[arch-projects] [netctl][PATCH] Move away from using wpa_actiond

[Emil Velikov]

Hi team,

On Sat, 9 Feb 2019 at 09:19, Thomas Bächler <thomas.baechler at gmx.de> wrote:
>
> Am 8. Februar 2019 20:17:52 MEZ schrieb Jouke Witteveen <j.witteveen at gmail.com>:
> >On Fri, Feb 8, 2019 at 3:36 PM Emil Velikov <emil.l.velikov at gmail.com>
> >wrote:
> >>
> >> On Thu, 7 Feb 2019 at 14:50, Jouke Witteveen via arch-projects
> >> <arch-projects at archlinux.org> wrote:
> >> >
> >> > The same functionality is provided by wpa_supplicant, so we do not
> >need
> >> > an extra and Arch Linux specific dependency.
> >>
> >> The introduction of wpa_actiond [1] hints there are issues with
> >wpa_cli. Namely:
> >> - partial or missing logging capabilities
> >> - race conditions
> >>
> >> Sadly I don't know much more about this. It'll be great if can use
> >> wpa_cli, while not introducing new issues.
> >>
> >> -Emil
> >>
> >> [1]
> >https://git.archlinux.org/wpa_actiond.git/commit/?id=c5c587771403d31ab4538e1c756e0b88a4641a92
> >
> >Thanks! This detective work is highly appreciated! For some reason I
> >had assumed wpa_cli would have gained this functionality only after
> >wpa_actiond came into existence. This couldn't really explain why both
> >use the same parameter names though...
> >I wonder if the issues with wpa_cli have been taken upstream. In my
> >experience, the maintainer of wpa_supplicant is very pleasant to work
> >with.
> >
> >autowifi and, later, wpa_actiond are important parts of the history of
> >netctl, but I tend to give over 10 years of development in
> >wpa_supplicant the benefit of the doubt. This means I would like to
> >try to move away from wpa_actiond anyway. Currently, except from those
> >using wpa_actiond directly, I think netctl is the only user of
> >wpa_actiond.
> >
> >@Thomas: do you have an oppinion in these matters?
> >
> >Thanks,
> >- Jouke
>
> I honestly don't remember why wpa_cli was insufficient at the time and why I wrote wpa_actiond. That must have been way over 10 years ago. I just remember that my first attempt used wpa_cli, and something obvious was missing.

After a bit of debugging I've noticed the move to wpa_cli broke my setup.
Namely: wpa_cli does not reliably detect when the laptop resumes into
another wireless network.
Issue happens fairly intermittently - from once a week to multiple
times a day - yet I have not seen it even once in the years of using
wpa_actiond.

Should I file a archlinux bug report, any suggestions how to proceed?

Thanks
Emil