Re: [boost] Re: How do regressions tests work?

From: Vladimir Prus (ghost_at_[hidden])
Date: 2003-11-19 10:15:59

Stefan Seefeld wrote:

> >>I had a look in array1.cpp and all it does is output to stdout. I
> >>don't see any asserts, and I cannot find a file of expected output to
> >>compare to. Is just running to completion enough to pass the test, or
> >>am I missing something?
> >
> > Nope, you're not missing anything.  It might be possible to add
> > expected results comparison, but that's not currently part of the
> > system.
>
> Did any of you have a look at qmtest
> (http://www.codesourcery.com/qm/qmtest) ? Using that it's relatively
> simple to define what 'to run a test' means, and to set up a test
> database suiting your needs. It's very portable and robust.
> One of its main 'clients' is the gcc development.
> A new release will be done shortly...

I think I've mentioned QMTest some time ago, but I'm not sure what followed. 
Speaking for myself, while QMTest is good for me (I use it for system tests), 
it lacks one of the primary advantages of Boost regression system: it does 
not track dependencies between sources. I.e. if you change a single source, 
Boost regression system will recompile/run only binaries that depend on it. 
With QMTest, you'll rerun everything. At least I don't know a way to prevent 
that.

And.. running unit tests during rebuild is much more convenient than  "Run All 
Tests" menu item in QMTest ;-)

- Volodya