[FernetMenta]

I recommend this for the next merge window. If there are no reasonable objections, we should do the switch.

[da-anda]

I noticed issues with stereo sound over HDMI to my TV - all it outputs is dissorted with some additional noise. This is only when using linux or WASAPI on windows - using directshow on windows music and gui sounds play fine via HDMI. Using SPDIF I didn't have any issues so far. Passthrou is working fine as well via HDMI and SPDIF from my first tests (AC3, DTS - avr is old and doesn't support HD formats).
Haven't found the time to debug the mentioned issue in detail - once I can confrim it's ActiveAE related (usually not using HDMI audio) I'll create debug logs

[Voyager]

I just discovered an issue with sync playback to display. When choosing audio clock you get heavy stuttering, e.g. When playing back 23.976 fps on a 24hz display, I got something around 5 fps... Note my audio is 5.1 digital over SPDIF. This was working fine with SoftAE.

Now switching to videoclock dupe/drop solved that issue for me.

[FernetMenta]

could you post a log please.

EDIT: I can't reproduce this here.

[WiSo]

Tested with my monitor speakers which are connected via hdmi. mp3 (stereo) and other formats worked fine with wasapi and directsound.

[da-anda]

Just did a little test session and compared SoftAE with ActiveAE while playing music both times. I noticed a ~10% higher CPU usage on ActiveAE (debug log enabled). Cycling output devices took a lot longer on ActiveAE and even maxed out the CPU for a short period of time, while in SoftAE everything ran smooth and on switches CPU max reached 60%. ActiveAE even crashed once while switching output devices while playing sound (see debug log below).

I tried to reproduce the borked WASAPI sound via HDMI and failed at first (sound was fine to my surprise), but after a second round of switching output devices back and forth I finally got the borked sound again.

Here are the according debug logs
- borked sound (the last output device switch in the log before exiting xbmc) http://xbmclogs.com/show.php?id=44259
- crash http://xbmclogs.com/show.php?id=44257

[fritsch]

From the log:

13:15:34 T:3764 INFO: CActiveAE::OpenSink - sink incompatible, re-starting
13:15:34 T:3764 ERROR: CAESinkWASAPI::InitializeExclusive: IsFormatSupported failed (AUDCLNT_E_UNSUPPORTED_FORMAT) - trying to find a compatible format
13:15:34 T:3764 ERROR: CAESinkWASAPI::InitializeExclusive: Unable to locate a supported output format for the device. Check the speaker settings in the control panel.
13:15:34 T:3764 INFO: CAESinkWASAPI::Initialize: Could not Initialize Exclusive with that format
13:15:34 T:3764 ERROR: CAESinkWASAPI::Initialize: WASAPI initialization failed.
13:15:34 T:3980 NOTICE: Thread AESinkNull start, auto delete: false

Btw. did you receive a real "segfault" or did it just hang there without reacting?

[da-anda]

it just hung

what I forgot to mention before - "output stereo to all speakers" didn't work for me at all. All I got was silence, even if turned off again, regardless of directsound or wasapi. Only a restart of XBMC brought sound back

[WiSo]

cycling through the devices works pretty fast on my setup and without hang even during music playback (under the debugger).

[FernetMenta]

Cycling output devices took a lot longer on ActiveAE

ActiveAE does a lot more in this case than SoftAE. It drains the sink instead just pulling the rag under your feet like SoftAE. So if you change the device while playing audio, ActiveAE has to wait until the buffer is empty, then it can change the device. As I mentioned in the other thread, we'll get rid of the spinner.

@da-anda can you elaborate on what you did to make it crash?

[sialivi]

I've only done some quick tests so far (with Windows 7 / WASAPI:HDMI), but I've already noticed that ActiveAE seems to have fixed an issue I've been having with popping and crackling in videos with aac, mp3 and a few other audio codecs. Good work!

[FernetMenta]

CAESinkWASAPI::EnumerateDevicesEx needs rework, it thinks it supports AE_FMT_S16NEP

[WiSo]

The code in CAESinkWASAPI line 794 to 813 checks if the device supports the format and if yes it adds it to the list. Does make ActiveAE use of deviceInfo.m_dataFormats? afaik SoftAE didn't make any use of it and the values recorded in EnumerateDevicesEx are just to log it into the logfile.

[FernetMenta]

CAESinkWASAPI::InitializeExclusive(AEAudioFormat &format) may have produced unpredictable results in case AE requested a planar format. I have created a pr for this.

[WiSo]

why? As far as I can see it tests what it supports? And with the changes of AEDataFormat we won't see in the log anymore if a planar format is supported by the hardware or not.