java unit vs. system tests

[Steve Vinoski <vi...@iona.com>]

Perhaps I'm missing something, but are the Qpid Java unit tests  
really unit tests? To me they seem more like system tests. One  
critical aspect of a unit test is isolation, typically achieved via  
mock objects, but when I look through the code I don't really see the  
isolation I expect to see.

Does anybody have any coverage figures for these tests? Perhaps  
someone familiar with the tests could file JIRAs for the parts known  
to lack coverage.

Again, assuming I'm not missing something, this is another reason why  
I advocate (and am still working on) moving to maven. It will force  
us to get our testing structures in order, will make it easier to add  
tests wherever required, and will also give us code coverage  
essentially for free.

--steve

Re: java unit vs. system tests

[Carl Trieloff <cc...@redhat.com>]

I would also like to see the Python systems test run across both Java 
and C++. They
do now for C++, but would like to see this to be the case with Maven.

Carl.

John O'Hara wrote:
> +1 for building the Java bits with Maven too (is this a vote?!?)
>
> We need to make sure that the build system for Java is nice for Java 
> users,
> and the packaging for C++ et al is good for them.  Ideally they would gel
> nicely.
>
> Distribution is, of course, separate from build.  So a distribution could
> make more compromises than the build system.  We need to make the code as
> accessable and easy to get going with as possible on all the platforms.
>
> John
>
> On 27/09/06, Robert Greig <ro...@gmail.com> wrote:
>>
>> On 27/09/06, Martin Ritchie <ma...@googlemail.com> wrote:
>> > Perhaps we should move the non-unit tests to a different package 
>> for now
>> so
>> > we can easily see what unit tests we have.
>>
>> That is a very good idea.
>>
>> We have some unit tests but the majority, particularly in the client
>> test directory are certainly system tests.
>>
>> RG
>>
>

Re: java unit vs. system tests

[John O'Hara <jo...@gmail.com>]

+1 for building the Java bits with Maven too (is this a vote?!?)

We need to make sure that the build system for Java is nice for Java users,
and the packaging for C++ et al is good for them.  Ideally they would gel
nicely.

Distribution is, of course, separate from build.  So a distribution could
make more compromises than the build system.  We need to make the code as
accessable and easy to get going with as possible on all the platforms.

John

On 27/09/06, Robert Greig <ro...@gmail.com> wrote:
>
> On 27/09/06, Martin Ritchie <ma...@googlemail.com> wrote:
> > Perhaps we should move the non-unit tests to a different package for now
> so
> > we can easily see what unit tests we have.
>
> That is a very good idea.
>
> We have some unit tests but the majority, particularly in the client
> test directory are certainly system tests.
>
> RG
>

Re: java unit vs. system tests

[Robert Greig <ro...@gmail.com>]

On 27/09/06, Martin Ritchie <ma...@googlemail.com> wrote:
> Perhaps we should move the non-unit tests to a different package for now so
> we can easily see what unit tests we have.

That is a very good idea.

We have some unit tests but the majority, particularly in the client
test directory are certainly system tests.

RG

Re: java unit vs. system tests

[Martin Ritchie <ma...@googlemail.com>]

I too notice this.

Perhaps we should move the non-unit tests to a different package for now so
we can easily see what unit tests we have.

-- 
Martin


On 27/09/06, Steve Vinoski <vi...@iona.com> wrote:
>
> On Sep 26, 2006, at 11:53 PM, Steve Vinoski wrote:
>
> > Perhaps I'm missing something, but are the Qpid Java unit tests
> > really unit tests? To me they seem more like system tests. One
> > critical aspect of a unit test is isolation, typically achieved via
> > mock objects, but when I look through the code I don't really see
> > the isolation I expect to see.
> >
> > Does anybody have any coverage figures for these tests? Perhaps
> > someone familiar with the tests could file JIRAs for the parts
> > known to lack coverage.
> >
> > Again, assuming I'm not missing something, this is another reason
> > why I advocate (and am still working on) moving to maven. It will
> > force us to get our testing structures in order, will make it
> > easier to add tests wherever required, and will also give us code
> > coverage essentially for free.
>
> Let me clarify that last sentence: maven will give us code coverage
> *measurement* essentially for free.
>
> --steve
>
>


-- 
Martin Ritchie

Re: java unit vs. system tests

[Steve Vinoski <vi...@iona.com>]

On Sep 26, 2006, at 11:53 PM, Steve Vinoski wrote:

> Perhaps I'm missing something, but are the Qpid Java unit tests  
> really unit tests? To me they seem more like system tests. One  
> critical aspect of a unit test is isolation, typically achieved via  
> mock objects, but when I look through the code I don't really see  
> the isolation I expect to see.
>
> Does anybody have any coverage figures for these tests? Perhaps  
> someone familiar with the tests could file JIRAs for the parts  
> known to lack coverage.
>
> Again, assuming I'm not missing something, this is another reason  
> why I advocate (and am still working on) moving to maven. It will  
> force us to get our testing structures in order, will make it  
> easier to add tests wherever required, and will also give us code  
> coverage essentially for free.

Let me clarify that last sentence: maven will give us code coverage  
*measurement* essentially for free.

--steve