a toolkit for testing software components written in Java programming language (limited support for C programming language is also available). With LimeTB, the user can specify behavioral aspects of the software component interfaces as annotations in the source code by using LIME Interface Specication Language (ISL). These specications can then be automatically monitored for violations during the execution of the program. LimeTB also has tools for automating the testing of a program, and for generating JUnit unit tests. This document introduces the tools provided by LimeTB and how they can be used with the provided graphical user interface. An example program is also introduced that will be used as a running example throughout the guide. The example program is a simple Java program with interface annotations, and will hopefully motivate the reader on how the tools can be useful also in real-life applications. All the LimeTB tools can be used from the user interface and also have a command-line variants that can be used independently. Conventions used in this guide are as follows: Monospace font will be used for both commands in the graphical user interface as well as command-line commands. Bold font will be used for dierent le names throughout the guide. Italic Italic font will be used to denote names of directories. The guide is structured as follows: In chapter 2 the example program will be presented on a source code level, along with explanations what the dierent parts of the program do. Chapter 3 will go through the initial steps of using the tool, mainly how its installed and how the graphical user interface can be started. Chapters 4 and 5 will go through the main features of LimeTB by presenting the tools and how they work with the example program. Finally, Chapter 6 will briey consider the utilities specic to the user interface that are designed to make it more usable. In every chapter where a feature of LimeTB is presented, also the equivalent command-line tools will be considered briey, mainly to let the user know what command does what.

The example program

In this section the example used during the rest of the guide will be demonstrated in detail with complete source code of all the les used. The les will be presented one at a time, each of them followed by notes about what it actually does and what we are trying to achieve. The example program is a simple lock implementation that can be locked and unlocked, and an associated main-method that generates one such lock instance and does some operations on it. It should also be noted that this example is also available in the LimeTB/examples/limtlct/example lock directory in the tool package.

package example_lock; import fi.hut.ics.lime.aspectmonitor.annotation.CallSpecifications; @CallSpecifications( regexp = { "StrictAlteration ::= (lock() ; unlock())*" } ) public interface Lock { public void lock(); public void unlock(); }

Figure 1: Lock.java In Figure 1 is the interface denition of a Lock interface. Apart from the extremely simple interface, the only interesting part in this le is the CallSpecications annotation at the top. This LIME Interface Specication Language annotation species a spec called StrictAlteration, which basically requires that the calls to methods lock() and unlock() must be done alternately, so a lock that is locked shouldnt be locked again before it is opened rst. Remember that this is just one example of what a user can do with Interface specications. For more information please check the other examples provided with LimeTB or the specication language documentation 1
1 Available

at http://www.tcs.hut.fi/Software/lime/

package example_lock; public class LockImpl implements Lock { private boolean locked; public LockImpl() { locked = false; } public void lock() { locked = true; } public void unlock() { locked = false; } public boolean getLocked() { return locked; } }

Figure 2: LockImpl.java In Figure 2 is shown a class le that implements the interface in Figure 1. The main thing to note here is that there is no need to write anything special in the classes implementing the interfaces with annotations to make them work with LimeTB tools.

package example_lock; public class Main { public void testme (int x, int y) { Lock lock = new LockImpl(); lock.lock(); System.out.println("x = " + x + ", y = " + y); if (x > y) if (x * y == 1452) if (x >> 1 == 5) lock.lock(); lock.unlock(); } public static void main(String[] args) { Main m = new Main(); if (args.length < 2) { System.out.println("please give two integers as arguments"); System.exit(1); } int x = Integer.parseInt(args[0]); int y = Integer.parseInt(args[1]); m.testme(x, y); System.exit(0); } }

Figure 3: Main.java Finally in Figure 3 is the main class of the example program. The main method here takes two arguments, and then calls a testme(int, int) method. What this method does is that it rst creates a LockImpl object (that implements Lock interface) and then locks it. Then the method checks if the arguments fulll a specic criteria (these tests are carefully crafted to only accept certain specic values for the sake of demonstrating the capabilities of the testing tools) and if they do, the lock is locked a second time, leading to a violation of the specication. At the end the method unlocks the lock and exits. As the reader has propably noticed by now, the example doesnt really do anything useful. It should be noted however, that it demonstrates quite well

a case where we have a program with an erroneous execution only with very certain input values. Now imagine a case where such a component is part of some bigger software, and the error could be considerably dicult to locate. Next well go into the installation of LimeTB, and how the user interface can be launched. After that the tools themselves will be presented one at a time.

