cppunit-devel

[Cppunit-devel] fixture between different test methods?

[CppUnit d. m. l. <cpp...@li...>]

Hi cppunit developers and users!

I have a question about test execution. I have a test suit as shown
below with two test methods. The code is then compiled into a dll
and I run it with this command:

DllPlugInTesterd_dll.exe -c plugin_d.dll

I now have the following problem. During startup cppunit creates two
instances of the ExampleTest class. Then it calls setup(), sampleTest()
and finally teardown(). Then it does the same on the second instance 
with the method sampleTest2().

This looks quite strange. I thought I could initialize some class in the
setup method which is then shared between all the different test method
calls?

The reason why I would like to share an instance between them is that I
have to start an external program during test execution (which takes
very long and therefore should not be started and killed for each test
method invocation). I therefore used a shared pointer which I thought
would be accessible for all test invocations. But since each invocation
uses a different instance this does not work.

Any suggestions how this should be done would be really great!

Please also tell me if what I wrote / want is unclear

Thanks a lot, Fabian



class ExampleTest : public CPPUNIT_NS::TestFixture
{
  CPPUNIT_TEST_SUITE( ExampleTest );
  CPPUNIT_TEST( sampleTest );
  CPPUNIT_TEST( sampleTest2 );
  CPPUNIT_TEST_SUITE_END();

public:
  void setUp();
  void tearDown();

  // The different test methods
  void sampleTest();  
  void sampleTest2();  
};

Re: [Cppunit-devel] fixture between different test methods?

[CppUnit d. m. l. <cpp...@li...>]

CppUnit development mailing list wrote:
> Hi cppunit developers and users!
>
> I have a question about test execution. I have a test suit as shown
> below with two test methods. The code is then compiled into a dll
> and I run it with this command:
>
> DllPlugInTesterd_dll.exe -c plugin_d.dll
>
> I now have the following problem. During startup cppunit creates two
> instances of the ExampleTest class. Then it calls setup(),
> sampleTest()
> and finally teardown(). Then it does the same on the second instance
> with the method sampleTest2().

Each test case has its own instance to ensure each test case execution is 
independent.

> This looks quite strange. I thought I could initialize some class in
> the setup method which is then shared between all the different test
> method calls?
>
> The reason why I would like to share an instance between them is that
> I
> have to start an external program during test execution (which takes
> very long and therefore should not be started and killed for each test
> method invocation). I therefore used a shared pointer which I thought
> would be accessible for all test invocations. But since each
> invocation
> uses a different instance this does not work.

Possible solutions I see are:
- make a single test case and add log to know which "internal" test case 
failed (see ClockerPlugIn as an example to integrate a listener listener 
hook with test result). Downside is you won't get precise reporting of what 
failed in the results without reading the log.
- make your "shared" instance static. Though, you would still need to figure 
out when to clean it up. Thinking about it, may be you could use a rough 
algorithm: if last executed test did not ask for it, then destroy the shared 
instance. Makes the component that managed the "shared" instance a test 
listener.

Baptiste.

> Any suggestions how this should be done would be really great!
>
> Please also tell me if what I wrote / want is unclear
>
> Thanks a lot, Fabian
---
Baptiste Lepilleur <bl...@us...>
CppUnit maintainer
OpenTest and CppUnit 2 developer.