Hello,

JUnit is a defacto standard in Java circle, which means it has a rich

ecosystem. I would like to "hijack" this ecosystem for other

languages/environments than Java. I actually did something for running

Javascript tests with htmlunit and would now like to to do the same

thing for C/C++ tests: use JUnit to drive and report test executions

that encapsulates native tests.



I am no sure this is a good idea, but it kind of worked for my JS

tests. For the C tests, I am confronted with an issue probably due to

my misunderstanding of JUnit architecture: It seems that Runner

instances need Class objects to work on, being passed them by the

framework. But of course, in my particular case I do not have specific

Class objects to work with as tests are written in native C/C++ code.



Could somebody point me to the right direction, or give me some good

reasons to forget about this ?



Thanks



Arnaud

Why not use an existing C/C++ framework? Gtest is my

current favorite.



Charlie



> -----Original Message-----

> From: junit@yahoogroups.com [mailto:junit@yahoogroups.com] On

> Behalf Of Arnaud Bailly

> Sent: Friday, July 03, 2009 8:16 AM

> To: junit

> Subject: [junit] Hijacking JUnit infrastructure

>

> Hello,

> JUnit is a defacto standard in Java circle, which means it

> has a rich ecosystem. I would like to "hijack" this ecosystem

> for other languages/environments than Java. I actually did

> something for running Javascript tests with htmlunit and

> would now like to to do the same thing for C/C++ tests: use

> JUnit to drive and report test executions that encapsulates

> native tests.

>

> I am no sure this is a good idea, but it kind of worked for

> my JS tests. For the C tests, I am confronted with an issue

> probably due to my misunderstanding of JUnit architecture: It

> seems that Runner instances need Class objects to work on,

> being passed them by the framework. But of course, in my

> particular case I do not have specific Class objects to work

> with as tests are written in native C/C++ code.

>

> Could somebody point me to the right direction, or give me

> some good reasons to forget about this ?

>

> Thanks

>

> Arnaud

>

>

> ------------------------------------

>

> Yahoo! Groups Links

>

>

>

>

Thanks for your answer,

We use a framework for writing and executing tests (eg. at present

CUnit and CPPUnit). But we are in a mixed language environment and

would like to be able to integrate smoothly my native tests w/ java

build system (eg. maven, CI).



Regards,

Arnaud

2009/7/6 Arnaud Bailly <arnaud.oqube@...>:

> Thanks for your answer,

> We use a framework for writing and executing tests (eg. at present

> CUnit and CPPUnit). But we are in a mixed language environment and

> would like to be able to integrate smoothly my native tests w/ java

> build system (eg. maven, CI).



You could look at the Test Anything Protocol, which is used to gather

results from tests written in different languages.  There is a Java

implementation.



http://testanything.org



--Nat



--

http://www.natpryce.com

On Fri, Jul 3, 2009 at 11:16 AM, Arnaud Bailly<arnaud.oqube@...> wrote:

> Hello,

> JUnit is a defacto standard in Java circle, which means it has a rich

> ecosystem. I would like to "hijack" this ecosystem for other

> languages/environments than Java. I actually did something for running

> Javascript tests with htmlunit and would now like to to do the same

> thing for C/C++ tests: use JUnit to drive and report test executions

> that encapsulates native tests.

>

> I am no sure this is a good idea, but it kind of worked for my JS

> tests. For the C tests, I am confronted with an issue probably due to

> my misunderstanding of JUnit architecture: It seems that Runner

> instances need Class objects to work on, being passed them by the

> framework. But of course, in my particular case I do not have specific

> Class objects to work with as tests are written in native C/C++ code.

>

> Could somebody point me to the right direction, or give me some good

> reasons to forget about this ?



Arnaud,



There are definitely ways to do what you want.  Runner doesn't require

a Java class, it just happens that almost all of the provided

subclasses do have a Java class.



How are you planning to run the overall test suite?  Depending on that

answer, it may work best to have a "stand-in" Java class that serves

as the root of your C++ suite.



    David Saff



>

> Thanks

>

> Arnaud

>

>

> ------------------------------------

>

> Yahoo! Groups Links

>

>

>

>

hello david,

thanks for your encouraging mail. Right now, I just used the Result

and RunListener classes and interfaces to manage feedback from my

tests, but the basic idea is that I would run JUnit with some custom

runner that would discover "test classes" that would report results.



In a previous JS framework I wrote, the wrapper classes were defined

statically (ie. each known html test was associated to a defined

JUnit4 test class) and run with a custom Runner deriving from

JUnit4BlockRunner (and making use of parameterized btw).



I could have a java class "proxy" that would report results on a given

runner. Is that what you are thinking of ?



Arnaud

I have something similar for antunit (unit test written as ant script).  I

wanted to integrate antunit test with the IDE (mostly to run the ant tasks

in a java debugger).



I have written a Runner (AntUnitSuiteRunner) that you can use to run a

'proxy' that can be detected by your IDE.

Then, there is a simple protocal between my runner and the 'proxy' class.



It looks like this :



@RunWith(AntUnitSuiteRunner.class)

public class EatYourOwnDogFoodTest {



     public static TestSuite suite() {

         File script = new File("src/etc/testcases/antunit/java-io.xml");

         return new AntUnitSuite(script, EatYourOwnDogFoodTest.class);

     }



}





Gilles Scokart





2009/7/6 Arnaud Bailly <arnaud.oqube@...>



>

>

> hello david,

> thanks for your encouraging mail. Right now, I just used the Result

> and RunListener classes and interfaces to manage feedback from my

> tests, but the basic idea is that I would run JUnit with some custom

> runner that would discover "test classes" that would report results.

>

> In a previous JS framework I wrote, the wrapper classes were defined

> statically (ie. each known html test was associated to a defined

> JUnit4 test class) and run with a custom Runner deriving from

> JUnit4BlockRunner (and making use of parameterized btw).

>

> I could have a java class "proxy" that would report results on a given

> runner. Is that what you are thinking of ?

>

> Arnaud

>

>





[Non-text portions of this message have been removed]