Installation and starting up

Installation of LimeTB should be relatively straightforward depending a bit on the system where its going to be installed. The tarball LimeTB.tar.gz holds basically everything a user should need to start using the tools, apart from few dependencies. Most of the dependencies are provided in the LimeTB package both as precompiled binaries (for x86 linux) and as source code, but some still need to be installed separately. Dependencies needed by and not provided by LimeTB: Java 1.6: Can be obtained from Sun website 2 . ACC: (AspeCt oriented C) only required for using LIMT with C programming language, can be obtained from ACC website 3 (read the installation instructions carefully). Boolector: Is a SMT-solver used by the test generation tool. It can be downloaded at Boolector website 4 . After downloading Boolector, a shared library used by LimeTB must be compiled. A makele for this task is provided in the LimeTB/dependencies/solvers/ folder. See the LCT README for more detailed instructions to compiling the library. Yices: Yices is alternative to Boolector and it is available from SRI website 5 . The instructions for compiling a shared library for Yices are also given in the LCT README. Dependencies provided by LimeTB: Scheck: In case scheck doesnt work, necessary sources and patches for recompilation are provided in the LimeTB/dependencies directory. Doxygen: Only required for using LIMT with C programming language, sources available under LimeTB/dependencies and precomipiled binaries under LimeTB/bin directory. The installation should be as simple as extracting the LimeTB package to a directory of the users choosing and compiling either Boolector or Yices libraries is the test generation functionality is needed. Once this has been done, the tool should be basically ready to use. The contents of LimeTB package are as follows:
2 http://java.sun.com/ 3 http://research.msrg.utoronto.ca/ACC 4 http://fmv.jku.at/boolector/ 5 http://yices.csl.sri.com/

bin/ ChangeLog dependencies/

doc/ examples/ lib/ README LICENSE source/

This directory holds all the runnable scripts and binaries LIME-tools are used with. User can see how the tools have been updated over time from this le. From this directory the user can nd the necessary dependencies to run LIME-tools that are shipped with LimeTB. Documentation (including the READMEs for individual tools) is contained in this directory. All the example programs are under this directory, sorted by the tool they are designed for. This directory holds all the .jar-les and other library items necessary for LIME-tools. The general README le for the package, read this rst. The license LimeTB uses is MIT, and the license text can be found in this le. From this directory the user can nd all the source les for LIME-tools that are included in LimeTB.

The easiest way to use LimeTB is to put the bin/ directory in the search path for commands in the current shell (variable $PATH with bash for instance). After this has been done, the tools can basically be run from any directory in the system. In the following chapters well present how LimeTB utilities can be used in the provided graphical user interface. To start up this interface, run the script named limegui in the bin directory. If the directory was added to path as advised above, then just running the command limegui in any directory should also work.

Figure 4: Freshly started user interface Once the user interface (Figure 4) pops up, the rst thing we need to do is to congure it. In order to do this, select Configuration from the menu on the top of the window, select Set toolkit path... and point the opening le dialog to the extracted LimeTB directory. Once this is done, select Save config from the Configuration menu and the path will be saved into a conguration le, and need not be set again after closing and restarting the user interface. The Save config command saves all the settings that have been set in the user interface thus far (including source directories that are open among other things), so whenever changes have been made, using this command quarantees they will be remembered.

Monitoring tools

In this chapter the compilation and running of an annotated program will be demonstrated. Well also show how you can give arguments to the program and how the specication violations are reported.

4.1

Opening and compiling the example program

Once the graphical user interface has been started, the rst thing that must be done is to show it where the source les are located. This can be done by selecting Monitor, and then choosing Open source directory... from the menu. Doing this will pop up a le chooser, that can be navigated to the source les. The thing to note here is that the tool needs the directory that holds the les as an argument, not the les themselves. In the case of our example, we

shall point it to the LimeTB/examples/limt-lct/example lock directory (as can be seen in Figure 5).

