September 2001 - Wine-Devel

dlls/advapi32/crypt.c
by Morten Welinder Sept. 13, 2001

Sept. 13, 2001

Someone added code to CryptGenRandom so that it generates very bad random data (with <10 bits randomness in them). Don't do that. It's a known security risk. Read from /dev/urandom instead. If that does not succeed, just abort(). Linux and newer Solaris have /dev/urandom. I would guess the BSDs have it too. Morten

2 1

wcs* functions within Wine
by Huw D M Davies Sept. 13, 2001

Sept. 13, 2001

Hi, I want to use wcsrchr and wcstrol in oleaut32.dll . If I just use them then we do indeed call the ntdll versions rather than libc which is all very nice. The question is, what include file am I supposed to use to get their protoypes? Huw.

1 0

unmanaged windows
by Bobby Bingham Sept. 13, 2001

Sept. 13, 2001

I think there was a discussion a while back about possibly rewriting the "unmanaged" mode (or adding a new mode, I don't recall which) to make it more like XMMS, but I can't find it now. This would really help on my system, as with KDE 2.2, wine windows are always on top and on every desktop and XMMS fits in a little better. I was wondering what the result if any, of that discussion was. Bobby Bingham -- Microsoft Windows: Made for the internet The Internet: Made for UNIX

1 0

Re: negative listview column width in WinZip 8.0
by Malte Starostik Sept. 12, 2001

Sept. 12, 2001

Am Thursday 13 September 2001 03:37 schrieben Sie: > WinZip 8.0 requests a negative column width for one of the columns in > the WinZip classic Zip file contents list. This stuffs up the list, > causing the other columns to be overwritten, and resulting in a blank > list. > > This patch prevents column widths from being negative. AFAIK, -1 and -2 and maybe others are special values that tell the list view to adjust the contents to the widest item and something else I cannot remember exactly (sorry, no Windows available ATM). -- Malte Starostik PGP: 1024D/D2F3C787 [C138 2121 FAF3 410A 1C2A 27CD 5431 7745 D2F3 C787] http://malte.homeip.net/justice-not-war.html M0ò0ò0

1 0

winedump prints a strange export table
by Francois Gouget Sept. 12, 2001

Sept. 12, 2001

If I run 'winedump dump ws2_32.dll -j export' I get a strange result: ... 00000000 125 <by ordinal> 00000000 126 <by ordinal> 00000000 127 <by ordinal> 00000000 128 <by ordinal> ... Are these really significant? It seems like they go all the way up to the highest ordinal of a 'real' exported function, here 500, WEP. Also if I run 'winedump dump msvcrt.dll -j export -C' winedump seems to crash. It looks like it tries to interpret mangled names as glibc mangled names? Hmm, '-C' seems to have trouble even with other dlls. Christian Jiresjö expressed interest in working on winedump (see bug #91), maybe these would be good bugs for him to takle. Btw eric, thanks for the thread patch in winedbg. -- Francois Gouget fgouget(a)free.fr http://fgouget.free.fr/ Computers are like airconditioners They stop working properly if you open WINDOWS

