brokenness of dri1

[Dave Airlie]

On Thu, Sep 27, 2012 at 2:27 AM, Matt Dew <marcoz at osource.org> wrote:
> I'm guessing this question has been asked a bunch of times, please forgive
> me for asking yet again.
>
> Is it possible to have automated regression and functional tests for this
> stuff?   Are machines needed with those particular cards?  This tinderbox
> just verifies that things compile correctly right? Not actual functional
> test? (Correct me if I'm wrong.)

The problem in a lot of cases is its mostly ancient hardware that
nobody cares about, so we don't get reports for a year after release,
so when people make changes to improve the maintainability of the
codebase, they forget that its actually making it worse for the
suckers who are actually maintaining the codebase, and fixing things.

In this case so far this cleanup of the extensions has broken DRI1 and
GLX in Xvfb etc, so we could probably have caught the second case with
some sort of Xvfb test suite, and I think something that tested Xvfb
and Xorg + dummy or nested might be a worthy start, before trying to
tackle real hw.

Dave.