After entering the Coach's mind (as others have reported, there was no sound during the cut scene), the film rolls, but doesn't display the coach's image. Instead I am looking at a blank wall. (It is only because I have played this before that I knew what was supposed to be happening.)
Log from session is attached.
I have this same problem. Coach film and coach doesn't appear.
Running on RV770 [Radeon HD 4850]
Libraries:
mesa 8.0.3 (ati-dri, lib32-ati-dri)
libtxc_dxtn 1.0.1-2
Linux 3.3.7 x86_64
Archlinux, current as of this post.
Applying the psychonuts-linux-test-06032012 patch doesn't help.
This is still happening in 0.6/2013-01-07 on nVidia 313.18 (and everything else I've tried).
I had a bit of a poke around with ApiTrace: the coach is definitely being rendered (to a 512x512 framebuffer), and then bounces around in several FBOs to get downscaled to 256x256 and filtered to give the "film" look. The final texture (1048 in my trace), however, is never bound again (at least until the next frame).
As an aside, while swimming through all of the FBOs, I noticed a couple of things which seem a little bit odd: most notably there are a lot of calls to glActiveTexture(GL_TEXTURE4), when texture unit 4 never has a texture bound.
If you want the trace, I'll send it your way when I can work out how the 'apitrace trim' command works. :)
-- David
Just discovered the 0.8 patch, which fixes this.
It works perfectly on nVidia, but on Intel/Mesa you can see a line of what could be uninitialized memory along the top of the texture as in #5693.
(In reply to comment #16)
> Just discovered the 0.8 patch, which fixes this.
>
> It works perfectly on nVidia, but on Intel/Mesa you can see a line of what
> could be uninitialized memory along the top of the texture as in #5693.
I'm not sure if we officially released the 0.8 patch, but you can go ahead and run it if you like.
(The filmstrip bug is that it used a TEX opcode in the original shader to sample the texture, but the original shader was Shader Model 1.1 in this case, which does something different depending on a rendering state you set in C++. We changed this to be a ps_2_0 shader, with the opcode we actually wanted in the first place--TEXLDP to do a projected sampling--and poof, there was the film.)
--ryan.
I know this bug has been fixed in the Linux Steam version. However, it's still present in the DRM-free version sold on the Humble Store (last update, September 13 2013), as well as the DVD release from The Adventure Company (released in 2014, and as far as I know never updated). Probably in the GOG version as well, but I don't own that one so I can't check.
I tried asking Double Fine about the problem back in March 2015 [1], and on February 9 2018, almost three years later, they finally concluded that "unfortunately we can't address this any time soon".
Is there any chance at all that the non-Steam versions will ever be fixed?
[1] https://forums.doublefine.com/topic/7315-psychonauts-patch-linux-and-mac-update
Created attachment 3153 [details] Log from session