Figure 5: Select a source directory to be opened. Once the directory has been set, the program can be compiled. To do this, the Compile command in the Monitor menu can be used, or the Compile button in the main interface that does the same thing. Compiling the example program (providing all the dependencies are working correctly), should produce output similar to Figure 6.

Figure 6: Output of Compile.

4.2

Running the example program

Now that the example has been compiled (with the monitors created automatically during the process), it can be executed. To do this, either the command Run under Monitor menu or the similarly labeled button on the main interface can be used. Notice now that the tool automatically assumes the name of the main class (class containing the runnable main-method) to be package.name.Main, which it is for this example. If the running is successful, the mainwindow should look like Figure 7.

Figure 7: Output of a failed Run of the program. The program didnt run properly, since as seen in Chapter 2, the program needs two parameters. To set these parameters, open the Monitor menu again, and choose the command labeled Set main class.... As can be guessed from the name of this command, this is mainly used for setting the name of the runnable class (if it isnt Main), but it can also be used for giving arguments for the program. Now insert any two integer values after the class name (13 and 42 are used in Figure 8) and try running the program again.

Figure 8: Setting some arguments for the example program.

10

Figure 9: Output of a successful Run of the program. Now the program should actually nish, and the output it produces can be seen on the interface window (Figure 9). Now any specication violations didnt happen (since 13 and 42 dont fulll the criteria needed for the second call to lock() in the code), but we can try that again. Through some math or by previous knowledge, it can be deduced that the only values for arguments that will lead to a specication violation are 10 and -858993314 (in the testing section of the guide it is shown how these values can be obtained using the LimeTB tools). Now lets try setting the arguments to these values (Figure 10) and see what happens.

Figure 10: Setting dierent arguments for the example program.

11

Figure 11: Output of Run with an exception trace. As can be seen from Figure 11, a specication violation happened. Specication violations cause a special exception to be thrown in the code, and our tools clean up the normal java exception output to a more readable form automatically before showing it. From the output we can now see that the class LockImpl.java caused the exception (since the lock() method is implemented there), and it was called from Main.java on line 15 (which is where the second call to lock() is). From the output we can also see the type of specication that was violated was a Call specication, and that the name of the spec was StrictAlteration. By looking the source code of the example in Chapter 2 we can see that this indeed is the case.

4.3

Other monitoring tools

Now for the other previously unmentioned features of the monitoring toolkit. The user interface has buttons for Emacs and Tracer. The Emacs command opens the text editor emacs on the current output on the screen, which is helpful for these exception violations since the user can navigate straight to the violating source les by just clicking the name of the le with the line number in the editor. This kind of integration isnt provided for any other text editor at the moment, but the user can of course edit the source les in any way he likes. The Tracer feature is a simple text-based tracing utility, that can be used to view the run of the program from the eyes of the specication monitors. The tracing utility can be launched by clicking the button, and it will automatically open the log le for the most recent run of the program as can be seen in Figure 12.

12

Figure 12: Tracer with simple verbosity. The tracer has several ltering options, including only showing the output of monitors for certain specications (the Specications: eld), or for specic observer instances (the Observers: eld). To nd out all the names of the specications monitored and the observers created during the run, the user can check the Summary option and a list of them will be printed, along with information about which observer noticed the specication violation. The ltering elds should be lled with the names of the specications (or observers) to be followed separated by commas (,). The Verbosity setting has three dierent values, simple, propositions and stacktraces, which respectively print the basic output, information how specic propositions were evaluated each time an observers state was updated, and the stacktraces for each function call an observer noticed. Dierent observers have also dierent layers of indentation in the output to make it more readable, which can be disabled with the No indent option. In Figure 13 is the output for the previous run with the verbosity set to maximum level to show the eect.

13

Figure 13: Tracer with stacktraces verbosity.

