[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#491601: madwifi-source fails to build against 2.6.24-etchnhalf.1-486

Hi dann, Michael,

On Thursday 24 July 2008 06:46:14 dann frazier wrote:
> On Wed, Jul 23, 2008 at 08:42:20PM +0200, Michael Goetze wrote:
> > Hi,
> >
> > On 21.07.2008, at 08:40, dann frazier wrote:
> >>
> >> Also note that the ath5k driver is included in the 2.6.24 kernel - I
> >> don't personally know how much hardware works with madwifi but not
> >> ath5k though.
> >
> > Well, for one thing, it's missing some exotic features like, say, Master 
> > mode. ;)
> Yep, that's a big one :) It also seems to not work well w/ WPA (at
> least, not on the laptop I tried).

Precense of ath5k also requires that madwifi-tools recieves a stable update
to include a modprobe blacklist file to avoid conflict with madwifi. Attached
debdiff for proposed update.

> Again, we're open to adding a madwifi-etchnhalf package in a point
> release. For 4.0r4, the window is closing very quickly since the
> release is scheduled for this Friday, but a 4.0r5 can hopefully happen
> ~2 months later.
> The requirements for such a package are:
>  * module either adds new hardware support for etch, or achieves
>    feature parity with the etch 2.6.18 kernel. etchnhalf is
>    specifically targeted at adding hardware support, so adding things
>    like new filesystems would be out-of-scope.
>  * module must not be rc-buggy (obviously)
>  * module version must be <= version in lenny

I backported the 1:0.9.4~rc2-1 package which was in lenny archive for a long
time until just recently. I then gave it a version bump to actual upstream
release version 0.9.4.

>  * source and binary packages should have -etchnhalf appended
>  * module source should get installed in /usr/src/etchnhalf (this
>    lets the package coeexist w/ its etch counterpart)
>  * Included documentation should explain how to point m-a at the new
>    location (e.g., # MOD_SRCDIR=/usr/src/etchnhalf m-a b-i). Note
>    that a-i and other commands that attempt to fetch source won't
>    work - the -etchnhalf namespace breaks that :(

This is a bit crap. It is also able to be overcome by providing a modass
override script similar to:


. ${MA_DIR:-/usr/share/modass}/packages/generic.sh

$1 "$@"
at /usr/share/modass/overrides/foo-source-etchnhalf. This allows the foo-source
module package to work without overriding variables, only providing full
package name including -etchnhalf:
m-a a-i foo-source-etchnhalf

Providing an override file for each modules backport may improve user
experience, and solve a possible conflict described below.

>  * source package should omit any redundant packages
>  * module has been tested in etch w/ etchnhalf kernel (i.e.,
>    compatible with etch userspace)
> squashfs-etchnhalf, atl2-etchnhalf and loop-aes-etchnhalf are examples
> of this.

squashfs-source and proposed squashfs-source-etchnhalf [0] both seem to provide
the same symlink:

gspca-source and gspca-source-etchnhalf are probably similar. This seem like
a problem to me. 

[0] http://people.debian.org/~dannf/etchnhalf-oot/squashfs.patch

Please test/consider the below backports for one of the upcoming stable
updates. Michael if you could test them that would help alot, i've not able to
test the driver, on etch, with supported hardware, have only been able to
ensure it is compiling in etch with Linux 2.6.18 and 2.6.24.


Thanks, Kel.
diff -u madwifi-tools-0.9.2+dfsg/debian/control madwifi-tools-0.9.2+dfsg/debian/control
--- madwifi-tools-0.9.2+dfsg/debian/control
+++ madwifi-tools-0.9.2+dfsg/debian/control
@@ -2,7 +2,7 @@
 Section: contrib/net
 Priority: optional
 Maintainer: Debian madwifi team <pkg-madwifi-maintainers@lists.alioth.debian.org>
-Uploaders: Loic Minier <lool@dooz.org>, Kel Modderman <kelrin@tpg.com.au>, Matt Brown <debian@mattb.net.nz>, Alex Wallis <awol@otaku42.de>
+Uploaders: Loic Minier <lool@dooz.org>, Kel Modderman <kel@otaku42.de>, Matt Brown <debian@mattb.net.nz>
 Build-Depends: cdbs, debhelper (>= 5), dpatch
 Standards-Version: 3.7.2
diff -u madwifi-tools-0.9.2+dfsg/debian/changelog madwifi-tools-0.9.2+dfsg/debian/changelog
--- madwifi-tools-0.9.2+dfsg/debian/changelog
+++ madwifi-tools-0.9.2+dfsg/debian/changelog
@@ -1,3 +1,12 @@
+madwifi-tools (1:0.9.2+dfsg-1etch1) stable; urgency=low
+  * Install /etc/modprobe.d/madwifi for the purpose of blacklisting
+    ath5k and allowing madwifi to function without conflict.
+  * Update my maintainer email address in debian/control. Also remove
+    Alex from Uploaders field, may he rest in peace.
+ -- Kel Modderman <kel@otaku42.de>  Thu, 24 Jul 2008 19:38:28 +1000
 madwifi-tools (1:0.9.2+dfsg-1) unstable; urgency=low
   [ Kel Modderman ]
diff -u madwifi-tools-0.9.2+dfsg/debian/rules madwifi-tools-0.9.2+dfsg/debian/rules
--- madwifi-tools-0.9.2+dfsg/debian/rules
+++ madwifi-tools-0.9.2+dfsg/debian/rules
@@ -27,6 +27,8 @@
 	install -D -m 0755 debian/post-down \
+	# modprobe.d blacklist
+	dh_installmodules -p madwifi-tools --name=madwifi
only in patch2:
--- madwifi-tools-0.9.2+dfsg.orig/debian/madwifi-tools.madwifi.modprobe
+++ madwifi-tools-0.9.2+dfsg/debian/madwifi-tools.madwifi.modprobe
@@ -0,0 +1,19 @@
+## ath5k (mac80211)
+## Comment out the following line, and uncomment all of the
+## madwifi modules below to use the ath5k module
+blacklist ath5k
+## madwifi (non-free)
+#blacklist ath_hal
+#blacklist ath_pci
+#blacklist ath_rate_amrr
+#blacklist ath_rate_onoe
+#blacklist ath_rate_sample
+#blacklist wlan
+#blacklist wlan_acl
+#blacklist wlan_ccmp
+#blacklist wlan_scan_ap
+#blacklist wlan_scan_sta
+#blacklist wlan_tkip
+#blacklist wlan_wep
+#blacklist wlan_xauth

Reply to: