Support speaker configuration panel and variable number of channels

[amurzeau]

This PR is rather large but still.

These are the features implemented:

• Fix static analysis warnings reported by VS 2017
• Enable POOL_NX_OPTIN as reported by the driver verifier on Windows 10 (this has an effect only on Windows 8+)
• Add a topology filter. This complicate things a bit but should be better as existing audio devices always have both a wave filter and topology filter (both real hardware or virtual audio cables)
• Remove the unused clockRegister (replaced by a reserved field)
• Support the "Configure" button in the audio devices control panel
  • This allow to have the benefit of Windows mixing taking care of different channels. Like playing 4.0 audio on a 5.1 configured endpoint, this will output the rear channels on the matching rear channel of the 5.1 setup instead of center and LFE channels.
  • This is implemented by changing the endpoint category to SPEAKER from LINE_CONNECTOR
• Allow to playback with less channels than the endpoint has. Like playing 2.0 stereo on a 6 channels endpoint.
  • This is implemented by adding a register containing the actual number of channels on the WaveRT side, and filling only the enabled channel in SarAsio (silencing others)
• Fix crashes and various bugs when hammering endpoint creating and destruction
• Support Int16, Int24 and Int32 on SarAsio side. Each sample type is represented by its size (2, 3 or 4 bytes) on WaveRT side.
• Add more logging categories. I actually added somewhat verbose logs but I don't think it is a good idea to remove them all. These logs could be useful in case of a failure encountered by an user. Same for the Irp dump, I just kept it non compiled inside #ifdef, this can be useful.
• I had to put some KS WAVERT properties in #if because they were not defined when targeting Windows 7. Like KSPROPERTY_RTAUDIO_PRESENTATION_POSITION. You were targeting Windows 10 before and it was still ok on Windows 7 ? The disabled properties are return STATUS_NOT_IMPLEMENTED so this should not have any effect.

I'm myself using my modified driver, but with windows in test mode as I don't have any certificate. I've tested it mostly (ie. including stress tests) in Windows 10 1511 and 1803, but also on Windows 8.1 and windows 7.

[pannal]

@amurzeau would you share your builds so we can test your changes? I'm more than willing to enable testsigning for that :)

[amurzeau]

I've added some commits to makes building on appveyor possible (with an upgrade to VS 2017 as I only have that one). Binaries will be available here: https://ci.appveyor.com/project/amurzeau/synchronousaudiorouter/build/artifacts

They are not signed, so to be able to run the driver, you will need to enable testsigning boot flag: https://docs.microsoft.com/en-us/windows-hardware/drivers/install/the-testsigning-boot-configuration-option

Thank's for your testing :) I'm using that driver myself now with jack so I can do audio routing between channels and DSP filters.

[amurzeau]

Also, ensure you uninstalled the previous SAR driver installation, else Windows might prefer to use the existing driver as it was signed. The driver version should be 23.48.50.371.

If the version is wrong, go in the Synchronous Audio Router device properties (in the Device Manager) and do an "update driver". Then in the dialog box:

• "Browse my computer for driver software"
• "Let me pick from a list of available drivers on my computer"
• Then choose the right driver version in the list.

[pannal]

I've contacted you on the SAR Discord to not further pollute this PR.

[amurzeau]

I've been running this on Windows 8.1 x64 since that PR without any issues while using hibernation, merging now.