1 0

HAVE_IPX cleanup
by Francois Gouget Sept. 12, 2001

Sept. 12, 2001

This patch attempts to clean up the ws_sockaddr_ipx to sockaddr_ipx conversions (and vice versa). I would really like if people could try this out: I don't have a working IPX network here. At least my very rudimentary tests indicate that this should still work for the TCP/IP case. More details on the patch: * all ws_sockaddr structures must be considered to have a potentially different layout from their Unix counterparts. Sure for now we just borrow the Unix structs anyway but even sockaddr_in does not have the exact same fields on FreeBSD and Linux (although the binary layout is the same). And hopefully we will not borrow the Unix structs for long anyway. FreeBSD testing is another aspect I would appreciate help with (sin_len worries me). * the ws_sockaddr_ipx fields did not match those used by windows. Also ws_sockaddr_ipx was declared in the wrong header. * all conversions are now done in two functions: ws_sockaddr_ws2u and ws_sockaddr_u2ws. If there is a conversion to do then these are the only two places that should be affected. SUPPORTED_PF returns whether a given protocol family is supported or not. * the sockaddr* pointers may point to structures that are more than 16 bytes, typically 28 bytes for IP v6. I tried to take this into account, I hope I did not mess up (but I'm not sure Wine would support TCP/IP v6 anyway). * I modified WSOCK32_accept to call WSOCK32_getpeername thus avoiding the conversion issue entirely in WSOCK32_accept. * all other functions that were doing such conversions were modified to use the above functions. I also tried to make the error handling code more consistent wrt. WSAENOTSOCK, WSAEAFNOSUPPORT, etc. * removed ugly casts in the traces -- Francois Gouget fgouget(a)free.fr http://fgouget.free.fr/ "Utilisateur" (nom commun) : Mot utilisé par les informaticiens en lieu et place d'"idiot".

1 0

Russian keyboard layout
by Andriy Palamarchuk Sept. 12, 2001

Sept. 12, 2001

Hello everybody, I want to fix message about not found keyboard layout for my configuration. The FIXME message is not shown after I add a phantom key definition to russian keyboard layout main_key_RU in file windows/x11drv/keyboard.c I need your advise how to do it correctly. Should I change existing layout - main_key_RU or it is better to add a separate one with phantom key - main_key_RU_phantom, like it is presently done for US layouts? I could not suppress the FIXME message by changing keyboard model in XF86Config from "pc102" to "pc101". My system: Wine release 20010824, installed without Windows, gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-81) The layout, defined in XF86Config is "ru(winkeys)". Thank you, Andriy Palamarchuk

1 0

Solaris winsock support
by Robert Lunnon Sept. 12, 2001

Sept. 12, 2001

Trying Eudora 4.3 on Solaris solicits this error err:winsock:WsControl Unable to open /proc filesystem to determine number of network interfaces! Looking about in the code WsControl grubs about in the /proc filesystem to get the names of, and count the number of network interfaces. Since /Proc only hosts processes on Solaris this constitutes a linux dependancy we don't need. Studying the ifconfig source gives another possibility instead which look like it may be more portable using ioctl SIOCGIFCONF on an arbitrary IP socket The return from this call is a structure containing the length of the reply buffer, which seems to contains an array of ifreq structures (one for each interface) which contains among other things the name of the interface. By dividing the size of the buffer by the size of the returned objects we know how many interfaces there are... Questions Anyone know if this is Is this workable ? Is this workable on linux, (I ask because I don't have linux and don't know linux ioctl implementation details)?

3 2

X output problem while displaying splash screen
by Andreas Schuldei Sept. 12, 2001

Sept. 12, 2001

The part of the screen where the splash screen of the internet explorer should be displayed freezes, but the debug messages keep scrolling for some time. That is why I send the whole output. I used the command line wine -dll=ole2,ole2conv,ole2disp,ole2nls,ole32,oleaut32,olecli,olecli32,olecnv32,oledlg,olepro32,olesvr,olesvr32,olethk32,compobj,rpcrt4,shell,shell32,shfolder,shdocvw,commdlg,comdlg32,commctrl,comctl32,urlmon,SHLWAPI=n .wine/fake_windows/Program\ Files/Internet\ Explorer/IEXPLORE.EXE --debugmsg +relay,+x11

1 0

Re: Focus livelock
by Alexandre Julliard Sept. 11, 2001

Sept. 11, 2001

Gerard Patel <gerard.patel(a)nerim.net> writes: > Now, when after all this the X events are processed, Wine handles a X > focusin event for A, that is generating a WM_GETFOCUS, then a focusout > for A, generating other WM_KILLFOCUS and WM_ GETFOCUS, > focusin for B, etc..; the sequence of events when creating windows is > perfect, but there are extra events after. > > IMO what's wrong is that X focus events are not ignored when they are the > following of program action. IMO the right fix is to add support for the take focus window manager protocol, and avoid messing with the X focus ourselves. This is somewhere on my todo list... -- Alexandre Julliard julliard(a)winehq.com

