Re: [boost] More tests running with Boost.Test than registered

Subject: Re: [boost] More tests running with Boost.Test than registered
From: Jon Black (jon_black_at_[hidden])
Date: 2011-09-23 02:53:39

On 23-9-2011 7:42, boost-request_at_[hidden] wrote:
>> So I put a print statement in the test in each DLL and ran it again, and
> You could have used --log_level=all to see what is being tested
Thanks. That's much nicer.
>> >  The project for MyDll1 depends on the other projects. That can't be a
>> >  coincidence! I'm confused though, why is the console_test_runner
>> >  executing tests in other DLL's when I've only specified MyDll1 on the
>> >  command line? Each DLL has its own unit_test_init() function declared in
> You probably are using automatically registered test units. These are registered
> in single shared test framework as soon as you load these libraries and they are
> loaded in memory as they are dependencies of the library you want to test. Thus
> you end up with all test units in all loaded libraries.
Yes, they are automatically registered tests. In the init_unit_test() 
method I also give the master test suite a unique name for each DLL. If 
there is on shared test framework, does this mean I'm just renaming the 
same master test suite each time?
> You can specify the test case belonging to the Mydll1 using --run=TestMyDll1 and
> you'll get only 1 test case being executed.
Perfect! I've used run_test to limit to a test suite which I've auto 
registered per project.
> I think this is correct behavior overall. What you got here is implicit
> dependency between test units you can't be sure TestMyDll1 working until you
> tested all of it's dependent test cases which are in this case reside in
> corresponding libraries.
Agreed. I'm running it as a post build step. Adding the run_test option 
to limit what is executed solves my problem. Thanks.
> Gennadiy