Emulator

The errors and bad_ps_0000 files are different between the backends. Not a whole lot but, might have some new information, so I should include it. The ones above are for D3D11, so here's OpenGL and D3D9. OpenGL appears to have the most information (actually talks about an error in it).

OpenGL
Error - http://img823.imageshack.us/img823/6541/opengl.png
bad_ps_0000.txt - http://pastebin.com/8aaZRVh5

D3D9
Error - http://img542.imageshack.us/img542/3576/d3d9.png
bad_ps_0000.txt - http://pastebin.com/yk5VmsWD

there are problems with FPS in this game and the sound studders. It goes up to 120fps :P I play the PC version now.

Using the 32 bit version of Dolphin(3.5) and it still crashes at the same spot.

Vastly different hardware as well (6950 2gb, Intel 2500k)

This bug happened to a PAL user on the forum. - https://forums.dolphin-emu.org/Thread-gc-beyond-good-evil--26920?pid=285708#pid285708

It appears it happens regardless of version. How some people were able to get past this point to reach Dark Island (read wiki test reports) remains a mystery.

I'm happy to report that the game can now pass the 'Hilly's screen and work fine with both the OpenGL and D3D11 backends. However only the OpenGL backend runs the game at 100% speed pass the Hilly's area. Just to be clear the D3D11 backend is 100% speed but runs extreamly slowl;y after the Hilly's screen.

please see images:
OpenGL: http://imgur.com/I7baBqm,wuxeDTQ,dfKKuFd,VeczPF9
D3D11: http://imgur.com/I7baBqm,wuxeDTQ,dfKKuFd,VeczPF9#3

Settings used: http://imgur.com/I7baBqm,wuxeDTQ,dfKKuFd,VeczPF9#1
http://imgur.com/I7baBqm,wuxeDTQ,dfKKuFd,VeczPF9#2

Here it still happens:
http://www.mediafire.com/view/e4picgt45agyt5g/bad_ps_0000(2).txt
If you disable panic handles though the game will continue fine.

Can you record a fifo log over the place where the shader compile error shows up? I guess it should be able to pass that scene by disabling panic handlers (if it's reproducible enough).

You can record over many seconds and just re-record the fifo log on playback if that makes things any easier.

Here's a fifolog. As kosta said, you can get past it now (I couldn't when I first made this report), so I recorded it as I went through it. You can now see the errors yourself in this handy fifolog!

http://www.mediafire.com/download/tzedi36ts7746ze/beyondgoodandevilfifolog.7z

I did a comparison of the area where the errors appear.

Dolphin - http://i.imgur.com/Z2KSOn8.jpg
Console - http://i.imgur.com/eGefIpT.jpg

Compared in 4.0-1516 win64. I knew something was fishy about the water. As you can see... no reflections in Dolphin. Apparently that's what the errors are fussing about.

More testing. 4.0-1535 win64 this time.

D3D slows WAY down on the ocean. In 4.0-1516 it would speed up randomly in certain places, like when being hauled away by Mamagos. But 4.0-1535 doesn't do that, it's always slow throughout. Weird. Of course, reflections are not present the entire time, and internal resolution as no effect on the speed. And there are very slight visual difference between the backends when the camera angles down though. Oh, and black isle leads to the slowdown getting so bad the game hangs. Weird. OpenGL has fullspeed in those same areas and doesn't hang.

Also, it should be noted that in D3D you can't skip past the bad_PS error warnings by just clicking (they respawn infinitely), but in OpenGL you can just click through them. Disabling panic handlers works fine in both backends.

Interpreter still doesn't boot. However, I used the savestate trick to get around it. The game hangs when going into the transition where the bad_ps errors show up, and sits there forever. Loading a savestate in interpreter out on the open water does not make the reflections appear. JITIL behaves identically to JIT.

Real XFB doesn't fix the reflections. EFB to Ram, per-pixel lighting, on and on, all have no effect.

Software shows the reflections... sort of. Has to be in single core to get around the bobbing ship thing (see above). - http://i.imgur.com/ey2CY8s.jpg

Interesting, something that looks like reflections appears at low angles, such as when you go into the first person view on the boat (Y). You can also see it on Black Isle. Both of these shots were taken in OpenGL.

First Person Boat - http://i.imgur.com/EY5oUFb.jpg
Black Isle - http://i.imgur.com/LOjcbXI.jpg

And here's a fifolog of Black Isle that shows the reflections sort of working - http://www.mediafire.com/download/r1d0ctx3hh9u0pv/Beyondgoodandevilfifolog2.7z

3.5-663 (last working) - https://dolphin-emu.org/download/dev/8767b30f75db5732023d9ebf9d93573f3b839e50/

3.5-1025 (GLSL merger, broke it) - https://dolphin-emu.org/download/dev/e1a081ad2dee7127e6c1fe2827eb1eb59ca24c6e/

The very old OGL backend didn't changed the shader to the broken one, the new one and the D3D backend just doesn't render anything.
I've hacked in the old behavoir and the rendering seems fixed, so I think it is because of the ERROR9 msg in the shaders.

IMO it's wrose to continue rendering with the wrong shader, so I don't like this hack.

MaJoR confirmed that this issue is also valid on Interpreter.

This should be fixed as of 4.0-1577, if I understand the issue correctly. I don't have the game, so I can't verify it myself, though.

I just gave it a try in 4.0-1761. It...."technically" did render the reflection (no shader compile issue appeared, with panic handlers enabled), but the entire scene was completely janky, slow and unplayable. The water would flicker, all geometry would disappear randomly, rendering would clip out of bounds, etc.

http://i.imgur.com/nyWHdqk.jpg

Ignore my last comment. I was dumb and had dumb things like frameskipping turned on. Here it is on 4.0-1761:

D3D: http://i.imgur.com/Xvb3LbU.jpg
OGL: http://i.imgur.com/j92k1h4.jpg

The scene runs about 50-60% of the speed of normal gameplay, but it does seem to be rendering correctly.

What are the specs of your PC, logan?

Tested and confirmed that the water is working.

" degasus: feel like giving any sort of context of why you marked that issue as fixed?"

This issue was no regression, it just has worked by accident. The older GL code did continue with the last successful compiled shader, the current code (both d3d and ogl) skip the drawing with broken rendering. So the issue was all the time the "ERROR9" message of the shader compilation (why are panicAlerts non-fatal atm?)

magumagu did some HW tests to verify the undocumentated behavior of this ERROR9 issue, and merged his changes in 4.0-1577, which fixes the issue. It was tested by myself iirc + loganstromberg in June 1st (see his correct screenshots in #22), but it seems I haven't resolved this issue report.