Re: iw - set freq regression

[Johannes Berg <johannes@sipsolutions.net>]

On Mon, 2019-12-09 at 20:43 +0100, Janusz Dziedzic wrote:
> Hello,
> 
> For v4.9 this command works correctly:
> ./iw wlp1s0 set freq 5220 20 5220 0
> ./iw wlp1s0 set freq 5180 40 5190 0
> 
> With latest master this fail with err msg:
> Usage:    ./iw [options] dev <devname> set freq <freq> [NOHT|HT20|HT40+|HT40-|5MHz|10MHz|80MHz]
>     dev <devname> set freq <control freq> [5|10|20|40|80|80+80|160] [<center1_freq> [<center2_freq>]]
> Options:
>     --debug        enable netlink debugging
> 
> After bisection and revert this one, works as before:
> 4871fcf iw: parse_freqs: check how many center frequencies to expect
> 
> I know we don't need to set this all arguments and finally I can change my script (don't set cf1/cf2).
> Anyway report this.

Thanks. Hmm. So ... dunno. It seems a bit more user-friendly to not have
the center_freq2, but OTOH having it as 0 at least won't really matter?

I think we needed this for e.g. handle_start_ap() to not be (quite as)
ambiguous since the chandef is followed by the beacon interval. If the
center2_freq is _always_ consumed by the chandef, then a beacon interval
of 200 would be like 200 MHz, and the result would fail here ...

So not really sure. I guess we could add another argument or check if
the last arg ('parsed') is non-NULL to change the behaviour, but is it
worth it?

johannes