[cairo] A theme (and challenge) for 1.8.2: Eliminating XFAILs!

[Carl Worth]

On Thu, 2008-09-25 at 16:37 -0700, Carl Worth wrote: 
> So we finally got 1.8.0 out. Congratulations, everybody! I think it
> really is a fantastic release.

And so far, nobody has contradicted me on that. I'm still hearing
positive reports from early adopters.

Adrian Johnson did notice that there was a regression in the
fallback-resolution test case, but we missed this because the test suite
wasn't automatically verifying the result of that test. Chris Wilson has
now rewritten that test to do automatic verification, so we won't have
that problem in the future.

> 	DISPLAY= make test
> 
> The xlib failures I just ignored. I know that there are at least some X
> driver bugs there, (and for me using Intel graphics I'm on the hook for
> those bugs). So I'll be fixing those up soon.

I won't be so lazy with cairo-xlib testing for future releases. Instead
I'll ensure that running cairo-xlib against an all-software X server,
(like Xfake or Xvfb) passes the entire test suite. This is exactly what
I've done for previous major releases.

Any remaining cairo-xlib test-suite failures with a "real" X server
should be either minor rasterization differences, (which we can
hopefully avoid with a suitable tolerance value in the test suite), or
else actual bugs in the driver. For example, I fixed a bug this week in
the Intel driver that was exposed by the device-offset-fractional test
case.

> Instead I'd like for us all to focus on
> cleaning up some known bugs and get out a 1.8.2 release soon with many
> fixes. The goal is to have no remaining XFAIL tests.
...
> So that's 20 failures. What do you think? Can we fix 2 per week and ship
> a failure-free 1.8.2 in 10 weeks?

I'm amazed! It's only been one week and already 7 of those 20 failures
have been eliminated, and I believe we have fixes in hand for two more
already, (extend-pad and self-intersecting). This is phenomenal
progress. Way to go, everybody!

During that same week, Chris Wilson has been putting in some fantastic
triage work for our bugs in bugzilla, reducing the number of open bugs
against cairo from 212 to 173. The signal-to-noise ratio is already
improving and should only keep getting better, (particularly as I've now
suggested that we copy any wishlist items to
http://cairographics.org/todo and close their bug entries).

So I'm thinking that within another week we'll perhaps be ready for
1.8.2? We may not have every XFAIL fixed by then, but maybe we'll know
that any that are left will require significant effort.

If so, then we can start landing all of the really interesting code that
has been piling up, (Joonas' polygon rasterizer, Chris' total rewrite of
the test suite, Adrian+Chris' support for embedding an image only once
into a PDF file[*], etc.).

This is fantastic stuff, and I *love* seeing the cairo community so
active. Keep up the great work! or rather, Keep up the great fun!

-Carl

[*] And yes, this means I'll actually write the copy-on-write code I
promised at the Cairo Summit.
 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://lists.cairographics.org/archives/cairo/attachments/20081003/cddfc83b/attachment.pgp