Re: [Madwifi-devel] release

[John D. <js...@av...>]

Michael Renzmann wrote:

> * port missing features from CVS HEAD to trunk
>   * fix at least every "more-than-just-annoying" bug reported by public
>     testers
>   * have trunk/ in a similar stable state as madwifi-old
....
> (at least most of the bugs that lead to kernel oopses are,
> IMO, more than just annoying). 

Yes, kernel panics are more-than-just-annoying.

> Suggestions for such a reordering are highly welcome.

Let me concur/suggest/beseech/implore that the priority queue should be
reordered so that known panic scenarios get immediate, serious, scrupulous
attention.  IMHO such issues should take priority over any imaginable
new features or missing features.

John Bicket wrote:

>> I feel like for the most part the current svn version is stable for use
>> in plain managed or master mode, which is what 90% of people do when
>> they use the driver.

Well, I must be in the other 10%.

Being able to set up AP *and* STA functionality is a documented feature
of madwifi-ng.  I need this feature.  Carrying out the documented commands,
verbatim from the documentation, causes a kernel panic.  For details, see
  http://madwifi.org/ticket/182

The existence of open tickets at this level of severity disqualifies the
product from being considered "stable" IMHO.

This issue obviously affects those of us who need AP+STA functionality,
but it less-obviously affects everyone, because it raises questions
about the design and/or coding of the basic functionality.  A vulner-
ability that can be easily provoked by one method might well be less
easily -- but still occasionally -- provoked by other methods, leading
to nasty irreproducible crashes.  Code is either correct or it is not.
Correct code doesn't crash.  Crashable code is not stable code, almost
by definition.