2 1

Re: Dialog painting
by Niclas Karlsson MATE Sept. 10, 2001

Sept. 10, 2001

> Wine currently seems to have a problem painting splash dialogs properly > (or at all). I took a closer look at this. I just ripped out the interesting parts from Speak freely in order to find a simple program exhibiting the problems. What I ended up with was a program which does nothing interesting except to show a simple dialg using CreateDialog(). It still doesn't work under Wine. The source is attached. I'm somewhat at a loss about what to do now. I'm just not clear about what should happen and why it doesn't. Some browsing around in the source suggests that it may related to the focus problems Gerard has been talking about, but I'm far from sure. Any comments? Nicke -- -[ nkarlsso(a)abo.fi ]- Seek simplicity, and distrust it. (A. Whitehead)

1 0

Developer offering to help.
by Marco Bizzarri Sept. 10, 2001

Sept. 10, 2001

Hi all. I've seen in the recent past a task list for newbie developer for wine. I would like to know what are still open, how I could contribute. I've very few skills in Windows development. However, I've good knowledge of development under linux, and usage of gnu tools (autoconf). Bye Marco -- Marco Bizzarri - Responsabile Tecnico - Icube S.r.l. Sede: Via Ridolfi 15 - 56124 Pisa (PI), Italia E-mail: m.bizzarri(a)icube.it WWW: www.icube.it Tel: (+39) 050 97 02 07 Fax: (+39) 050 31 36 588

2 1

Re: Focus livelock
by Gerard Patel Sept. 10, 2001

Sept. 10, 2001

At 10:09 AM 09/09/2001 +0300, you wrote: >I posted a patch that would fix a focus related livelock. >However, I have to admit that the patch is quite hacky and likely >not to be included in Wine. Since I would like to get this bug fixed, >I thought to ask opinions about the correct way to fix the bug. > >The short description of the bug is that setting >active/foreground/focus window in Wine calls >X11DRV_SetFocus, which calls XSetInputFocus. >This makes Wine receive FocusIn event, the handler of >which calls SetForegroundWindow. Even though SetForegroundWindow >does nothing if its parameter matches current foreground window, >it can be easily seen that having more than one window can lead >into livelock. I think that the problem with focus is that when a window is shown by Wine, synthetic WM_GETFOCUS and WM_SETFOCUS events are generated, that are themselves generating X focus events. The X events are kept in the X queue until Wine finds time to handle them. Wine creates window A : WM_GETFOCUS for A Wine creates window B : WM_KILLFOCUS for A, WM_GETFOCUS for B Wine creates window C : WM_KILLFOCUS for B, WM_GETFOCUS for C Now, when after all this the X events are processed, Wine handles a X focusin event for A, that is generating a WM_GETFOCUS, then a focusout for A, generating other WM_KILLFOCUS and WM_ GETFOCUS, focusin for B, etc..; the sequence of events when creating windows is perfect, but there are extra events after. IMO what's wrong is that X focus events are not ignored when they are the following of program action. The following hack disables focus handling in this case. I have no clue on its value since I am currently struggling with other bugs. and I will not come to these problems until some time - maybe a lot of time if current Wine progress (aka regressions :-)) continues. --- event.c.orig Fri Aug 24 09:30:46 2001 +++ event.c Mon Sep 10 13:00:28 2001 @@ -534,6 +534,8 @@ if (!hWnd) return; + if (!event->send_event) return; + bIsDisabled = GetWindowLongA( hWnd, GWL_STYLE ) & WS_DISABLED; /* If the window has been disabled and we are in managed mode, HTH Gerard

1 0

Re: Monkey Island 3
by Gerard Patel Sept. 10, 2001

Sept. 10, 2001

At 03:11 PM 09/09/2001 +0200, you wrote: >The installation of Curse of Monkey Island seems to be broken in latest >CVS. All I get is a dialog saying that setup requires a different version >of Windows. > >A trace is available on the web at >http://www.lysator.liu.se/~johane/monkey3.txt > >/Johan Gill, johane(a)lysator.liu.se > I have looked at your trace -the app is trying to use a Win32s function. Maybe the problem is that it succeeds. However, Win32s is present in Wine since years and has not been touched since a long time. I think unlikely that it could be a new problem - when was the last working Wine version before current Cvs ? You can find in documentation/cvs-regression a few tips on how to search for the exact date of a regression. FYI the Wine app database list this game as 'playing perfectly directly off the CD' :-) Gerard

1 0

Re: [PATCH] DestroyWindow16
by Gerard Patel Sept. 10, 2001

Sept. 10, 2001

At 06:40 PM 30/08/2001 +0200, you wrote: >Gerard Patel wrote: >> >> As DestroyWindow does a lot of handle comparaisons, it is necessary >> to give it a real 32 bit handle. This avoids a systematic crash when >> ForteFreeAgent 16 bit exits. I have not looked for other places where >> this can be necessary - but it would not be surprising that this happens. >perhaps as an extension of the relay code ? >(it would mean typing >correctly >windows handles in spec files though) What do you mean by typing ? There is no hwnd type in the spec files, AFAICT. That's not a matter of typing GetWindow(hwnd word) instead of GetWindow(word word) - I think it is necessary to actually change winebuild to define a new type and write a bunch of code to support it. I have looked at the points where it current cvs crashes and it's necessary to fix 16 to 32 bits handles and I have found that 2 changes cure all my new crashes : the one in DestroyWindow16 and another in DefWindowProc16 The reason for these problems is the same : Wine is keeping window handles in several variables (active window, focus window) and Wine code does several comparisons between applications provided handles and these values. In the last case for example, the test /* activate hwndTop if needed. */ if (hwndTop != GetActiveWindow()) { in focus.c fails because of the high word part (and creates an infinite focus loop) I'd say that by adding also handle fixes in SetWindowPos16, SetFocus16 and SetActiveWindow16 as well as fixing the handles setters in queue.c, everything would be covered. IMO your solution is 1) somewhat complex for a fix to apps not widely used anymore 2) overkill because it's not necessary to convert window handles for each and every Api. OTOH your solution is more elegant. I don't know what to do. Gerard

1 0

Dialog painting
by Niclas Karlsson MATE Sept. 10, 2001

Sept. 10, 2001

Hello! Wine currently seems to have a problem painting splash dialogs properly (or at all). Is this a known problem (with a known cause)? An open source program exhibiting problems of this kind would be Speak Freely (http://www.fourmilab.ch/speakfree/windows/) I will try to single out the cause of the problem when I get a little spare time, unless the cause is already known that is. Also, I haven't used Wine for some time so I don't know if this is a long-standing problem. The version I used was a cvs snapshot from last friday before the commits. Cheers, Nicke -- -[ nkarlsso(a)abo.fi ]- Seek simplicity, and distrust it. (A. Whitehead)

1 0

Monkey Island 3
by Johan Gill Sept. 9, 2001

Sept. 9, 2001

The installation of Curse of Monkey Island seems to be broken in latest CVS. All I get is a dialog saying that setup requires a different version of Windows. A trace is available on the web at http://www.lysator.liu.se/~johane/monkey3.txt /Johan Gill, johane(a)lysator.liu.se

1 0

what if MS is forced to publish all API's?
by Scott Bradford Sept. 9, 2001

Sept. 9, 2001

with all the discussion going on in the press about what to do with Microsoft to remove their Monopoly power or at least see to it that they don't abuse their power, how to folks here feel about forcing them to release all their API's? I'm not a programmer myself, but it's my understanding that not having access to the API's is what makes wine such a challenge. If this is the case, would getting access to the MS API's allow Wine to be completed? I feel that a MS breakup wouldn't have done much to change things because they can still give information only to whom they choose, leaving MS to decide who can play and who can't. Having other ways to run the application base of windows software would remove the need to run windows itself, and in my mind, this is the easiest way to remove monopoly power of MS. What do you folks, the Wine developers, think should be done to MS?