4.4

Command-line usage

All the tools mentioned above can be also used from the command line. limejc is a script that takes care of instrumentation and compilation of the target program, (same as Compile command. The usage of this script is pretty straightforward; just run limejc path/to/source/dir where path/to/source/dir points to the directory with the source les to be compiled. If the user wants to compile the les in the current directory, limejc . can be used. The script creates a directory called spoon, which will hold all the produced class les. This directory will be removed automatically by the script if it exists already, so some care should be taken when the script is used. In order to run the compiled program, a script called limejava is provided. This script should be run from the spoon directory created in the previous step, and must be run with the complete package and class name of the le (e.g. for the example_lock used in this guide, rst go to the directory with the source les and run limejc ., then change to the created spoon directory and nally run limejava example_lock.Main 13 42). In order to achieve the parsed output seen in the previous Section, a script called beautify.py can be utilized. This script takes the output of a program as input, and if it notices a LIME exception, it beauties that part. The usage varies with each shell a bit (since the exceptions are printed in the error stream and that must be directed to the beautier), but for bash a command like limejava example_lock.Main 10 -858993314 2>\&1 | beautify.py will do the trick. Finally, the tracer can be used with command tracer (use tracer -h for information about additional options). The log le of the last run the tracer needs will be generated under the spoon directory by the name runlog, notice that this le will be replaced with each additional run so if the user wants to save a log for a certain run, the le should be copied somewhere safe before 14

running the program again. In the next chapter we shall move on to the testing tools, and demonstrate how the program can be tested automatically (and how the input values that caused the exception in the example can be found).

Testing tools

This Chapter focuses on the usage of the testing utilities provided by LimeTB.

5.1

Using Lime Concolic Tester

In order to use the Lime Concolic Tester (LCT) tool, the source le for the Main class needs to go through some small changes. Basically, the tester works by generating input values for methods and trying to lead the execution to an erroneous state by altering those values. The LCT doesnt actually know what values it should generate unless we specically request them, so the class le in Figure 14 shall serve as the Main class for test generation purposes (the le is available in the example directory under the name MainLCT.java).

15

package example_lock; import fi.hut.ics.lime.tester.LCT; public class MainLCT { public void testme (int x, int y) { Lock lock = new LockImpl(); lock.lock(); System.out.println("x = " + x + ", y = " + y); if (x > y) if (x * y == 1452) if (x >> 1 == 5) lock.lock(); lock.unlock(); } public static void main(String[] args) { Main m = new Main(); int x = LCT.getInteger(); int y = LCT.getInteger(); m.testme(x, y); System.exit(0); } }

Figure 14: MainLCT.java The input values for testme() are no longer read from arguments, but instead they are set to be generated by LCT. LCT supports generating primitive type input values with the exception of oating point numbers and also some support for generating input objects is provided. To read an input value from LCT, method calls LCT.getInteger(), LCT.getShort(), LCT.getLong(), LCT.getByte(), LCT.getBoolean(), LCT.getChar() and LCT.getObject() can be used. For more information about these methods, see the LCT README le in the documents folder of LimeTB. It should also be noted that this le is also completely runnable. If ran by itself the values LCT generates will just be random, which is basically a single test run for the program. The usage of testing tools in the user interface follows the same track as the monitoring tools. First, we need to select the directory where the tests should be executed (this step can be omitted if a source directory has already been selected

16

for the testing tool, the path will be lled out automatically). This can be done by opening the Tester menu, and choosing the command Set test path... which will open a new le chooser. When selecting this directory manually, the user should remember to set it to the parent directory of the source les if they belong into a package (the directory that contains the package directory). Once the path has been set, the program can be instrumented and compiled by using the Compile and/or instrument... command in the Tester menu. Now unlike the monitoring tools, this compilation process takes a single class or java le as an argument. So in the case of this example, point the le chooser rst to example lock directory (under the spoon directory where the le chooser opens), and then to either MainLCT.java or MainLCT.class (Figure 15). If the compilation process in the monitoring tools hasnt been run before and these les dont exist, for the sake of this example the user should go back to Chapter 4 and do the necessary steps mentioned there. It should also be mentioned that even though we use LCT here with the monitoring tools, also programs that dont have specication language annotations in them and havent been compiled by the monitoring tools can be tested with LCT. However, as the error in our example depends on the interface specications, we must use the monitoring tools in this case.

Figure 15: Choose a le to be compiled and/or instrumented. A new window will be opened where the output of the compilation and instrumentation process (providing it was successful) should look like in Figure 16.

17

Figure 16: Output of Compile and/or Instrument... The next step now is to run the tester for the instrumented program. In order to do this, choose Run... command from the Tester menu. The correct class le should automatically be selected in the opening le chooser (Figure 17), so accepting that will usually be enough to run the tester. If something goes wrong however and the le isnt selected automatically, the user must navigate to it manually. LCT compilation and instrumentation process creates a directory called output under the directory which has been set as the test path with Set test path... command. Under this directory the class les generated by the process should be found (in their respective package directories), and can then be selected for the tester to run tests on.

