[boost] Re: Boost.Test choice

From: Markus Schöpflin (markus.schoepflin_at_[hidden])
Date: 2002-09-19 12:04:22

Rozental, Gennadiy wrote:> --- Beman Dawes <bdawes_at_[hidden]> wrote:
 >
 >> Peter,
 >>
 >> Unfortunately, the new Boost.Test Version 2 breaks the old
 >> regression test.  You will need to switch to the new bjam based
 >> regression tests.
 >>
 >> I've been trying to convince Gennadiy that it is a bad idea to
 >> gratuitously break old code, but so far he has resisted changes to
 >> recognize BOOST_INCLUDE_MAIN and revert to the old behavior.
 >>
 >> --Beman
 >
 >
 > On one hand I would like the users to use latest, more expressive,
 > test framework. While on the other hand I would not want to force any
 >  undesired burden on existent Boost.Test users. So I will yield to
 > any community decision if we will be able to find the consensus on
 > one. Here is the choice.
 >
 > 1. Leave as is. Bjam based testing already new Boost test aware and
 > is working. Old regression testing does not work. It should not be
 > used any more at all or it will need to be changed to link with
 > appropriate Boost.Test component. Out of Boost users need to change
 > their projects/makefiles to link with appropriate Boost.Test
 > component or use boost/test/minimal.hpp instead of
 > boost/test/test_tools.hpp + BOOST_INCLUDE_MAIN.
 >
 > 2.Incorporate switch to boost/test/minimal.hpp into
 > boost/test/test_tools.hpp based on BOOST_INCLUDE_MAIN Jamfiles will
 > need to be changed back to not link with Boost.Test components. In
 > all other cases it should work as is This option has also one more
 > decision point: should we keep this switch temporary for this
 > release. or permanently. In later case only those who intentionally
 > do not define BOOST_INCLUDE_MAIN will use complete Boost.Test
 > functionality.
 >
 > Comments?
 >
 > Gennadiy.

You have all my votes for option 2 (if I have any). I just started to 
try out a whole lot of code with the upcoming boost release and suddenly 
most of the tests start breaking. This isn't very nice.

I don't see a problem with the macro BOOST_INCLUDE_MAIN. It's not as if 
you have to write something extra to get the new thing, you have to 
write the macro to get the old stuff. And you get the bonus that a lot 
of old code will keep working as it is.

I'm quite sure we will get a lot of complaints here about this if we 
don't do it backwards compatible but you may have another point of view 
on this.

Markus