1 0

Focus livelock
by Jukka Heinonen Sept. 9, 2001

Sept. 9, 2001

I posted a patch that would fix a focus related livelock. However, I have to admit that the patch is quite hacky and likely not to be included in Wine. Since I would like to get this bug fixed, I thought to ask opinions about the correct way to fix the bug. The short description of the bug is that setting active/foreground/focus window in Wine calls X11DRV_SetFocus, which calls XSetInputFocus. This makes Wine receive FocusIn event, the handler of which calls SetForegroundWindow. Even though SetForegroundWindow does nothing if its parameter matches current foreground window, it can be easily seen that having more than one window can lead into livelock. This happens repeatably every time you start Colonization for Windows. I'm currently thinking about calling directly some WINPOS_* routine from X11 FocusIn handler instead of SetForegroundWindow. If the routine acted exactly like SetForegroundWindow except it made sure that it causes no calls to X11DRV_SetFocus, focus problem would be fixed. Unfortunately I have not figured any good way to pass an extra flag to default handler for WM_ACTIVATE, so that the routine would call something like SetFocus_DontCallDriver instead of regular SetFocus... -- Jukka Heinonen <http://www.iki.fi/jhei/>

1 0

Re: Option to disable XVidMode
by Ove Kaaven Sept. 8, 2001

Sept. 8, 2001

On Fri, 7 Sep 2001, François Gouget wrote: > > I had a report of a crash and it turned out to happen during the > XVidMode initialization. > What happened is that for some unknown reason XF86VidModeGetAllModeLines > returned a single mode with a width and/or height of zero > (htotal/vtotal). This caused a division by zero in convert_modeline. Last time this happened, we considered this a bug in the X server, as under these circumstances, xvidtune also crashes. > Then will come the work of trying to reproduce the problem and see > how best to deal with this server (XFree86-SVGA 4.1.0 apparently). I don't think such a server exists. There's either XFree86 4.x with loadable driver modules, or there's XFree86-SVGA 3.x. The latter sounds more likely to have this kind of bug.

2 2

Re: ukrainian layout
by Dmitry Timoshkov Sept. 8, 2001

Sept. 8, 2001

"Andy Rysin" <arysin(a)yahoo.com> wrote: > Also I had to modify unicode/cp20866.c (KOI8-R) to allow conversion > of additional 4 pair of letters in Ukrainian language (KOI8-U, > RFC2319) as well as one additional pair of letters for Belorussian > language (AKA KOI8-RU, which does not have RFC though). > Strictly speaking KOI8-R is a subset (if not to take to account > pseudographics) of KOI8-U wich is a subset of KOI8-RU. > That's why I applied it to cp20866.c instead of creating smth new. > (I took it from my patches for linux kernel nls system) You can't modify cp*.c files, they are auto generated from the unicode.org file set. You need to send your patch to the unicode.org maintainers and hopefully it will be included in the next Unicode Standard release. > Another problem is that I could not test it practiacally because for > some reason Win32 programs under wine does not input cyrillic > letters, but if I swicth to X program and then back it stops input at > all if layout is cyrillic. If I swicth back to latin everything's ok. It's a known problem, that Wine's x11 keyboard driver does not work with "non-sticky" switches. I have tested cyrillic input with XFree86 3.3.6, 4.0.3 and 4.1.0 with KDE and KiKbd, all works fine for me. Proper support for non-sticky modifiers, dead-keys and the real keyboard layout support requires almost complete keyboard driver rewrite. > BTW, why not to take layouts from X system itself? > Those wich are under /etc/X11/xkb/symbols Because XFree86 is not the only X11 implementation out there. -- Dmitry.

