for this bug: http://bugs.winehq.org/show_bug.cgi?id=3091 [ HL2 D3D ]

you only have to copy MSVCR70.dll from a windows installation (isn't a memory problem)

Well, I've seen HL2 work before without this dll... just using Oliver's patch. Steam only seems to need it when displaying Update News (crashes).

Anyway, with msvcr70.dll from a winxp machine, it gets a little further, and crashes again.

fixme:dbghelp:sffip_cb NIY on 'U:\dev\src_main\launcher_main\Release \launcher_main.pdb' fixme:dbghelp:sffip_cb NIY on 'C:\relHL2\src\launcher\Release \launcher.pdb' fixme:dbghelp:sffip_cb NIY on 'C:\relHL2\src\vstdlib\Release \vstdlib.pdb' fixme:dbghelp:sffip_cb NIY on 'C:\Valve\HL2\Build\src\tier0\Release \tier0.pdb' fixme:dbghelp:sffip_cb NIY on 'C:\Valve\HL2\Build\src\filesystem \filesystem_stdio\Release_STEAM\filesystem_steam.pdb' fixme:dbghelp:sffip_cb NIY on 'u:\p4clients\rel_beta\Projects \GazelleProto\Client\Engine\VC70_Release_Static \SteamEngine.pdb'fixme:dbghelp:sffip_cb NIY on 'C:\Valve\HL2\Build\src \materialsystem\Release\materialsystem.pdb' fixme:dbghelp:sffip_cb NIY on 'C:\relHL2\src\vguimatsurface\Release \vguimatsurface.pdb' fixme:dbghelp:sffip_cb NIY on 'C:\Valve\HL2\Build\src\engine\GLRelease \engine.pdb' fixme:dbghelp:sffip_cb NIY on 'C:\Valve\HL2\Build\src\vgui2\src\Release \vgui2.pdb' fixme:dbghelp:sffip_cb NIY on 'U:\dev\src_main\soundemittersystem \Release\soundemittersystem.pdb' fixme:dbghelp:sffip_cb NIY on 'C:\Valve\HL2\Build\src\materialsystem \shaderdx8\Release\shaderapidx9.pdb' fixme:dbghelp:sffip_cb NIY on 'u:\p4clients\taylor_valve_cser \ThirdPartyCode\DebugNet\Release\BugslayerUtil.pdb' fixme:dbghelp:sffip_cb NIY on 'msvcr70.pdb' wine: Call from 0x60004545 to unimplemented function dbghelp.dll.StackWalk64, aborting

On 06.07.2005 09:02, Raphael wrote:

Why it need StackWalk64 ?

StackWalk64 is used on both 32 and 64 bit Windows; for a while now, MS says to use StackWalk64, probably because it's more portable or so.

-f.r.

Steam and Half-Life 2 have a utility that dumps debugging information when either crashes and this is what uses msvcr70 and the dbghelp functions. Turn on +seh logging and you'll likely see a STATUS_ACCESS_VIOLATION before this.

With the newest patches on d3d, this is no longer the problem - the problem is that HL2 runs out of memory on startup and the OOM killer takes it out. So, it does seem like there is a memory management problem after all..

Hi Ivan, Yesterday I read your posts on the list saying that you're able to get HL2 to run with some patches. That's great--I can't wait to see it actually be playable. :) Can you send me the patches you're using so that I can try them?

Thanks, James Liggett

On Sun, 2005-07-10 at 14:57 -0400, Ivan Gyurdiev wrote:

...

Steam and Half-Life 2 have a utility that dumps debugging information when either crashes and this is what uses msvcr70 and the dbghelp functions. Turn on +seh logging and you'll likely see a STATUS_ACCESS_VIOLATION before this.

With the newest patches on d3d, this is no longer the problem - the problem is that HL2 runs out of memory on startup and the OOM killer takes it out. So, it does seem like there is a memory management problem after all..

On Thu, 2005-07-14 at 17:29 -0400, Ivan Gyurdiev wrote:

On Thu, 2005-07-14 at 13:54 -0700, James Liggett wrote:

...

Hi Ivan, Yesterday I read your posts on the list saying that you're able to get HL2 to run with some patches.

The only patch that I'm using outside wine-cvs is here: http://www.winehq.org/hypermail/wine-patches/2005/03/0504.html

It's necessary to let you type the Steam password and download HL2 over the web (since MSI still doesn't work).

...

That's great--I can't wait to see it actually be playable. :)

You'd have to wait a bit more, currently it's way too slow (I really mean this...). I'll post some screenshots on appdb when it becomes playable with cvs.

Have you tried setting the heapsize to something a bit larger? Sometimes that can help.

You might want to try: directxwine.sf.net

I'll see if there's anything new there, update my CVS, and give it a go... Right now I've got HL2 on my account already; I tried it and it hangs, but anyway, thanks a lot. I'm sure the wait will be worth it in the long run.

On Sat, 2005-07-16 at 16:34 -0700, James Liggett wrote:

Wow, you're not kidding... ;-) I did have some better luck with HL2 deathmatch and CS:Source though. I can create servers and actually try to play, but currently the mouse movement is way too sluggish. If you give it a second or two to compensate for the slowness, weapon firing animations are pretty fluid, but only for a few seconds. You can even walk around somewhat. Hey, it's a start :) I can give you some screenshots of the multiplayer stuff if you like.