Figure 17: Choose an instrumented class le to run. As the testing commences, a new window is opened where all the single test case outputs will be printed as can be seen in Figure 18. In the main window a test server will be run, which will print out important information about the performed tests. Output for the example should look like Figure 19. Here can be seen (among other things) the following: One error was found using input values 10 and -858993314 (which are the only ones leading to an error and LCT found them automatically), total number of execution paths explored was 4,

18

and the test coverage was 44%. The low value for test coverage is caused by the fact that we now have two dierent Main classes in the package (Main and MainLCT), and the other one wasnt explored at all during the run.

Figure 18: Output of Run command; information about individual tests.

Figure 19: Output of Run command; found errors and general summary. If something goes wrong when running the LCT tester and the server is left running, it can be killed by closing the output window so the run can be started again (with default values only one server can be running at a time on one machine). The Tester menu also has a Tester configuration submenu, where many of the settings used by the tester can be changed. For normal usage the default 19

values should be ne but should the user want to change them, more documentation on what they do can be found in the README le of LCT.

5.2

Limitations of LIME Concolic Tester

LCT has some limitations that aect what kind of programs it can test eciently. One of the current major limitations with LCT is that it does not instrument any of the standard Java libraries by default. This means that it is not possible to use, for example, String objects as inputs created by LCT. There is, however, support for Integer, Boolean, Short and Byte classes provided. The support for these classes is implemented by providing a custom implementation of the classes and replacing the original Java standard classes in the program under test with these new classes. This has the downsize, that if the program under test uses a piece of code that has not been instrumented by LCT and it receives one of the supported standard class objects from the instrumented code, the program may crash as the types of the original and replaced classes are dierent. Therefore it is recommended to use the above mentioned classes in the program under test only if the program can be fully instrumented. For a more detailed information about this limitation, see the research report on the test generation method available at the LIME website 6 . Another limitation of the tool regards handling arrays. Currently the tool provides approximative support for generating tests for programs that use arrays. More detailed explanation of how arrays are handled is presented in the above mentioned research report in section 3.3. As an example of this limitation, consider the following piece of code. // variable a is an array of Strings int x = LCT.getInteger(0,10); System.out.println(a[x]); The program prints out the element in array a based on the input values generated by LCT. However, for this program LCT generates only one test case where the value of x is selected randomly between 0 and 10, instead of generating test cases that print all the values in the array. This is because LCT tries to generate possible execution paths to explore based on the branching statements (e.g., if-statements) written in the program under test. In this particular example there are no such branching statements and therefore LCT thinks that there is only one possible execution path in the program.

5.3

Using JUnit Tools

The LIME JUnit Tools (LJUT) provides tools for generating test drivers for unit testing and interface testing. The generated test drivers retrieve input from the LCT to drive testing. LJUT will also save the tests that were run with the LCT as a static JUnit test case to be included in a unit test suite. The LJUT tools can be accessed from Test drivers and JUnit tests... menu item in the Tester menu. A new window for accessing the LJUT tools opens (Figure 20) . This window contains a row of buttons for accessing the
6 http://www.tcs.hut.fi/Software/lime/

20

tools and two le path elds for providing the JUnit test generator tool with the input les it needs.

Figure 20: LJUT window, some of the elds empty. To be able to generate test drivers the toolset must be pointed to the correct project folder. LJUT uses the same folder as the LCT, which is set from the Set test path... menu item in the Tester menu. Notice that if you have been following this guide from the beginning, the path doesnt need to be set again at this point. The Generate method unit test driver... button in the Test drivers and JUnit tests window provides access to generation of straightforward unit test drivers for single methods. Pressing the button will open a dialog box asking for a constructor and method signature. These are written in a Java like syntax. Examples of valid constructor signatures are java.lang.String() and my.package.hierarchy.MyClass(int,long). Valid method signatures include foo(boolean,byte) and bar(java.lang.Object). For our example, ll in the values example lock.Main() as a constructor signature, and testme(int,int) as the method signature in order to generate testcases for the testme method. After pressing the OK button in the dialog, an output window is opened (see Figure 21) and the test driver is generated in the project folder.

