Right way to cope with user error in make test?

[Steven Edwards]

Hey Dmitry,

I'd like to reply point by point but Rolf's message sums up the way I
feel about the framework. I've avoided using it more than I could have
and am going to be doing a lot more work with it in the future so I
have in interest in seeing it pass 100% on more than just Julliards
box. I did want to reply to one question you asked but I'll leave the
rest rather than just adding more noise to my original point on fixing
the tests for soft dependancies.

On Sun, May 18, 2008 at 11:36 PM, Dmitry Timoshkov
<dmitry at codeweavers.com> wrote:
> Why? Does it magically make something useful? Again, what about other
> packages
> built from source?

Other packages built from source are not trying to run third party
closed source applications on a platform for which they were not even
designed. Its not like there is a whole suite of open source win32
applications out there that use Winelib and are constantly being
maintained where breakages are spotted quickly. If someone builds QT
or GTK and then builds an application using that, odds are they are
going to have much less trouble spotting the breakage because they
have the source for both the library and application. In essence the
existing poll of Free Software out there that is being constantly
built and repackaged (sometimes daily, see gentoo for example), acts
as a testing framework for those libraries. Your comparing apples to
oranges. We simply don't have the luxury of a large ecosystem catching
a regression the way most Linux libraries do. Accepting brokeness in
the test suite in any form is simply flawed IMHO.

-- 
Steven Edwards

"There is one thing stronger than all the armies in the world, and
that is an idea whose time has come." - Victor Hugo