Re: [Qemu-devel] [PATCH] Workaround for broken OSS_GETVERSION on FreeBSD, part two

[malc]

On Sat, 9 Jan 2010, Juergen Lock wrote:

> Turns out on those versions of FreeBSD (>= 7.x) that know OSS_GETVERSION
> the ioctl doesn't actually work yet (except in the Linuxolator), so if
> building on FreeBSD fall back to using SOUND_VERSION as defined in
> <sys/soundcard.h> (which atm is 0x040000) if the ioctl is defined but
> fails.

I've changed the code in the meantime, so this wont apply. Also this
is wrong version is dynamic by nature, if someone runs qemu on FreeBSD
where SOUND_VERSION is less than 4, things will get ugly.

In the new code i mentioned one can opt for not considering POLICY
failure a hard error and proceed with a SETFRAGMENT path, which is
what i would prefer to see on FreeBSD.

> 
>  Signed-off-by: Juergen Lock <address@hidden>
> 
> --- a/audio/ossaudio.c
> +++ b/audio/ossaudio.c
> @@ -289,8 +289,25 @@ static int oss_open (int in, struct oss_
>  
>  #ifdef USE_DSP_POLICY
>      if (ioctl (fd, OSS_GETVERSION, &version)) {
> +#ifdef __FreeBSD__
> +        /*
> +      * Looks like atm (20100109) FreeBSD knows OSS_GETVERSION
> +      * since 7.x, but currently only on the mixer device (or in
> +      * the Linuxolator), and in the native version that part of
> +      * the code is in fact never reached so the ioctl fails anyway.
> +      * But since it just (attempts to) return SOUND_VERSION as
> +      * defined in <sys/soundcard.h> we can fall back to getting it
> +      * from there.  (On FreeBSD, /usr/include is part of the base
> +      * system which is assumed to be always in sync with the kernel.)
> +      * XXX What if the user runs the original OSS from ports instead
> +      * of FreeBSD's own code?  Well I hope we can assume the ioctl
> +      * works there... :)
> +      */
> +        version = SOUND_VERSION;
> +#else
>          oss_logerr2 (errno, typ, "Failed to get OSS version\n");
>          version = 0;
> +#endif
>      }
>  
>      if (conf.debug) {
> 
> 

-- 
mailto:address@hidden