[cairo] Cairo meeting tomorrow 13:00 EDT

[Carl Worth]

On Mon, 02 Apr 2007 18:09:08 -0400, Behdad Esfahbod wrote:
> The agenda is to discuss the upcoming 1.4.4 release, general 1.6 plans,
> and last but not least, recent patches done by Chris Wilson and others.

Thanks to everyone that came to the meeting. Here is my writeup of the
discussion that occurred and the decisions made.

The IRC nicknames of those who contributed during the meeting are
behdad, cworth, desrt, gpolo, ickle, jrmuizel, ssp, tbf, tor, vlad,
and xan. Some of those nicknames might be unfamiliar as they bear little
resemblance to either names or email addresses of regular contributors
to the cairo mailing list. I believe that "ickle" corresponds to Chris
Wilson, for example, and "tbf" corresponds to Mathias Hasselmann.

Some unknown number of others might have followed the meeting
silently.

The major decisions of the meeting are as follows:

cairo 1.4.4: Scheduled for next Monday, 2007-04-09. Consisting of
what's currently in cairo's master branch plus the following:

  • Mathias Hasselmann's "static region" patch series, (fewer calls to
    malloc for pixman regions), to be reviewed by Behdad, and
    performance tested by myself.

  • Chris Wilson's many recent cleanups, (to be reviewed by Behdad and
    myself).

  • I'll ensure that Monty tests the latest code with his
    multi-threaded application, (which was working with 1.4.2 but
    broke sometime afterwards).

  • Other bugfix patches as existing in bugzilla or on the mailing
    list, (we've marked some of these as "blocker" in bugzilla---let
    us know if we've missed others).

cairo 1.6: Scheduled for July 2007, (the two proposed dates were "end
of June" to be useful for Mozilla and "before GUADEC", which is the
third week of July). Useful things that are candidates for 1.6, and
that are beyond the scope of 1.4.x include:

  • Change from 16.16 to 24.8 (or so) for fixed-point storage.

  • Implement a faster lookup-table-based rasterizer

  • New cairo-xlib API to support IncludeInferior Pictures

  • Make cairo-quartz a supported backend

  • Make cairo-xcb a supported backend

  • A significant improvement to pixman compositing code, (see
    jrmuizel's recent work and other ideas going on).

  • New, user-font API (behdad and krh)

  • Other proposed API additions (Behdad, who has been disappointed to
    not get much feedback on these yet---Behdad, care to summarize
    these in a single mail with links to the proposals?)

We also discussed other useful work to be done, which could be part
of some subsequent 1.4.x release or else in 1.6.x:

  • Fix all remaining XFAIL failures from the test suite

  • Eliminate problems with "unsupported image formats" for xlib,
    (very frequently duplicated bug reports and a regression since
    cairo 1.0.x apparently)

  • Discard (and clip) trapezoids that extend beyond the clip region,
    (this relates to the slow, giant focus rectangle behavior recently
    reported on the GTK+ devel list).

  • Perhaps add dashing support to the fast-path rectilinear stroking
    code, (again, for GTK+ focus rectangles).

Other issues discussed at the meeting:

How should multiple concurrent branches, (say, "1.4" and "1.5" or
"master"), be distinguished. Options discussed included restricting
the stable branch to only bug fixes vs. restricting the master branch
to only changes that add API. The consensus was to allow things like
performance fixes as well as bug fixes on the stable branch, but to
restrict API changes and "large" changes to a separate branch. (This
consensus was used in the above breakdown of features.)

I proposed a different model for the stable branch than we have used
in the past. Previously, we had applied all changes to master and then
cherry-picked back to the stable branch. I floated the idea of instead
committing stable-branch-appropriate changes directly to that branch
and then merging it into the master branch.

There was a side conversation about cairo performance. Cairo 1.4 is
seen as largely solving the performance problems for embedded
systems. Work that remains could include:

  • Fixing the X server to accelerate Render with GPU (outside of
    cairo scope, but some of the players may be the same)

  • Potential new project to get GPU acceleration for cairo image
    surface

  • Current work to improve compositing software performance in pixman
    (jrmuizel and others)

  • Current work to share pixman and X server's fb (ssp)

  • Potential new work to improve polygon compositing by doing it in a
    single scanline pass rather than tessellating, rasterizing, then
    compositing the entire polygon.

Thanks again, I feel this was useful and would be glad to participate
in future meetings as necessary (or even regularly, if desired).

-Carl
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.freedesktop.org/archives/cairo/attachments/20070403/48a974e7/attachment.pgp