Subject: Re: [boost] Boost Test Policy
From: Robert Ramey (ramey_at_[hidden])
Date: 2010-08-25 03:51:26
Dave Abrahams wrote:
> On Tue, Aug 24, 2010 at 7:47 PM, Edward Diener
> <eldiener_at_[hidden]> wrote:
>> I also found the lightweigt_test.hpp in Boost detail, so I was
>> wondering whether or not that is now considered a viable enough
>> choice for designing Boost component's tests.
> It always has been. While Boost.Test has some cool features, they
> don't seem to markedly improve the quality of testing in a library
> that uses Boost.Test, which mostly depends on the skill of the test
> designer. Also, if you use lightweight_test.hpp you are liable to
> continue to get useful test results even if Boost.Test should be
> broken ;-)
Actually this can be a big problem. I was having trouble testing
on my own machine and in the trunk because boost.test was
being worked on while I was using it. So I moved to lightweight
test so that I would be dealing with only one thing at a time.
Now, I run all my local tests against the release branch of
all the libraries not my own. So this would not be a problem
for me to move back (except the work of actually moving back).
Of course the trunk tests still test all the library changes
"all at once" so the problem can still occur. But I don't
check into the trunk until I've tested here so it's not such
a big deal anymore.