I think the problem is Vertex Shader support (or lack of it).. in the code I see several versions of the drawprim code, and I seem to end up in the slowest path (DrawStridedSlow)

Screenshots? Sure, if you want to - send in a private mail. I'll update appdb if this reflects CVS, rather than Oliver's tree - which did you test?

Thanks for the tip. I have noticed that the initial load on any of the Source games isn't that bad on my machine (about 10 seconds.) Are you actually able to play any of these games?

James

On Sun, 2005-07-17 at 11:01 +0200, Andreas Schneider wrote:

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

To make it easier entering HL2 menu, you can disable the animated background and kill the start video.

Here is my hl2 tweaking guide:

http://www.linux-gamers.net/modules/wfsection/article.php?articleid=60

I have to update it after my exams. Suggestions are welcome :)

The howto section will be wiki like soon (I hope end of july).

-- andreas

James Liggett wrote:

...

Wow, you're not kidding... ;-) I did have some better luck with HL2 deathmatch and CS:Source though. I can create servers and actually try to play, but currently the mouse movement is way too sluggish. If you give it a second or two to compensate for the slowness, weapon firing animations are pretty fluid, but only for a few seconds. You can even walk around somewhat. Hey, it's a start :) I can give you some screenshots of the multiplayer stuff if you like.

James

On Thu, 2005-07-14 at 17:29 -0400, Ivan Gyurdiev wrote:

...

On Thu, 2005-07-14 at 13:54 -0700, James Liggett wrote:

...

Hi Ivan, Yesterday I read your posts on the list saying that you're able to get HL2 to run with some patches.

The only patch that I'm using outside wine-cvs is here: http://www.winehq.org/hypermail/wine-patches/2005/03/0504.html

It's necessary to let you type the Steam password and download HL2 over the web (since MSI still doesn't work).

...

That's great--I can't wait to see it actually be playable. :)

You'd have to wait a bit more, currently it's way too slow (I really mean this...). I'll post some screenshots on appdb when it becomes playable with cvs.

You might want to try: directxwine.sf.net

---

http://www.linux-gamers.net - your online gaming resource

-----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (GNU/Linux)

iD8DBQFC2h5kf94+j/M+P8YRAgTJAJ9fgPUJ0jMvjjJMIWq5wc7/Uk8nxwCfUHSV eWZIiNLQto9mRlAWzyy2uqM= =+3eP -----END PGP SIGNATURE-----

I have single player HL2 running decent. Not quite excellent, but playable. Try running steam with this command:

WINEDEBUG="-all" wine steam.exe

And you should get anywhere from 10-30 FPS. It's not the best in the world, but at least you can play. :) I have started a new game, and I am now in the beginning of the car sequence, and aside from some textures being lit up and the flashlight not working, it's pretty nice. But, beware of problems with pixel shaders. The make CS and HL2 deathmatch totally useless (they give a white screen in menu. It's not frozen, but you can't see anything.) I don't see that behavior with single player because of the animated menus. Plus, pixel shaders screw things up big time anyway. It inverts colors and makes it look like NPCs are holding tree trunks. I'll put up some shots up later so you can see what I mean.

James

On Sat, 2005-10-29 at 17:48 -0500, Evil wrote:

One thing I've noticed while trying to get CS:S to work is that if I set the resolution to something other than that of my desktop (1280x1024), WINE does resize the screen to the appropriate rez... BUT moving my mouse to the edge of the screen lets me scroll across the rest of my virtual desktop.

Maybe this is because I'm using xOrg 6.8.2? (I'm running Mandriva 2005LE)

I would suspect this is part of the reason Counter-Strike gives me 1frame per second: If OpenGL works like DirectX, maybe it's not actually writing directly to video RAM and is instead copying from a backbuffer in system RAM? I'm not familiar with OGL at all, so I don't know if that's true or not.

Things seem better when I run the game in my native resolution in that at least I don't see my desktop newsticker crawling along the top. However, performance is relatively unchanged (i.e. about half a frame a second or less at the higher rez)

I'm using an ATI 9600XT card with the latest (8.18) drivers. Native Linux 3D apps run very nicely.

-J

On Sun, 2005-10-30 at 11:34 -0600, Evil wrote:

Thanks for the tip, James! Disabling all the debugging had a much bigger impact than I would have expected. I've included my observations and comments below, in hope that they might be of use to the devs working on DX.

Yeah, you wouldn't think turning off debugging messages would do much of anything but it does. Don't ask me why though... :)

The Counter-Strike: Source Video Stress Test is now giving me 15.74fps in 1024x768 windowed mode (vs. 24 in Cedega and 44 in Win98). It's a definite step in the right direction, but still a little too low to prevent you from getting shot by any newb. :) I get the same speed at lower resolutions, which indicates that I'm CPU-limited. I can believe that, since it's pegged-out the entire time I run the test.