2 2

Re: Make cmdline writable (cmdline4)
by Alexandre Julliard Sept. 8, 2001

Sept. 8, 2001

Francois Gouget <fgouget(a)free.fr> writes: > In the case below. Works on Win98 but not in Wine. I believe the > reason is that Windows does not need to modify the command line anyway. In some cases it does; for instance your test program should fail under NT when using Unicode. Do you have a real app that requires this? -- Alexandre Julliard julliard(a)winehq.com

3 2

WM_[LR]BUTTON{DOWN,DBLCLK} in menus
by Bobby Bingham Sept. 7, 2001

Sept. 7, 2001

I was writing a stupid app to help my friend learn to use menus in windows apps. It would call TrackPopupMenu on a WM_LBUTTONDOWN message. After some experimentation, I found that if I leftclicked in the client are, the menu would appear, as it should. But if, while the menu was still active, I leftclicked in the window's nonclient area (titlebar), the menu would appear elsewhere. By clicking the menu receives a WM_LBUTTONDOWN message, and determines that it should close the menu as it's not inside the menu, and it doesn't remove the message. This is normally fine, but if the click falls on the nonclient area of a window, the message should not be passed through. I believe this also applies to WM_LBUTTONDBLCLK, and the right button equivalents. Is there an easy way to determine what window the click would be passed through to, so we can check if it'll be in the client area or not? Bobby Bingham -- Microsoft Windows: Made for the internet The Internet: Made for UNIX

1 0

Re: Problems with WM_PAINT and threads
by Alexandre Julliard Sept. 7, 2001

Sept. 7, 2001

Ove Kaaven <ovehk(a)ping.uio.no> writes: > After Wine dispatches this WM_PAINT message from the wrong thread, > 3DMark2000 fails to locate its thread-local data, gets confused, and > crashes. Does this help? Index: windows/win.c =================================================================== RCS file: /opt/cvs-commit/wine/windows/win.c,v retrieving revision 1.147 diff -u -r1.147 win.c --- windows/win.c 2001/08/29 00:16:00 1.147 +++ windows/win.c 2001/09/07 19:34:24 @@ -387,15 +387,17 @@ hwndRet = pWnd->hwndSelf; /* look among siblings if we got a transparent window */ - while (pWnd && ((pWnd->dwExStyle & WS_EX_TRANSPARENT) || - !(pWnd->hrgnUpdate || (pWnd->flags & WIN_INTERNAL_PAINT)))) + while (pWnd) { + if (!(pWnd->dwExStyle & WS_EX_TRANSPARENT) && + (pWnd->hrgnUpdate || (pWnd->flags & WIN_INTERNAL_PAINT)) && + GetWindowThreadProcessId( pWnd->hwndSelf, NULL ) == GetCurrentThreadId()) + { + hwndRet = pWnd->hwndSelf; + WIN_ReleaseWndPtr(pWnd); + break; + } WIN_UpdateWndPtr(&pWnd,pWnd->next); - } - if (pWnd) - { - hwndRet = pWnd->hwndSelf; - WIN_ReleaseWndPtr(pWnd); } TRACE("found %04x\n",hwndRet); return hwndRet; -- Alexandre Julliard julliard(a)winehq.com

2 1

Re: Bugs involving returning strings from API functions
by Alexandre Julliard Sept. 7, 2001

Sept. 7, 2001

"Justin Santa Barbara" <justinsb(a)hotmail.com> writes: > As I see it, the advantage of a helper function is that for about the same > amount of work of fixing GetSystemDirectory and GetCurrentDirectory, all the > functions should work the same way meaning any bugs would quickly be found. > It might also be possible to optimize the case where buflen=0, while this > isn't practical per-function. Should I offer a patch only for wine/files/* > ? I'm afraid there are too many variations to make this useful. It would be nice if all APIs that return strings behaved the same way, but unfortunately this is not the case: some count the terminating null, some don't, some copy partial strings when the buffer is too small, some don't copy anything, etc. We would need about as many helpers as there are API functions. Also be aware that the MSDN doc cannot be trusted at all on this matter, the correct behavior needs to be verified for each function. -- Alexandre Julliard julliard(a)winehq.com

1 0