new d3d9/device.ok test always fails here, but not a regression?

[Alexandre Julliard]

Dan Kegel <dank at kegel.com> writes:

> On Sun, Jan 1, 2012 at 6:32 PM, Austin English <austinenglish at gmail.com> wrote:
>> That _particular_ test never worked, so it's not a regression. A code
>> change in d3d itself that broke a previously working test/application
>> would be a valid regression.
>
> IMHO a code change in a d3d test that broke a previously working
> "make test" also counts as a regression.
>
> The test suite is *important*.  Breakage in the test suite is not trivial.

Obviously it's important, and we have an importance field in bugzilla
for that. But the regression keyword has a specific meaning: there has
to be a piece of code that worked and then got broken. Bugs in newly
added code cannot be regressions.

Similarly, if we add a dll and some app starts using it and breaks,
technically that's not a regression even though the behavior got worse,
because it has always been broken, it just wasn't exercised before.

-- 
Alexandre Julliard
julliard at winehq.org