Re: [Boost-users] Boost.Test - problems overriding main

Subject: Re: [Boost-users] Boost.Test - problems overriding main
From: Gennadiy Rozental (rogeeff_at_[hidden])
Date: 2015-09-07 01:39:38

 <ramonpadilla1 <at> yahoo.co.uk> writes:

> >This is because your library was built without 
BOOST_TEST_ALTERNATIVE_INIT_API. 
> 
> Having looked at the documentation, it suggests that I 
> only need to build the library with
> BOOST_TEST_NO_MAIN if I want to supply my own main. Why do 
> I also need to also build with BOOST_TEST_ALTERNATIVE_INIT_API?

Why did you specify it in your test module?

The value of this macro should be the same both during library build and 
when you build test module.

> >> 
> >> I found out that including the following file:
> >> 
> >>     #include <boost/test/included/unit_test.hpp>
> >
> 
> >Single header version will work indeed. 
> 
> I'm not sure I understand. Can you elaborate?

Single header version build both framework and your test module 
simultaneously. Obviously this resolved the issue with discrepancy in
BOOST_TEST_ALTERNATIVE_INIT_API value.

> >> resolves this problem but I'm confused about the implications. 
> >
> >Your only implication is that you will have to spend a bit more time on 
> >compilation (maybe negligible, if you have enough power) and you have 
to 
> >stick with single file test modules.
> 
> Again, can you please elaborate?

What specifically you want me to elaborate on? Single header might take a 
bit more time to build and only allow you to use one test file per test 
module.

> 
> And just to clarify, if I want to provide my own main function, 
> I *must* build the library with at least
> BOOST_TEST_NO_MAIN. 

If you are building static library - yes.

> Under what circumstances would I also build it with
> BOOST_TEST_ALTERNATIVE_INIT_API?
 
It is up to you. You need to make sure you set the same value when you 
build library and test module.