Bugzilla Bug 112420

This is just a bug because currently the masked madwifi-driver packages are
working off an old snapshot that seemed to crash alot, and the madwifi-tools
package doesn't install the now vital wlanconfig program.

These ebuilds are for the 1325 svn snapshot of 2005-11-11.  There are no unusual
patches for them, but when a card is insert only a wifi0 device is automatically
created and this is not a normal wireless device.  The new madwifi-ng drivers
now allow the creation of different actual devices based off this underlying
wifi0 device.  So for instance, to create a standard adaptor/station, you must
issue the command:

wlanconfig ath0 create wlandev wifi0 wlanmode sta

(This should mean that you can create an access point, and a standard adaptor
and have them both run from the same card simultaneously, but I haven't tried this).

Unfortunately wpa_supplicant-0.4.6 doesn't work with this correctly, because the
scan initiated by the supplicant never terminates.  This has been fixed in the
CVS version and will be included in 0.4.7.  I have also attached an ebuild for a
CVS snapshot that I created from:

http://hostap.epitest.fi/cgi-bin/viewcvs.cgi/hostap/hostap.tar.gz?tarball=1

To create it, examine the Makefile in the wpa_supplicant directory and move all
required files from the hostap directory to the wpa_supplicant directory, then
modify the Makefile to remove all references to ../hostapd (and due to a type
../hosapd as well).  There is a final patch which the ebuild applies (also
attached), which modifies the makefile to include /usr/include/madwifi as an
include location.  If anyone doesn't want to go through the effort of creating
their own wpa_supplicant-0.4.7-pre1.tar.gz, mail me and I can send you mine.

Created an attachment (id=72823) [details]
madwifi-driver-0.1_pre20051111 ebuild

This is the main driver ebuild, there are no associated patches

Created an attachment (id=72824) [details]
madwifi-tools-0.1_pre20051111.ebuild

The madwifi-tools, which also builds and installs the wlanconfig tool.	There
are no associated patches.

Created an attachment (id=72825) [details]
wpa_supplicant-0.4.7_pre1.ebuild

This is an experimental wpa_supplicant with madwifi-ng enabled.  Note that the
tar.gz referenced in the sources must be MANUALLY created from the hostap CVS
snapshot, and has only one associated patch to ensure that the madwifi include
files are correctly used.

Created an attachment (id=72826) [details]
wpa_supplicant-0.4.7_pre1-makefile.patch

Patches what a normal wpa_supplicant build's makefile would look like, to
ensure the use of gentoo installed madwifi include files.  For use with
attachment 72825 [details] posted above.

I added the new madwifi-driver and -tools version to portage.

brix, can you please take care of wpa_supplicant?

(In reply to comment #5)
> brix, can you please take care of wpa_supplicant?

No, I will not add an unofficial pre-release of a development version of a
package to portage in order to maybe fix something as fragile as madwifi support
at this point.

Upstream is aware of the problem with madwifi and madwifi-ng and has promised to
release a new development snapshot as soon as it has been tested that it doesn't
break more than it fixes. I will add the official 0.4.7 developement snapshot
once it is released.

This will be fixed when a wpa_supplicant development-snapshot is available.
makring LATER for the meantime.

That's fine, not a problem what-so-ever.  I've posted the instructions and
ebuilds here for anyone who really wants to try out the cutting edge stuff, and
I've also covered the problems I found, so getting an ebuild out when the
wpa_supplicant people do release 0.4.7 should be very easy.  Thanks for your
help everybody, you've all been very speedy.

In case anyone's interested I've also tried to update kismet to work with
madwifi-ng (again the developers have recently added support to the stable
version), but whilst it compiles fine, it doesn't seem to see anything from the
madwifi card (the other cards seem unaffected).

Mike,

I appreciate your work on this, I'm just afraid to add to the confusion by
adding an unofficial pre-release of wpa_supplicant at this point due to all the
problems we've had with madwifi support for the last couple of weeks.

Hopefully an official snapshot is not far away.

0.50 is out and in portage, but without madwifi flag.

can it be readded?

I've asked the same question in bug 113432, which deals only with that issue,
rather than with the madwifi ebuilds as well.  Hopefully the owner of the bug
will reopen it and wpa_supplicant will get madwifi support again.