Hmmm... I can usually get double that. My card isn't all that different from yours (I have a Radeon 9800 pro) Personally I think it has more to do with the speed of the code and yet to be implemented features. I can imagine that since Oliver is still trying to stabilize things and implement stuff, speed isn't yet the priority.

I get pretty much the same performance whether I am windowing the game by setting wine to emulate a desktop, or just passing -window to the hl2.exe (15.2fps).

The odd thing to me is: I get the same speed when running in fullscreen mode. This is even the case if I used XRandR to set my desktop to the same resolution before starting Steam, so that I don't end up accidentally scrolling across my virtual desktop when I move the mouse down or to the right. Doing that is also a help because it stops my newsticker from drawing over the top of the game screen, but it seems to have no impact on performance.

Do OpenGL applications typically perform the same in windowed mode as fullscreen mode? My experience programming with DirectX made me expect fullscreen to run faster since the video card only has to flip the backbuffer pointer instead of blitting from one chunk of memory to another. Of course, that all being handled by the video card, I suppose this too is probably just a side-effect of my CPU-limited speed.

I wouldn't think running a game in a window versus full-screen would change much of anything. Then again, I'm no 3D expert either.

Here are the parameters I'm feeding to CS:S by the way, in case anyone's interested: "-heapsize 512000 -fullscreen -width 1024 -height 768". I've tried passing "-dxlevel 70" to the hl2.exe, but don't see much if any performance difference. It is interesting to not that if you don't pass this parameter, you will see little white boxes indicating each light source. These boxes aren't clipped by any object.

Passing "-dxlevel 80" causes the menu screen to come up completely white, making it unuseable. Forcing DirectX 8.1 or 9.0 has the same white-screen effect.

Try disabling pixel shaders in winecfg to get around that. They never did work for me. I tried to use them yesterday with HL2 to get some screenshots, and they froze my computer. If you're lucky enough to even star a game using them, they mess things up anyway.

Restarting CS:S without exiting Steam completely can have flaky results. It might be related to the amount of time it actually takes the game to cleanup and exit. I've noticed that right after exiting the game and then exiting Steam, it can take several minutes or longer (of maxed out CPU utilization) before a command prompt returns.

Maybe the Valve minidump program is intercepting a crash. I see that a lot with Source multiplayer games like CS:S and HL2:DM. I believe that it might be because of the web browser control used to display MOTDs. BTW, what did you do to get transgaming's mozctl working? Thanks.

Visually, I would say it looks better than I recall getting from Cedega. The textures on the monoliths were always plain and flat when I was running Cedega. I'm tempted to renew my Cedega subscription to make sure I'm remembering right.

Yeah, I know what you mean. I've never been a fan of Cedega, so I say to hell with it. I would stick with free WINE. ;-)

Anyway, it's looking very promising. Keep up the great work!

I'm not really involved with development on the DX emulation. I just follow Oliver's work, test, cross my fingers, and hope it works. :)

-Jesse

James Liggett wrote:

...

I have single player HL2 running decent. Not quite excellent, but playable. Try running steam with this command:

WINEDEBUG="-all" wine steam.exe

And you should get anywhere from 10-30 FPS. It's not the best in the world, but at least you can play. :) I have started a new game, and I am now in the beginning of the car sequence, and aside from some textures being lit up and the flashlight not working, it's pretty nice. But, beware of problems with pixel shaders. The make CS and HL2 deathmatch totally useless (they give a white screen in menu. It's not frozen, but you can't see anything.) I don't see that behavior with single player because of the animated menus. Plus, pixel shaders screw things up big time anyway. It inverts colors and makes it look like NPCs are holding tree trunks. I'll put up some shots up later so you can see what I mean.

James

That should have said "regsvr32 mozctlx.dll", BTW. That's what I get for writing from memory. :)

Evil wrote:

James Liggett wrote:

...

BTW, what did you do to get transgaming's mozctl working? Thanks.

All you need to do is go to the control directory (~/.transgaming_global/mozcontrol, by default) and run "regsvr mozctlx.dll" to register the DLL. Many thanks to Fabian Bieler, who gave me that info.

Since I think I've gotten CS:S and HL2 working as good as they can with the current code, I think I'll see if I can get anywhere with Anarchy Online (I hate the game, but would like to see WINE running it. ;) ).

-Jesse

Ivan Gyurdiev wrote:

I see that the patch from Oliver which implements pixel shaders hasn't been merged yet, which might explain why they aren't working...

This probably explains my results with Anarchy Online tonight too, if I don't misunderstand about shaders.

I got the AO patcher working, and the main game too after some tweaking (basically, I had to force some native DLLs back to the built-in ones after playing with Sidenet yesterday). It even runs at a respectable speed, except there are a lot of textures coming out completely black.

I see the following errors repeating while it's running:

err:ddraw:set_render_state Specular Lighting not supported yet. err:ddraw:GL_IDirect3DDeviceImpl_7_SetLight dvTheta not fully supported yet ! err:ddraw:GL_IDirect3DDeviceImpl_7_SetLight dvTheta not fully supported yet !

There didn't appear to be any existing entries in the bug database for AO, so I added this info there. I'll re-test as I see patches coming into CVS.

-Jesse