May 2006 - Wine-Devel - list.winehq.org

re: thanks for wine
by Dan Kegel May 16, 2006

May 16, 2006

>> Copy-paste didn't work between wine and Gnome, but everything else was >> without a hitch! > > Doh! What exactly failed? Could you give me the exact steps to reproduce > the problem? Yes, please do! Let's get a bug filed about this. There are quite a few copy and paste bugs in Wine, e.g. http://bugs.winehq.org/show_bug.cgi?id=633 http://bugs.winehq.org/show_bug.cgi?id=2382 http://bugs.winehq.org/show_bug.cgi?id=3486 http://bugs.winehq.org/show_bug.cgi?id=4523 http://bugs.winehq.org/show_bug.cgi?id=5061 I wonder if just getting copy and paste working well would be a sufficient summer of code project! Too bad submissions are closed. Ah, well, next year :-) - Dan -- Wine for Windows ISVs: http://kegel.com/wine/isv

8 15

Re: 4979: wine 9.11 make fails on AMD64, more info
by David Anderson May 16, 2006

May 16, 2006

I just installed kubuntu 6.06 on a new laptop (no binary avail), AMD64. Compiling wine failed as 4979 says, no libfreetype.so that is compatible. The problem is not in the -L or in gcc or in wine. The problem is in the m32 libfreetype. /usr/lib32 is searched (as /usr/lib/../lib32) by gcc automatically with -m32. The libfreetype developer(s) provided libfreetype.so.6.3.8 and link libfreetype.so.6 ->libfreetype.so.6.3.8 However they forgot to create the symlink libfreetype.so -> libfreetype.so.6 Create this by hand in /usr/lib32 and regrettably wine still will not build as 4 externs are left unresolved: inflate, inflateReset,inflateEnd, inflateInit2. This appears to also be a libfreetype issue. It's been reported several times in various contexts, google for 'libfreetype inflate' to see them. I don't have a solution, sorry.

2 2

bugzilla report changes
by Saulius Krasuckas May 16, 2006

May 16, 2006

I just have submited an additional attachment on my report page [*]. After this bugzilla said to me: Changes Submitted -- Attachment #2466 to Bug #2082 Created Email sent to: lionel.ulmer(a)free.fr, jave27(a)gmail.com, z77y16(a)hotmail.com, danielv(a)freeshell.org, ivg2(a)cornell.edu, pq(a)iki.fi, chuzzum(a)gmail.com, flerchjj(a)ieee.org Excluding: saulius.krasuckas(a)elst.vtu.lt Well, I don't understand, why this hasn't been sent to wine-bugs(a)winehq.org. Did that happen by a design? [*] http://bugs.winehq.org/show_bug.cgi?id=2082#c23

2 1

Re: [wined3d] Corrected definition for parameters (Coverity)
by Ivan Gyurdiev May 16, 2006

May 16, 2006

Paul Vriens wrote: > Hi, > > don't know anything about wined3d, but the code in > IWineD3DVertexShaderImpl_ExecuteSW could make use of 6 parameters. The > definitions should cover this. > I don't think so... the case for 6 parameters should be removed. I can't find 6-parameter instructions in the ins table. [otherwise this would have to be fixed in a lot of other places as well]. ==== I find the multi-argument calls rather ugly, though I understand they might be necessary for speed. However, I am currently trying to get relative addressing tokens in shaders 2.0+ properly accounted for. We can have up to... 5 extra tokens per instruction for relative addressing. So, either all those multi-arg calls need to switch to unpacking structures, or the number of arguments should be increased up to 10. I think I'll ignore software shaders for now though - not sure how to test them.

2 2

OGL Crash
by Robert Lunnon May 16, 2006

May 16, 2006

I have a problem with OGL , during initialisation I have a segfault occurring in the following code segment, note that I have instrumented the code XVisualInfo *X11DRV_setup_opengl_visual( Display *display ) { XVisualInfo *visual = NULL; /* In order to support OpenGL or D3D, we require a double-buffered visual and stencil buffer support, */ fprintf(stderr,"setup visual\n"); int dblBuf[] = {GLX_RGBA,GLX_DEPTH_SIZE, 16, GLX_STENCIL_SIZE, 8, GLX_DOUBLEBUFFER, None}; if (!has_opengl()) return NULL; fprintf(stderr,"Choosing Visual display=%p DefaultScreen=%p dblBuf=%p\n",display, DefaultScreen(display), dblBuf); wine_tsx11_lock(); ************************Crash Occurs in the following line visual = pglXChooseVisual(display, DefaultScreen(display), dblBuf); wine_tsx11_unlock(); fprintf(stderr,"Visual is %d\n",visual); Output is setup visual Choosing Visual display=81ad320 DefaultScreen=0 dblBuf=7fd1ef70 Handling exception at EIP 7f0434b7 accessing 2a0 wine: Unhandled page fault on read access to 0x000002a0 at address 7f0434b7 (thread 0009), starting debugger... Can't attach process 8: error 5 I suspect the problem is DefaultScreen returning 0 Anyone know where this is defined ?

