GNOME on NetBSD/sgimips 5.0

"Jared D. McNeill" <[email protected]> Mon, 02 Mar 2009 16:55:45 -0500
Newsgroups gmane.os.netbsd.ports.sgimips
Message-ID <[email protected]>
Hey folks --

I've spent the past ~week or so building GNOME against native Xorg on my 
O2. Still not finished yet, but I have enough installed that I can at 
least do some testing. Some notes:

  o I am using the Feb 21 crime_drv.so build, along with macallan's
    sample xorg.conf with one change:
      Options "XAANoOffscreenPixmaps" "true"
    This is now the default in native Xorg on -current and 5.0.

  o My O2 firmware insists on configuring the display at 1280x1024
    instead of the panel's native resolution, 1680x1050. Everything is
    distorted as a result.

  o In my limited testing with GNOME I haven't seen any of the rendering
    glitches that I encountered with pwm.

  o If you're using anything that uses fontconfig, you probably want to
    generate a cache in advance, otherwise startup time for the first
    application in the session takes a serious hit as it has to generate
    a temporary copy. I timed 'fc-cache -fv' on my system and it took 88
    seconds, so this should result in a big win for all. If you're using
    GDM you will want to run the same command as root to populate
    /var/cache/fontconfig otherwise GDM will see the same hit.

  o http://gnats.netbsd.org/40814
    lang/mono fails to build on mips, so I had to comment out a few
    things from meta-pkgs/gnome/Makefile (nothing critical).

  o http://gnats.netbsd.org/40932
    gnome-power-manager is built with -Werror, and some of the
    gstreamer0.10 headers are reporting alignment problems.

  o famd performance is awful, not sure why, but top shows it sitting
    around 18% CPU at all times while GNOME is running on my R5k 180MHz.
    Turning it off made things much better.

  o firefox-2.0.0.19 doesn't start, no error messages (epiphany built
    with the firefox2 SDK works fine, however).

  o There's a bug with the volume control applet (seen on x86 as well)
    where it only shows the first mixer control it finds; in mavb(4)'s
    case that is the microphone, so the volume slider doesn't work.

That's all for now, I'll send updates as I get around to testing more 
things.

Cheers,
Jared