Re: [ALSA PATCH] 1.0.9b+

[Takashi Iwai <tiwai@suse.de>]

At Thu, 28 Jul 2005 17:07:38 -0400,
Dave Jones wrote:
> 
> On Thu, Jul 28, 2005 at 03:06:02PM -0400, Lee Revell wrote:
> 
>  > > The git-alsa.patch in -mm which I obtain from
>  > > master.kernel.org:/pub/scm/linux/kernel/git/perex/alsa-current.git is
>  > > empty.  So we're now wanting to merge 4,000 lines of unreviewed code which
>  > > hasn't been tested in -mm at approximately the -rc4 stage.
>  > 
>  > Lots of people install ALSA independently from the kernel (like all the
>  > audio oriented distro users), probably a lot more than run -mm, so it's
>  > not completely unreviewed.
> 
> There's a big difference between 'unreviewed' and 'untested'.
> With drivers that support n variants of a piece of hardware,
> hearing back that a driver works fine from someone isn't really
> worth anything if the changes break for all the other users
> of that driver that have a different variant, which haven't tested it yet.
> 
> We have a testing/review process in place, attempts to short-circuit
> it should be prevented. Especially from a subsystem that
> historically has had a number of issues wrt regressions each release.

Oh, I missed your talk in OLS ;)

Basically I agree with the argument above.  Some hardware-specific
regressions are very hard to find only from review.  The famous
problem with ThinkPad in the last update was such a kind.  More tests
in a bigger place like mm tree would be required, indeed.

Unfortunately, the sync between ALSA and mm tree didn't work well
because of troubles with git/cogito at this time.  The sync process
shall be improved from now on.  So, please consider this as an
exception.


Takashi