[FEAT] OSS sound support on NetBSD

[nbvins]

Description

The native NetBSD sound subsystem - audio(4) - is based on the SunAudio API derived from OpenSolaris.
However, a compat library providing OSS emulation - ossaudio(3) - is available for commodity on any NetBSD system.

While a native implementation may be preferred in the future[1], the following patch allows building sound_bsd.c on NetBSD. For this to work, fastfetch needs to be linked with -lossaudio. The default sound unit is detected through the command line using audiocfg(1).

--- src/detection/sound/sound_bsd.c.orig	2025-01-13 07:57:52.000000000 +0000
+++ src/detection/sound/sound_bsd.c
@@ -5,13 +5,39 @@
 #include <fcntl.h>
 #include <sys/soundcard.h>
 
+/* Obsolete macros */
+#ifndef SOUND_MIXER_MUTE	// unavailable on some platforms
+#define SOUND_MIXER_MUTE	SOUND_MIXER_NONE
+#define SOUND_MIXER_READ_MUTE	MIXER_READ(SOUND_MIXER_MUTE)
+#endif
+
 const char* ffDetectSound(FFlist* devices)
 {
     char path[] = "/dev/mixer0";
+
+#if defined(__FreeBSD) || defined(__DragonFly__)
     int defaultDev = ffSysctlGetInt("hw.snd.default_unit", -1);
 
     if (defaultDev == -1)
         return "sysctl(hw.snd.default_unit) failed";
+#elif defined(__NetBSD__)
+    const char* const cmd = "audiocfg list | grep [*] | cut -d: -f1";
+    char buf[32];
+    long defaultDev = -1;
+
+    FILE* f = popen(cmd, "r");
+    if (f == NULL)
+        return "popen() failed";
+
+    while (fgets(buf, sizeof buf, f) != NULL) {
+	defaultDev = strtol(buf, NULL, 10);
+        if (defaultDev == -1)
+	    return "audiocfg: failed to get default sound unit";
+
+    if (pclose(f) != 0)
+        return "pclose() failed";
+#endif
+    }
 
     for (int idev = 0; idev <= 9; ++idev)
     { i
@@ -26,7 +52,7 @@ const char* ffDetectSound(FFlist* device
             continue;
 
         uint32_t mutemask = 0;
-        ioctl(fd, SOUND_MIXER_READ_MUTE, &mutemask); // doesn't seem to be available on DragonFly
+        ioctl(fd, SOUND_MIXER_READ_MUTE, &mutemask);
 
         struct oss_card_info ci = { .card = idev };
         if (ioctl(fd, SNDCTL_CARDINFO, &ci) < 0)

[1] TODO; aiomixer(1) sources may represent a good starting point.

Motivation

To avoid an additional dependency on pulseaudio on NetBSD.

Additional context

Using the patch above, on a NetBSD 10.1_STABLE system running on a Thinkpad T460s, I get:

~ $ fastfetch -s sound --format json
[
  {
    "type": "Sound",
    "result": [
      {
        "active": true,
        "main": true,
        "volume": 68,
        "name": "Realtek ALC293 01h",
        "identifier": "/dev/mixer0",
        "platformApi": "OSS"
      },
      {
        "active": true,
        "main": false,
        "volume": 0,
        "name": "Intel HDMI/DP 01h",
        "identifier": "/dev/mixer1",
        "platformApi": "OSS"
      }
    ]
  }
]

If the default sound unit is changed to the onboard intel device:

~ $ audiocfg default 1
setting default audio device to audio1
~ $ audiocfg list
0: [ ] audio0 @ hdafg0: Realtek ALC293
       playback: 2ch, 48000Hz
       record:   2ch, 48000Hz
       (PR) slinear_le 16/16, 2ch, { 32000, 44100, 48000, 88200, 96000, 192000 }
       (PR) slinear_le 20/32, 2ch, { 32000, 44100, 48000, 88200, 96000, 192000 }
       (PR) slinear_le 24/32, 2ch, { 32000, 44100, 48000, 88200, 96000, 192000 }
       (  ) ac3 16/16, 2ch, { 32000, 44100, 48000, 88200, 96000, 192000 }
1: [*] audio1 @ hdafg1: Intel HDMI/DP
       playback: 2ch, 48000Hz
       record:   2ch, 48000Hz
       (P-) slinear_le 16/16, 2ch, { 48000 }
       (P-) slinear_le 16/16, 4ch, { 48000 }
       (P-) slinear_le 16/16, 6ch, { 48000 }
       (P-) slinear_le 16/16, 8ch, { 48000 }
       (PR) slinear_le 16/16, 2ch, 48000-48000Hz

~ $ fastfetch -s sound --format json
[
  {
    "type": "Sound",
    "result": [
      {
        "active": true,
        "main": false,
        "volume": 68,
        "name": "Realtek ALC293 01h",
        "identifier": "/dev/mixer0",
        "platformApi": "OSS"
      },
      {
        "active": true,
        "main": true,
        "volume": 0,
        "name": "Intel HDMI/DP 01h",
        "identifier": "/dev/mixer1",
        "platformApi": "OSS"
      }
    ]
  }
]

[CarterLi]

Please test the dev build

[nbvins]

Thank you for the extremely quick reply and commit. Using readlink() is indeed better than spawning a subprocess. One thing, the array size for mixerp, as well as the character count returned by readlink should be larger (to store /dev/mixerN), otherwise the call will fail.

--- src/detection/sound/sound_bsd.c.orig	2025-01-26 13:09:45.000000000 +0000
+++ src/detection/sound/sound_bsd.c
@@ -15,10 +15,10 @@ const char* ffDetectSound(FFlist* device
     #else
     int defaultDev;
     {
-        char mixerp[8];
-        if (readlink("/dev/mixer", mixerp, ARRAY_SIZE(mixerp)) != 6)
+        char mixerp[12];
+        if (readlink("/dev/mixer", mixerp, ARRAY_SIZE(mixerp)) != 11)
             return "readlink(/dev/mixer) failed";
-        defaultDev = mixerp[5] - '0';
+        defaultDev = mixerp[10] - '0';
         if (defaultDev < 0 || defaultDev > 9)
             return "Invalid mixer device";
     }

[CarterLi]

One thing, the array size for mixerp, as well as the character count returned by readlink should be larger (to store /dev/mixerN), otherwise the call will fail.

Strange. As I tested in VM, readlink reported mixer0 instead of /dev/mixer0

[CarterLi]

Anyway 4e9fecc should work for both cases. Please test it again

[CarterLi]

BTW, are you vins?

[nbvins]

Strange. As I tested in VM, readlink reported mixer0 instead of /dev/mixer0

I think readlink(2) should just store the contents of the symlink's path in the buffer (so it's the full path), while readlinkat() stores the path relative to a directory.

If you do:
printf("'%s' points to '%.*s'\n", "/dev/mixer", (int) plen, mixerp);

it returns: '/dev/mixer' points to '/dev/mixer0'

Anyway 4e9fecc should work for both cases. Please test it again

It works, thank you.

BTW, are you vins?

Yes, I will update the patch to include your changes while updating the package to 2.35.0.
EDIT: done [2]

[2] https://mail-index.netbsd.org/pkgsrc-changes/2025/01/26/msg315346.html