audio: Don't call SDL_AudioStream callbacks for empty data sets.
Fixes #8095.
diff --git a/src/audio/SDL_audiocvt.c b/src/audio/SDL_audiocvt.c
index 3df1465..68eae68 100644
--- a/src/audio/SDL_audiocvt.c
+++ b/src/audio/SDL_audiocvt.c
@@ -751,7 +751,10 @@
stream->flushed = SDL_FALSE;
if (stream->put_callback) {
- stream->put_callback(stream, SDL_GetAudioStreamAvailable(stream) - prev_available, stream->put_callback_userdata);
+ const int newavail = SDL_GetAudioStreamAvailable(stream) - prev_available;
+ if (newavail > 0) { // don't call the callback if we can't actually offer new data (still filling future buffer, only added 1 frame but downsampling needs more to produce new sound, etc).
+ stream->put_callback(stream, newavail, stream->put_callback_userdata);
+ }
}
SDL_UnlockMutex(stream->lock);
@@ -1047,7 +1050,9 @@
approx_request *= stream->src_sample_frame_size; // convert sample frames to bytes.
const int already_have = SDL_GetAudioStreamAvailable(stream);
approx_request -= SDL_min(approx_request, already_have); // we definitely have this much output already packed in.
- stream->get_callback(stream, approx_request, stream->get_callback_userdata);
+ if (approx_request > 0) { // don't call the callback if we can satisfy this request with existing data.
+ stream->get_callback(stream, approx_request, stream->get_callback_userdata);
+ }
}
// we convert in chunks, so we don't end up allocating a massive work buffer, etc.
