Re: [boost] Better testing; was type_traits_test

From: Beman Dawes (beman_at_[hidden])
Date: 2000-11-02 20:59:37

Jeremy Siek wrote:

 >In addition to running test suites for each library, I would
 >also like the regression test to compile any examples
 >and diff the results with the expected output. The purpose
 >here isn't to test the library... just to make sure the
 >examples still work as expected so users don't get confused.

Yes, that would be quite useful.  I gather from both Dietmar's experiences 
and from reading the docs, that DejaGnu is very good at that sort of test.

But that's just too much to tackle right now.  I think we can have basic 
tests running in another day or two on the files covered by the compiler 
status tests.  I'd rather have a bird in the hand than two in the bush.

Timing tests can also be useful.  In some industrial software I maintain, 
an unexpected change in the timings have several times been the first 
indication of a hidden program logic bug.  But likewise, that will have to 
wait until later.

--Beman