Emulator

I got a crash on exit also, from selecting exit from the Dolphin menu.

When debugging with Visual Studio 2010 the popup says:
Unhandled exception at 0x000007fefdb49670 in Dolphin.exe: 0xC0000005: Access violation writing location 0x0000000000200fe8.

and it shows me a green arrow pointed at line 132 of MemTools.cpp which reads:
return EXCEPTION_CONTINUE_SEARCH;

Yeah ... smart pointers solve all our memory leaks, right...
Especially since D3D11 had that many.[/cynicism]

Fwiw, not sure what's going on. Guess it's a regression caused by replacing all COM refcounting by smart pointers in the D3D11 code.

This problem starts from Dolphin version 7421 and onward.

Issue 4416 has been merged into this issue.

Nice info SAM, i didnt know in which rev problem appeared exactly.

I would deem this a 3.0 blocker personally - ugly regression that will affect a lot of users.

billiard!!11

Does this happen for other games as well?

Happens on all games in my experience. Upon starting a game, stopping it and exiting emu, I get the error message about 90% of the time.

I've noticed, If I play any game after playing one of those games (Disney/Pixar presents Cars or Pokemon Battle Revolution, some more..) and then exit, Dolphine.exe stops working.

I did some investigating, and it seems to be something with CMPR format textures. I've tested with PoP:T2T, which uses CMPR textures right away, and the crash always occurs with it, while Soul Calibur II doesn't use them until you see a character model and won't cause the crash if closed prior to the Select Character screen.

Correction: Soul Calibur II doesn't use the CMPR Textures until a match starts.

I think you have right. Dolphin doesent crash if you interrupt SC2 before a fight, then exit dolphin.

First interesting note:
I believe the video memory corruption isn't restricted in Dolphin, as it can corrupt the GFX memory of other 3D applications such as TrackMania United Forever.

I don't have SSBB or TMUF, but I've played many games more than once in a single session using DX11. I haven’t experienced any graphic corruption in Dolphin or in a PC game after using Dolphin.
I'm using ATI HD5670.

Fwiw, I'm already working on reverting r7421.

Although you're free to do it yourself as well (since I'm short of time these days) :P

fwiw, you should probably remove the whole SharedPtr stuff (rather than only parts of it), it was stupid to begin with anyway IMO.

I can't get dx11 to crash on stop here. HD3650 - DX 10.1.

@tommy
That is because they think it only effects NVIDIA.

Anyone with the problem, please check r7592.

I can confirm it too.

For the record, the problem looked like it was due to some erroneous reference counting done by SharedPtr on some resources (such as ID3D11DepthStencilState, ID3D11BlendState, and ID3D11RasterizerState, as reported by the DX11 debug libraries during SetPrivateData() calls).

Now what if the "video memory corruption" was caused by some buggy API's (NVIDIA's?) free'ing data based on obsoleted pointers? But whatever :p

It most likely was caused by that, and it would have been trivial to find out by just checking the output of the D3D11 debug layer (which would've even told us WHICH device object didn't get released properly).
But well. Like I said, I didn't feel like wasting my time debugging the issue myself.

Dude, the Radeon HD 5670 is ATI, not NVIDIA!

I am not exactly sure what you mean.

Do you mean, by chance, thay my old "NVIDIA-only" guess for the crashes on stop was wrong because I somehow overlooked the "/GPU HD5670" part of the very first post? I was making that guess mainly for the video memory corruption and/or display driver crashes, because I didn't see many people with such (odd) problems.

Just so you know, by "NVIDIA", I was referring to my own GPU (NVIDIA GeForce 8800 GTS 512), as well as to my old "perhaps NVIDIA-only" guess (which I made up because I didn't see many people getting crashes in "~SharedPtr()", let alone video memory corruption and/or display driver crashes...).