[Bug 8051] Sims 2 demo exits prematurely

[wine-bugs at winehq.org]

https://bugs.winehq.org/show_bug.cgi?id=8051

--- Comment #124 from swswine at gmail.com ---
Created attachment 52904
  --> https://bugs.winehq.org/attachment.cgi?id=52904
Partial workaround to fix game start, models display and get a sensible trace

I eventually tried to run Sims2 and make it work under Wine. It seems to be an
interesting case of using various d3d9 features which are not implemented in
Wine yet. Shader constants limit is just the first but not the least reason. I
hope it is interesting to share what I found as actually previous discussion
does not reveal what is actually wrong with Sims2 under wine.

While I not yet succeded in making Sims2 fully work I found the major reason
why rendering did not work without any indication what's wrong or not supported
in the trace (after the trick similar to that from Comment #118). I am
attaching the patch which partially fixes rendering for me but more important
reveals actually unsupported features which the game is using. This patch by no
means can be thought as a fix proposal for anything, but lets Sims2 run and
draw models, and makes it usable for further testing and fixing.

As noted in Comment #118 here, the crash on startup can be worked around by
increasing the number of supported shader constants. The patch increases that
constant at the first place and also increases number of constants in wined3d.
Actually this workaround helps to run some other games involving 3d characters,
but not this time. After applying just this the game starts but model is
rendered in T pose and totally red (I am mainly running TS2BodyShop.exe to view
how the model renders). At the same time wine trace does not show up any
relevant problems. 

The reason is that Sims2 is using some shader compilaton function from d3dx
library (which has fancy static linkage in sims2 exe files, so no way to use
builtin d3dx.dll). That function requests undocumentated interface from d3d9
through Direct3DShaderValidatorCreate9, and fails if it is not available.
Unlike some redundant constants initialization at startup, this problem does
not show up as crash or something, the game just continues running and falls
back to some most basic shaders which draw red models in T pose, without any
visible errors. My patch adds a very dumb & dirty implementation of the
undocumented validator interface which returns OK for all the requests and does
the trick, the game starts to use real shaders.

After that, if shader model 3.0 is available, the model stops rendering at all
and the trace is full of indications of unsupported features. Not fully
supported ProcessVertices implementation goes first, then a few other things
like tangent, binormal and vertex blend indices vertex attributes, and some
more. These probably should be sorted out one by one. 

The last thing that the patch does is hardcoding maximum shaders version to 2
(caps->VertexShaderVersion and caps->PixelShaderVersion in wined3d/directx.c).
After that the model shows OK in model builder (with normal pose and texture).
I also started the main game in tutorial mode (I do not know how to play it),
and the model also showing OK in the beginning, but other glitches show up
quickly when doing some actions in the house (the glitches are now accompanied
by the informative trace output). So probably it is easier to fix the problems
showing up in shader 2.0 model first as it likely has less missing features. It
actually should be possible to limit shader model version within game
configuration files (the google is full of discussions on how people fixing
Sims2 on Windows through these).

-- 
Do not reply to this email, post in Bugzilla using the
above URL to reply.
You are receiving this mail because:
You are watching all bug changes.