1 files changed, 97 insertions, 0 deletions
diff --git a/results/classifier/118/device/1882787 b/results/classifier/118/device/1882787
new file mode 100644
index 000000000..3f087ccb3
--- /dev/null
+++ b/results/classifier/118/device/1882787
@@ -0,0 +1,97 @@
+device: 0.937
+socket: 0.908
+graphic: 0.904
+hypervisor: 0.881
+register: 0.874
+files: 0.872
+PID: 0.864
+peripherals: 0.842
+ppc: 0.841
+risc-v: 0.841
+performance: 0.824
+network: 0.821
+architecture: 0.818
+arm: 0.818
+assembly: 0.806
+debug: 0.778
+semantic: 0.771
+boot: 0.765
+permissions: 0.760
+vnc: 0.755
+TCG: 0.747
+kernel: 0.727
+user-level: 0.704
+VMM: 0.703
+virtual: 0.679
+x86: 0.658
+i386: 0.635
+mistranslation: 0.632
+KVM: 0.580
+
+AUD_set_volume_out takes SWVoiceOut as parameter, but controls HWVoiceOut
+
+There was a change in https://github.com/qemu/qemu/commit/571a8c522e0095239598347ac0add93337c1e0bf#diff-230ab01fa7fb1668a1e9183241115cb0R1852-R1853 (audio/audio.c) which breaks audio output on devices which have multiple software voices on the same hardware voice.
+
+When multiple software voices use the same hardware voice, then setting a volume / mute for any of the software voices, will affect all other software voices, too.
+
+I'm not sure if such a use-case exists in QEMU; however, it does exist in my fork.
+It's also easy to see that this is a bug in the QEMU audio subsystem.
+
+The API (and broken function) for this is:
+
+```
+    void AUD_set_volume_out (SWVoiceOut *sw, int mute, uint8_t lvol, uint8_t rvol)
+```
+
+So this is supposed to modify the SWVoiceOut.
+
+However, if the backend supports `pcm_ops->volume_out` this does not work as expected. It's always as if you had called:
+
+```
+    void AUD_set_volume_out (HWVoiceOut *hw, int mute, uint8_t lvol, uint8_t rvol)
+```
+
+*(Note how this modifies the hardware voice)*
+
+
+In my specific use case, I have 2 outputs (digital and analog audio on AC97), and I want to mute the digital audio output, but I still need to keep the voice activated for timing.
+However, if I mute the digital audio SWVoiceOut, it will also affect the other SWVoiceOut (for analog audio) on the same HWVoiceOut.
+
+---
+
+Old code - if the hardware supports volume changes, it will receive the software voice which should be modified, so changes can be restricted to that one voice:
+
+```
+        HWVoiceOut *hw = sw->hw;
+        [...]
+        if (hw->pcm_ops->ctl_out) {
+            hw->pcm_ops->ctl_out (hw, VOICE_VOLUME, sw);
+        }
+```
+
+New code - the hardware backend will have no way to differentiate software voices; so any change will affect all voices. The volume which was set last on any sw voice will set a global volume / mute for all voices on the hardware backend:
+
+```
+        HWVoiceOut *hw = sw->hw;
+        [...]
+        if (hw->pcm_ops->volume_out) {
+            hw->pcm_ops->volume_out(hw, &sw->vol);
+        }
+```
+
+The old interface was already broken with some (all?) backends, because they ignored the software voice, but at least the design made sense.
+However, the new design is fundamentally broken because it doesn't even tell the backend which voice is supposed to be modified.
+
+---
+
+Bug was introduced in cecc1e79bf9ad9a0e2d3ce513d4f71792a0985f6
+The affected code was touched since then, but still remains in 49ee11555262a256afec592dfed7c5902d5eefd2
+
+
+This is an automated cleanup. This bug report has been moved to QEMU's
+new bug tracker on gitlab.com and thus gets marked as 'expired' now.
+Please continue with the discussion here:
+
+ https://gitlab.com/qemu-project/qemu/-/issues/74
+
+