harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Geir Magnusson Jr." <g...@pobox.com>
Subject Re: [general] JUnit consistency, practices
Date Wed, 06 Dec 2006 13:31:57 GMT
+1

Konovalova, Svetlana wrote:
> IMHO we can start collecting info about running the classlib tests. 
> I suggest to create a new wiki page, let's say, "Running The Classlib
> Tests" and add a link to it front the wiki front page > components >
> class library.
> ASA we get the clear picture of this issue and verify all the ways to
> run tests, we'll be able to post this info on the web-site.
> Want do you think?
> I'd be glad to help.  
> 
> Best regards,
> Sveta
> 
> -----Original Message-----
> From: Alexei Zakharov [mailto:alexei.zakharov@gmail.com] 
> Sent: Tuesday, December 05, 2006 5:25 PM
> To: dev@harmony.apache.org
> Subject: Re: [general] JUnit consistency, practices
> 
> +1 for having the doc. But personally I don't know the way how to run
> tests from the particular test class (not to speak of individual test
> methods) using the current build system. So I don't really know what
> exactly should be copy/pasted.  Or you was talking about by-hand test
> invocation cmd mentioned above?
> 
> Thanks,
> 
> 2006/12/5, Tony Wu <wuyuehao@gmail.com>:
>> Agree! When I tried harmony on some applications, I found it is very
>> hard to run the test of application mainly because there is no
>> instruction for that at all. So I think it is also not very easy for
>> user who have interest to run tests of harmony. And a good instruction
>> may be good for having many user's help to run harmony tests on
>> various platforms.
>>
>> On 12/4/06, Oliver Deakin <oliver.deakin@googlemail.com> wrote:
>>> Perhaps some kind of "How To Run The Classlib Tests" section on
>>> the website would be useful? (Assuming there isn't already one with
>>> this information in) Then none of us would have to remember - it
>>> would be right there to copy/paste :)
>>>
>>> Regards,
>>> Oliver
>>>
>>>
>>> Ivan Popov wrote:
>>>> Alexei,
>>>>
>>>> I agree that it is still possible to run JUnit tests from command
> line
>>>> even without having main() in the code. But I think it is easier
> to
>>>> run test by convenient way
>>>>
>>>>  $ java -cp junit.jar TestClass
>>>>
>>>> rather than in a more complex manner
>>>>
>>>>  $ java -cp junit.jar junit.textui.TestRunner TestClass
>>>>
>>>> Actually, I constantly forget the right spelling of the full class
>>>> name for TestRunner class and have to look into JUnit doc to
> specify
>>>> proper name for such a command line. Also, it would be
> inconvenient if
>>>> someone runs test from an IDE that does not support JUnit
> environment,
>>>> but launches test as a usual Java application.
>>>>
>>>> I don't insist on adding main() to each JUnit testcase, but I see
> no
>>>> reason for removing this functionality from those test where it
>>>> already exists.
>>>>
>>>> Thanks.
>>>> Ivan
>>>>
>>>> On 11/29/06, Alexei Fedotov <alexei.fedotov@gmail.com> wrote:
>>>>> Ivan, Stepan,
>>>>>
>>>>> I personally set +1 for removing main() method. Any script or
> command
>>>>> line can be trivially modified to launch JUnit tests without
> main()
>>>>> method: one should just add junit.textui.TestRunner class before
> a
>>>>> test class name.
>>>>>
>>>>> $ java -cp junit.jar junit.textui.TestRunner TestClass
>>>>>
>>>>> I'm writing this trivial thing here because during our work on
> class
>>>>> library test enabling it was FAQ N1 for all C/C++ developers.
>>>>>
>>>>> Note, any JUnit test won't work without junit.jar anyway. If you
> have
>>>>> junit.jar, you have a standard test runner, which is also quite
>>>>> lightweight.
>>>>>
>>>>> --
>>>>> Thank you,
>>>>> Alexei
>>>>>
>>>>> On 11/29/06, Ivan Popov <ivan.g.popov@gmail.com> wrote:
>>>>>> -1 for removing main().
>>>>>>
>>>>>> I often run individual tests from command line or using scripts
> and
>>>>>> it's easier to launch them as a usual Java application. Also,
> this
>>>>>> facilitates creating separate bundle with test to attach to a
> bug
>>>>>> report or send to other people, who can just run it from
> command line
>>>>>> or use script with the all required options already specified,
> instead
>>>>>> of setting IDE for this test.
>>>>>>
>>>>>> Thanks.
>>>>>> Ivan
>>>>>>
>>>>>> On 11/29/06, Nathan Beyer <nbeyer@gmail.com> wrote:
>>>>>>> There is a large amount of inconsistency across the tests and
> I'd
>>>>> like
>>>>>>> to lobby for cleaning them up as much as possible. I'm of the
>>>>> opinion
>>>>>>> that test code should be clean, simple and transparent. Here
> are
>>>>> some
>>>>>>> of the more noticeable items that I'd like to cleanup.
>>>>>>>
>>>>>>> * Empty setUp/teardown methods - There are a number of tests
> that
>>>>>>> override setUp and/or teardown methods, but are either empty
> or just
>>>>>>> call the super implementation.
>>>>>>>
>>>>>>> * Singleton suite methods - There are some tests that contain
> a
>>>>> static
>>>>>>> "suite" method that creates a TestSuite and adds one test
> (the test
>>>>>>> class it's declared in). Are there any practical uses for
> these
>>>>>>> methods? TestSuites are for grouping together tests to treat
> them as
>>>>>>> one unit. Since these suites are just one test, it doesn't
> seem to
>>>>>>> provide much value.
>>>>>>>
>>>>>>> * main method launching text runner - There are some tests
> that
>>>>>>> contain "main" methods which run the enclosing test via a
> JUnit text
>>>>>>> runner. Most IDEs have built-in support for JUnit and can
> launch any
>>>>>>> test arbitrarily and Ant can do the same thing. Does anyone
> launch
>>>>>>> tests via these methods?
>>>>>>>
>>>>>>> My proposal would be to clean up these inconsistencies by
>>>>> eliminating
>>>>>>> them, but what does everyone else think?
>>>>>>>
>>>>>>> -Nathan
>>>>>>>
>>> --
>>> Oliver Deakin
>>> IBM United Kingdom Limited
>>>
>>>
>>
>> --
>> Tony Wu
>> China Software Development Lab, IBM
>>
> 
> 


Mime
View raw message