Figure 21: LJUT simple test driver generation output. The Generate interface test driver... tool is similar to the unit test driver generator. It also takes a constructor signature, but instead of a method signature it needs a qualied name of an interface. The class the constructor signature refers to must implement that interface. The test driver generated 21

will perform testing on sequences of calls to the provided interfaces functions. The dialog also asks for the number of iterations in the test driver that will determine the length of the call sequences to be tested. After a test driver is generated it can be used with the LCT. The test drivers are generated as les named ClassNameLCTDriver.java in the project folder (notice that for our example this isnt the folder where the other source les are since they belong to the package example lock, but one level higher in the le hierarchy), where ClassName is the name of the class under test. A second le ClassNameLCTDriver.properties is generated and is needed for JUnit test case generation (if this functionality is not needed the le can be discarded). As described in the section on using the LCT the test driver must be compiled and instrumented before use. When the tests are run with the LCT the server writes a testcases le in the bin folder in the LIME distribution folder. This le is used in the generation of JUnit tests. To generate JUnit tests the properties le and testcases le elds in the Test drivers and JUnit tests window must point to valid les (Figure 22). These are normally prelled during the generation of test drivers, but may not be correct in all situations. To generate the JUnit tests press the Save as JUnit tests... button. An output window is opened again (the output should look like Figure 23) and JUnit tests are written in to a le named ClassNameLCTDriverTestCase.java in the project folder.

Figure 22: LJUT window, all elds lled.

Figure 23: LJUT JUnit test generation output.

22

5.4

Command-line usage

Before running the tools from the command line, make sure bin folder is in $PATH as described in Chapter 3. LCT can be used from the command line with commands LCTcompile, LCTinstrument, LCTserver and LCTrun. Basically the testing process works like in the user interface: Change to the proper directory, compile the les if needed (if working with .java les youll need to do this, recompilation is not needed if LCT is used together with the monitoring tool), instrument the les, start the server (preferably in a dierent terminal) and run the tester. LCTcompile and LCTinstrument commands take .java and .class les as arguments respectively, and LCTinstrument saves the output les it produces in and output directory which will be generated automatically. In order to run LCTrun, the test server must be running. The server can be started with LCTserver command, and should be started in a new terminal window as mentioned above. LCTrun takes .class les as arguments, and these les should be instrumented so please use the les from the output directory. Using LJUT tools from the command line is essentially identical to using them from the GUI. All of the command line tools provide help on the possible options when given a --help option. The unit test drivers for single methods are created with ljut-simple. This tool needs a constructor and method signature. For example it might be called like this: ljut-simple "example_lock.Main()" "testme(int,int)" The interface test drivers are created with ljut-interface, which needs an interface name, a constructor signature and the number of iterations to do. For example: ljut-interface "example.File" "example.FileImpl" 2 JUnit test generation is done with the ljut-junit tool that expects a properties and a parameters le (the testcases le). It could be called for example like this: ljut-junit --parameters-file /home/user/LimeTB/bin/testcases TestClassLCTDriver.properties

Other utilities

The last set of utilities a use can nd in the interface is located under the Commands menu. Usually this menu shouldnt need to be used at all, but for instance in the rare cases where the build process for the program under test cant be handled by the Lime tools, the commands under this menu might prove useful. First there are the Add custom run command... and Remove custom run command... items. With these features the user can add any command thats

23

runnable from command line to a list of runnable commands, and run it from the user interface. In order to add a command, three things need to be specied: A name for the command which is used to just tell te commands apart from each other, the directory where the command should be run, and last the command itself. Some care should be taken when making and running custom commands, since a badly placed rm -fr for instance can do a serious amount of damage. Change PATH variable... and Change CLASSPATH variable... commands can be used to add something extra to the respective environmental variables if thats needed. In most cases the Lime scripts take care of having all the necessary variables set correctly, but if something ends up missing it can be added here. Another thing to note is that these extra variables take eect before the actual variables set by the scripts. So if the system has one version of Java installed and the user wants to use a dierent one thats installed somewhere else, the path to the preferred Java executable directory can be set with Change PATH variable... and it will be used instead of the default one.

24