[pkg-wpa-devel] Bug#374342: Needs more testing

[Reinhard Tartler]

On Wed, Jun 21, 2006 at 09:12:22PM +1000, Kel Modderman wrote:
> I am fine if you feel that this package needs more testing for a while, 
> however, I would like to point out what small problems occur if we roll 
> back to the 0.4 series of wpa_supplicant upstream:
> 
>    * non-backwards compatible wpa_supplicant.conf options may disrupt
>      the end users configuration, causing wpa_supplicant to fail when
>      parsing the file

This is another good point for having filed #374342.

>    * we can offer a self-contained roaming solution with the 0.5
>      series, that may fill the void for users of the old init style
>      roaming (even Felix!)
>    * 0.5 series contains some patches to enhance operation with some driver

These points stand for quick inclusion of the 0.5 branch to testing.

> Madwifi will eventually propagate to testing, and the wpa_supplicant 
> version in same archive will not support it.

You mean Version 0.svnr1644.0.9.0-2? Perhaps we can coordinate
transition of the two packages? On the other hand, we can tell users to
install wpasupplicant .deb from unstable in their testing system.

> Also, there have been no bug reports since last upload indicating large 
> problems with functionality. (I realise it is still early days)

I remember #373776, which was bounced from Elimar Riesebieter
<riesebie at lxtec.de> to pkg-wpa-devel. Ok, this is a madwifi-ng bug. Bugs
which directly relate from the 0.5 branch did not appear yet.

To bring up my reservations to this list (I already mentioned them on
irc). I think the 0.5 branch is a bit dangerous, because:

 a) the 0.5 branch didn't get as much test coverage than the 0.4 branch.
Bugs are more likely.

 b) Semantic changes in the code of new features, which are still under
development. I fear that behavior changes may disrupt our package

 c) config file changes like mentioned above.

> I am confident that the 0.5 series will be tagged as "stable" by Jouni 
> before the time that Etch is released (quite a few months away yet, 
> iirc), therefore, I'd like to concentrate more on testing and developing 
> infrastructure around 0.5 before that time comes.

Yes, afaik, etch is currently targeted for december this year. I think
that we are both (well, espec. you ;) active enough to stabilize a
wpasupplicant 0.5 for etch release. So yes, I mostly agree with you. I'd
still appreciate if we could have the 0.5 branch tested for say, 2 more
weekends and wait for reports. I don't expect many bugs, but I'm rather
conservative for users of 'testing'.

Gruesse,
	Reinhard