Re: ALSA: nm256: Fine-tuning for three function implementations

From: Takashi Iwai <tiwai@suse.de>
To: SF Markus Elfring <elfring@users.sourceforge.net>
Cc: alsa-devel@alsa-project.org,
	Arvind Yadav <arvind.yadav.cs@gmail.com>,
	Jaroslav Kysela <perex@perex.cz>,
	Takashi Sakamoto <o-takashi@sakamocchi.jp>,
	kernel-janitors@vger.kernel.org,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: ALSA: nm256: Fine-tuning for three function implementations
Date: Tue, 28 Nov 2017 16:21:20 +0100	[thread overview]
Message-ID: <s5h4lpeqvin.wl-tiwai@suse.de> (raw)
In-Reply-To: <a1cd0902-c8bb-9813-8d0c-d394975ef745@users.sourceforge.net>

On Tue, 28 Nov 2017 16:01:47 +0100,
SF Markus Elfring wrote:
> 
> >>> Give the test result before speaking too much.
> >>
> >> Which concrete data do you expect here?
> > 
> > Depends on the result.
> 
> How can this vary?

How?  Because I didn't see any test result from you, so I can't trust
you.

> > The bottom line is that you run your patched kernel on the real hardware
> 
> Which test configurations would you trust finally?

Do test whatever like the users do.

> > or equivalent (VM or emulation) for the device you touched.
> 
> Can all the devices for which I dared to adjust their source code a bit
> tested in desired ways within virtual machines?

No, you need to run the kernel exactly to be used by the user of the
target hardware.

> > Run your patched kernel and the driver code on the real machine with
> > the corresponding device.  Show the device is running.  That's the
> > very first step.  Then follow the more detailed tests, but it depends
> > on the subsystem.
> 
> How can such descriptions improve the trust situation?

It's the first step.  At least then I can see you did some test.
Currently nothing.  zero.  nada.  How can I trust it?

Takashi