[pulseaudio-discuss] [PATCH] alsa: Silence some Valgrind warnings

[Tanu Kaskinen]

On Mon, 2014-02-10 at 22:25 +0100, David Henningsson wrote:
> On 02/07/2014 03:53 PM, Tanu Kaskinen wrote:
> > On Thu, 2014-01-30 at 23:00 +0100, David Henningsson wrote:
> >> On 01/29/2014 07:58 PM, Tanu Kaskinen wrote:
> >>> I don't know if there's some bug in alsa-lib or not, but reading dB
> >>> values seems to irritate Valgrind a lot. These changes remove all
> >>> warnings that can be removed within our own code, but reading dB
> >>> values still causes warnings within alsa-lib code.
> >>
> >> I tried to fix this in Valgrind two years ago, but got stalled:
> >>
> >> http://valgrind.10908.n7.nabble.com/Missing-ioctl-for-SNDRV-CTL-IOCTL-TLV-READ-td42711.html
> > 
> > Interesting. So, is my patch OK if I add a reference to the valgrind
> > mailing list thread?
> > 
> 
> Well, fixing the root cause (in valgrind) would be better, but as I'm
> pragmatic, I'm okay with working around it until the root cause is fixed.
> 
> Btw, maybe something like this would be a more readable solution:
> 
> #ifdef HAVE_VALGRIND_MEMCHECK_H
>  /* Work around valgrind bug, see
> http://valgrind.10908.n7.nabble.com/Missing-ioctl-for-SNDRV-CTL-IOCTL-TLV-READ-td42711.html
> */
> 
> #define snd_mixer_selem_ask_playback_vol_dB(a, b, c) \
>  VALGRIND_MAKE_MEM_DEFINED(c, sizeof(*c)); \
>  snd_mixer_selem_ask_playback_vol_dB(a, b, c);
> /* More macros here */
> #endif
> 
> ...or inline functions if that makes more sense.

That's a good idea. I'll post a new version of the patch (unless it gets
buried in the todo list forever).

-- 
Tanu