1 0

Re: rpcrt4: Implemented NTLM authentication for rpcrt4 connections.
by Kai Blin May 16, 2006

May 16, 2006

* Mike McCormack <mike(a)codeweavers.com> [15/05/06, 23:04:25]: > + r = AcquireCredentialsHandleA(NULL, "NTLM", SECPKG_CRED_OUTBOUND, NULL, > + AuthIdentity, NULL, NULL, &bind->cred, &bind->exp); Sweet. Glad to see that code in use. :) > + /* tack on a negotiate packet */ > + RPCRT4_AuthNegotiate(Connection, &out); > + r = RPCRT4_SendAuth(Connection, Header, Buffer, BufferLength, out.pvBuffer, out.cbBuffer); > + HeapFree(GetProcessHeap(), 0, out.pvBuffer); Should that actually be a Negotiate packet? If it should, I can add Negotiate to the set of patches I sent Rob. Cheers, Kai -- Kai Blin, (blin at gmx dot net) Your wig steers the gig. -- Lord Buckley

2 2

Re: user32: Add more ShowWindow tests
by Alexandre Julliard May 16, 2006

May 16, 2006

Dmitry Timoshkov <dmitry(a)codeweavers.com> writes: > @@ -3217,7 +3295,7 @@ static void test_showwindow(void) > > trace("calling ShowWindow( SW_SHOWMAXIMIZE ) for invisible popup window\n"); > ShowWindow(hwnd, SW_SHOWMAXIMIZED); > - ok_sequence(WmShowMaxPopupResizedSeq, "ShowWindow(SW_SHOWMAXIMIZED):popup", FALSE); > + ok_sequence(WmShowMaxPopupResizedSeq, "ShowWindow(SW_SHOWMAXIMIZED):popup", TRUE); It would be better to avoid turning working tests back into todos, because it means a lot of things will no longer be properly regression tested, since todo applies to the whole sequence. -- Alexandre Julliard julliard(a)winehq.org

2 3

Help needed: Drawing of transformed vertices
by Stefan Dösinger May 16, 2006

May 16, 2006

Hello, As I have mentioned on irc some time ago, I have some Z order problems in a few d3d7 games. It seems to affect all games which mix transformed and untransformed vertices in one scene. The problem is that the transformed vertices show up in front of the untransformed, although the shouldn't. See http://stud4.tuwien.ac.at/~e0526822/messiah-broken.png, this is a screenshot from the game Messiah, and the symptoms are that the flames and the beams can be seen through boxes, units and other objects. I can remember that a simmilar problem was reported for a d3d8 or d3d9 game. Other affected games are Gothic 2(fire can be seen though walls) and Anarchy Online(water level is too high). With Lionel Ulmer I found the problem in the glOrtho setup that wined3d does in drawprim.c(and in surface.c). The -minZ, -maxZ z range scales the vertices and disturbes the z order. The old ddraw code uses a transform matrix which is equally to passing zmin = 1.0 and zmax = -1.0 to address this issue. Then the z coordinate are passed to gl unchanged and the z order is left intact. But as you undoubtately can guess, this breaks other games such as Prince of Persia 3D. Some investigation showed that the current z setup in wined3d seems to be to avoid clipping. First I thought that rhw vertices aren't supposed to be clipped at all, but opengl does that and it's impossible to deactivate clipping in gl. So the vertices are scaled into the depth range to prevent them from beeing clipped. Lionel Ulmer mentioned a DONOTCLIP flag for d3d7 vertex drawing, which pop3d indeed passes. But we can't avoid clipping and disturbing the z setup at the same time :-( The attached patch moves the glOrtho setup in a shared function for drawprim and unlockrect to avoid having double code. It also changes the default z range to 1.0, -1.0, with a dontclip flag which selects the old setup(but this is disabled for now, it will be activated for the DONOTCLIP flag and if ProcessVertices was supposed to clip the vertices). Can you test the patch on some d3d8 and d3d9 games? I tested it on warcraft 3 and half-life 2 and didn't see any difference. Can you report negative and positive changes? Thanks, Stefan

2 1

SoC idea: fix wine 2398 bug.
by Marcin Kardas May 15, 2006

May 15, 2006

Hi, I'm participate in Google summer of code and I'm going to send application with my own idea. I would like to fix popular wine bug - http://bugs.winehq.org/show_bug.cgi?id=2398. On http://wiki.winehq.org/OpenGL there are some ideas how to do this and I think that this one with glViewport and glScissor is the best one. I've made some test and it should work, but may be it will cause another problems to solve, for example conecting X graphic context with GLContext. Maybe I will have to rewritte wine graphics function to use only OpenGL (or may be I will found another way). What do you think about it? Cheers, Marcin.

8 11

ClipCursor implementation
by Remus May 15, 2006

May 15, 2006

Hello, Please excuse me if i didn't sent this mail to the right mailing list. I want to implement in Linux a function similar with ClipCursor, but without using XGrabPointer(). The first time i tried to process mouse-move events, but it doesn't look good. Second time i tried XGrabPointer(), it works, but an extra window is not wanted in my app. I think this functionality is already implemented in Wine, but i didn't find much about it in the sources, except for setting a rectangle, etc. Do you have any ideas or directions to point to me? Thanks, Remus

1 0