January 29, 2004

waproamd, Fedora Core 1, and cloaked access points

Here’s a message I sent to the author (I hope) of waproamd:

        Hi.  I just downloaded and started using waproamd.  Thanks
very much for this software.
        I've made RPM packaging for waproamd and modified its initscript
to work with Red Hat/Fedora (I'm running Fedora Core 1) tools.  I've
put the two patches I had to make against waproamd to allow me to
build the RPM (SYSV_INITDIR vs. SYSINITDIR in some files is probably a
bug; see waproamd-0.5b-initdir.patch) and the modifications to the
init script (waproamd-0.5b-initscript.patch) up at


        Also, I have to associate to some cloaked networks.  "iwlist
scan" and waproamd in logs reports these access points as having ESSID
"" (blank) before I manually set the ESSID.  After I set the ESSID
"iwlist scan" sees the cloaked access points twice: once with blank
ESSID and once with the ESSID I just set it to.  My script for such an
access point looks like:

case "$1" in
                ESSID=TheCloakedSSID ifup wlan0
                ifdown $IFACE
I put this file in /etc/waproamd/scripts/<MAC of AP>.  The interface
comes up successfully, but immediately after the script executed
waproamd seemed to be "unsetting" the ESSID (or setting it to blank).
I surmise this is because of the call to iw_set_essid in
iwapi.c's iw_tune(), which is called immediately after setting up the
AP.  I've made a patch which simply avoids the call to iw_set_essid if
the essid field in the ap_info structure is NULL or blank.  This patch
is at


        I hope this might be helpful to you or other users.  I'll
probably put this message up on the web somewhere as well, unless you
can identify a particularly grievous error I've made.
Thanks again,

All that stuff I mentioned is at http://www.codefu.org/people/darkness/waproamd/ for those of you too slack to type it in.

January 28, 2004

Cocoon, Resin, JDK 1.4.2_03, Fedora Core 1 Linux

Mmmm. Trying to set this up. The hello world JPEG and PNG examples were failing with an error quite like this:

        at java.lang.Class.forName0(Native Method)
        at java.lang.Class.forName(Class.java:141)
        at java.awt.GraphicsEnvironment.getLocalGraphicsEnvironment(GraphicsEnvironment.java:62)
        at sun.awt.font.FontResolver.<init>(FontResolver.java:46)
        at sun.awt.font.FontResolver.getInstance(FontResolver.java:120)
        at java.awt.font.TextLine.getFontAtCurrentPos(TextLine.java:984)
        at java.awt.font.TextLayout.standardInit(TextLayout.java:721)
        at java.awt.font.TextLayout.<init>(TextLayout.java:613)
        at org.apache.batik.gvt.text.BidiAttributedCharacterIterator.<init>(Unknown Source)

To fix, in Resin, put <system-property java.awt.headless="true" /> in resin.conf under the <resin> tag. I am amazed this needed to be done — and more amazed that I couldn’t easily find an explanation for it with Google. Finally found it after looking at this message on resin-interest.

Update: Oops: http://cocoon.apache.org/2.1/installing/index.html#Headless+UNIX+and+PJA. Guess I should read more carefully.

January 27, 2004

Fedora Core 1 on Thinkpad T40

Check my archives for what I did to get RH9 working on my laptop. I upgraded from RH9 to Fedora Core 1 (FC1; for the search engines). To fix suspend, I had to make sure I had CHANGEVT=7 commented out in /etc/sysconfig/apm, then do something like this in /etc/sysconfig/apm-scripts/apmcontinue:


case "$1" in
        chvt 1
	chvt 7

I couldn’t use CHANGEVT=7 because it would chvt before suspend, which would cause machine to lock up rather than suspend. It looks like DAG has all the RPMs you could want for FC1, such as thinkpad, tpctl, slmodem (slmdm), and kernel-module-thinkpad.

January 5, 2004

Periods in aliases with qmail

For search engines: periods in a dot-qmail file name

See http://herzl.nylug.org/pipermail/nylug-talk/2003-January/009824.html for the full story, but basically: use a colon instead of a period. So .qmail-j.doe should be .qmail-j:doe.