Minutes from August 2 Release Wranglers call

[Paul Anderson]

Release wranglers,

The minutes from today's call:


Attendees
---------
Stuart Kreitman
Jim Gettys
Keith Packard
Egbert Eich
Daniel Stone
Kevin Martin
Alan Coopersmith
Eric Anholt
Paul Anderson
Leon Shiman


The call begain at 8:00 Pacific, 11:00 Eastern.

Kevin thanked the group for all of the extra hours people put in
to get their code checked in before the feature freeze deadline
on Friday.  

Given that we'll only have an hour for today's call, Kevin would
like to plan to have another release-wranglers call on Wednesday.


Freeze status
-------------
Got most things in the tree now.  The main things that didn't make
it were the new trapezoid component to render.  Keith wants to add
some more optimizations in as well.  It was less critical to get
in before feature freeze than the Composite work.  The new trapezoid
code is done and it is self-contained, so Kevin asked if there were
any objections to putting that into the tree now.  Keith noted
that the optimizations aren't done, but the new implementation 
aligns better with the hardware that's out there, but the pending
optimizations will make things faster.  Egbert asked if the Xinerama
code for this is done.  Keith said that it isn't done yet, but it
will be.  The only thing using this part of RENDER is Cairo (Gtk
and Qt are not).  It was felt that this code is self-contained and
will have minimal impact.  Kevin is leaning towards including this
code, and there weren't any other objections to Keith's checking
this code into the tree.

The XAA RENDER hooks: There are some new hooks that can be added
to the tree that will allow some RENDER acceleration.  The question
is do we put this in the tree now, or come up with a new(er)
version of XAA that will allow RENDER to be more easily accelerated?
The hooks will mostly help font rendering, which they speed up by
about a factor of 7.  Kevin noted that this fix spans features
and bugs (we've had a bug open against this for a while).  He'd 
obviously like to fix the bug, as long as the affects on the rest
of the code are minimal.  Eric noted that the Matrox driver will
need the most testing.  Keith feels we should treat this as a
bug - the current RENDER acceleration hook is broken, and we should
either fix the current hook or disable it.  Kevin feels that we
should treat this as a bug, and that along with this work, we should
implement the flags to disable these hooks.  If this change goes in,
Kevin would like to get the change in early so that we can get as much
testing as possible.  We need to get testing coverage from as many
driver owners as possible as well.

Egbert mentioned a DPMS issue.  There is a problem with several
drivers that don't get out of the DPMSOff state when the server
shuts down.  This can prevent users from being able to use
the console after the server terminates.  Kevin said that the 
Radeon driver he saw did this correctly.  Egbert will take a closer
look.

pkg-config: Daniel has been making some fixes to this to add
some new targets to xorg.tmpl.  This is used by autotools for
externally-maintained libraries - they will benefit by having
this information available.  Kevin said that this won't affect
any functionality, but he wanted to discuss it here, since it's
out of the norm.  There were no objections to including these
changes.  Daniel will get the patch into bugzilla so that others
can see it.


Daniel raised an issue (bug 955):

http://freedesktop.org/bugzilla/show_bug.cgi?id=955

about some autoconfig/getconfig code that may be under a non-free
license and therefore should be removed.  Egbert believes that
the code this was published under the old license, in which case
the license issue doesn't apply.  Daniel asked if we can get some
outside opinions on this.  Leon will contact Steve Nunn for an
opinion on this.  Egbert will update the bugzilla entry to contain
his knowledge of this.


freetype:  People feel that the goal should be for things to
compile properly against old/new versions of freetype.  Egbert
volunteered to write the code if nobody else has done it yet. 
Egbert will also follow-up with Adam Jackson.  Egbert's patch
will take the information from the freetype header file about
the version, and the new code will be compiled conditionally
if the new version is 2.1.8 or higher.  Any 2.1.7 freetype
library will continue to build and work.  This will have no
affect on externally-maintained applications.  If people want to
get acceleration, they'll need to use 2.1.8 or higher, but
the compilations will no longer be broken with lower versions.
Egbert will test with versions 2.1.7 and 2.1.8 of the library,
but he would like to get others to test.  Egbert will also
ping Chisato to do some testing.


patches for dll modules:  Kevin noticed that these broke the
static build.  Matthieu Herrb has a patch for this.  Kevin
noticed that if you build a non-loadable server, the radeon
driver will pass the test suite, but a loadable version will
not.  He has not had time to track this down yet, but it's
an important issue that he wants to understand.


Kevin asked people to look through bugzilla prior to the next
meeting.  We have 181 bugs in there, and he'd like to start
discussing them during the next meeting. 

The next call will be Wednesday, August 4.
Unless you hear otherwise, the dial-in numbers will be:

    11:00EST, 8:00PST, 5:00PM CET
    Access numbers:
      1-866-639-4713 (toll free)
      1-574-948-0365 (toll)
    Passcode: 5517683


The call adjourned at 9:00 Pacific/12:00 Eastern.

-paul