Created attachment 28695 [details] [review]
patch to implement wireless details

Here is the first rendition of this patch. It has only been tested on my
laptop, and might have unforseen issues.

What distro are you on?  I think that importing wireless.h is probably a bad
idea, since the version of the wireless extensions will vary from kernel to
kernel.  On my machine (SUSE 9.1), I have wireless.h in /usr/include (from
wireless-tools) and /usr/include/linux (from glibc-devel).

Also, it's Linux-only, so other OSes will need their own way of setting this info.

I like your work very much Davyd... I'm on FC2 with GARNOME and used your patch
with the latest CVS (had to fix a few rejects)

One thing I noticed  is that "Channel" doesn't work, it displays "0 (0 khz)"
while iwconfig clearly says that I am using channel 9.

I really hope that this will make it into official CVS, even if only as an
compile time option. Then if we could just get a dropdown menu which would let
you select your ESSID it would be heaven ;)

Joe,

Ahh, I didn't notice there was a system installed version. Debian doesn't appear
to have one from wireless-tools, however it does have one in the kernel headers.
What is our policy about using to kernel headers?

That said, I don't think the particular ioctl()s I'm using have ever changed, I
had considered just taking the parts out of wireless.h that I'm interested in
(the ioctl() #defines and the structs) and moving them to netstatus-wireless.h.

So we know for a fact that the BSD don't use the same set of ioctl()s?

There appears to be at least one transient crash still in it. It just likes to
not happen too often, and I haven't yet worked out how to trigger it.

I don't think the ioctls have changed either, and they probably aren't too
likely to change.  Moving the #defines in is probably fine but there's little
reason to do that if wireless.h is reasonably installed.

I'm not 100% sure about the BSD wireless stuff, but I'd be willing to bet that
the  interfaces are different.

Ok, I think I've gotten the issues relating to random death sorted out. Non of
it appears to have been in the new code, but in the IPv6 support I wrote a while
ago (bug #138390). 

Yes, there were mistakes in that code, however it's interesting that this would
trigger it. I'm sure there is some sort of icky race there I can't see. I'll get
both patches updated tomorrow evening (after my semiconductors exam), it's a
little difficult separating them out in the tree, I have half of a script to do
it for me.

Hopefully I can get the wireless and IPv6 details stuff in there soon so people
can test it and work out what still needs work.

Linking against wireless.h on the system will probably do. It might also deal
with Michael (above)'s channel problem. I think wireless extensions before 16
handled ranges a little differently. More tomorrow!

Created attachment 29756 [details] [review]
Wireless configuration extensions to the Gnome-Netstatus applet

Comment on attachment 29756 [details] [review]
Wireless configuration extensions to the Gnome-Netstatus applet

This patch provides a set of wireless configuration dialogs for the
Gnome-Netstatus applet.  Please evaluate and let me know what you think. 
Thanks!

Davyd: I think we should use libiw from wireless-tools for this stuff - it
handless differences with kernel versions and stuff. If you look at bug #119472
you'll see we're coming to the conclusion that the signal strength indicator
reported by the kernel is more or less meaningless as it isn't an absolute
metric. So, I reckon I'll make gnome-netstatus use libiw very soon.

Mark: I'm pretty sure that I want to keep the applet as a pretty simple tool for
"monitoring" your network interface rather than a fully fledged network
configuration UI. I could imagine integrating some simple way of switching
between networks and specifying the ESSID but that's about it. I'd really prefer
to keep configuration in vendor supplied tools. But, lets discuss this in
another bug - could you submit your patch to a separate bug and include
screenshots of the bits you added?

I have this sick and twisted idea to try using the daemon from the netapplet (I
think it's called netdaemon) to talk with gnome-netstatus. The daemon seems to
know everything we want to know, I just don't like having 'applets' running in
my notification area. It's just an idea... it might not turn out at all.

It should be very straightforward to do.

Any news on this?

No, I've been slack.

gnome-netstatus development has been stalled [1]. Maintainers don't have
future development plan so i am closing the bugs as WONTFIX.

[1] http://mail.gnome.org/archives/desktop-devel-list/2011-June/msg00073.html