Fedora 9 (Sulphur)

Docs

Changes

   $ gconftool-2 --type bool --set /desktop/gnome/interface/cursor_blink false

Config

   # yum remove PackageKit

   # chkconfig --level 2345 NetworkManager off
   # chkconfig --level 345 network on

syncookies

At some point Fedora started enabling tcp_syncookies (/proc/sys/net/ipv4/tcp_syncookies) http://ipsysctl-tutorial.frozentux.net/ipsysctl-tutorial.html#AEN485

A recent thread on LKML brought up this message from David S. Miller (amongst over things, Linux's TCP/IP stack maintainer) http://www.ussg.iu.edu/hypermail/linux/kernel/0807.3/0050.html

Given this, I'm going to disable tcp_syncookies.

Issues

Showstoppers

Nautilus

http://bugzilla.gnome.org/show_bug.cgi?id=516429
https://bugzilla.redhat.com/show_bug.cgi?id=447927

Firefox 3

https://bugzilla.mozilla.org/show_bug.cgi?id=421482 good to see Andrew Morton has got involved.

I did strace the Fedora shipped Firefox 3 beta 5 the other day and only saw a few fsync()'s.

Could comment #137 be what happened to Paul?

Some chat on the fsync() issue from Mike Shaver.

Thread on linux-fsdevel about fsync()/fdatasync()

NOTE: This isn't directly a Firefox 3 issue but SQLite.

Others

Touchpad

Seems a change in the synaptics driver has changed the default behaviour of tap to click to be off.

There currently seems to be two options to resolve this.

Editing xorg.conf is preferable.

In the "ServerLayout" section add a line

    InputDevice    "TouchPad" "CorePointer"

Create a new "InputDevice" section

 Section "InputDevice"
        Identifier  "TouchPad"
        Driver      "synaptics"
        Option      "RTCornerButton" "2"
        Option          "RBCornerButton" "3"
        Option          "TapButton1" "1"
        Option          "TapButton2" "2"
        Option          "TapButton3" "3"
EndSection

Optionally add a line in the above section to allow synaptics driver changes from userspace tools.

   Option        "SHMConfig" "on"

Note: After these changes, extra mice plugged in continue to work.
Note 2: Seems the SHMConfig thing is regarded as a "disaster" http://lists.freedesktop.org/archives/xorg/2008-July/037070.html

The good news is that Peter Hutterer has been working on an input properties mechanism based on RandR output properties.