wine-devel

wine-devel@winehq.org

2 participants
84469 discussions

queue management problem
by eric pouech 19 May '01

19 May '01

after last night move of queue management to the server, one of my test app stopped working. What happens is that it's waiting on a WM_TIMER message, but which is no longer triggered. After some investigation, it turned out that message retrieval did change a bit: - in previous code, if the QS_PAINT bit was set, but no window to redraw was found, then the QS_TIMER bit was tested - current code, returns from the server a WM_PAINT message in any case, and if no window can be found, then we retry the whole loop, and we don't check the timer case for my, the wPaintCount was always > 1, meaning that the QS_PAINT bit was never reset then, I investigated why the wPaintCount was > 1. It seems that when a visible window, needing some redraw, is hidden before the update takes place, then we don't update the paint count (which should be decreased) the attached patch does take care of it (basically, if we find a hidden window with an update region, then we destroy the region, and clear the internal paint flag), but I don't think it's the best place to do it. any better idea ? A+ -- --------------- Eric Pouech (http://perso.wanadoo.fr/eric.pouech/) "The future will be better tomorrow", Vice President Dan Quayle Index: windows/win.c =================================================================== RCS file: /usr/share/cvs/cvsroot/wine/wine/windows/win.c,v retrieving revision 1.130 diff -u -r1.130 win.c --- windows/win.c 2001/05/18 22:51:57 1.130 +++ windows/win.c 2001/05/19 11:39:54 @@ -367,8 +367,15 @@ { if (!(pWnd->dwStyle & WS_VISIBLE)) { - TRACE("skipping window %04x\n", - pWnd->hwndSelf ); + TRACE("skipping window %04x\n", pWnd->hwndSelf); + if ((pWnd->flags & WIN_INTERNAL_PAINT) || pWnd->hrgnUpdate) + { + if (pWnd->hrgnUpdate) DeleteObject(pWnd->hrgnUpdate); + pWnd->hrgnUpdate = 0; + pWnd->flags &= ~WIN_INTERNAL_PAINT; + QUEUE_DecPaintCount(hQueue); + } + } else if ((pWnd->hmemTaskQ == hQueue) && (pWnd->hrgnUpdate || (pWnd->flags & WIN_INTERNAL_PAINT)))

1 0

Re: Patch for a bug in server/select.c
by Morten Welinder 19 May '01

19 May '01

> +#ifdef __NetBSD__ > + if (pollfd[i].revents & !(pollfd[i].revents & POLLNVAL)) > +#else > if (pollfd[i].revents) > +#endif I would hate to see this kind of __NetBSD__ pollution of Wine. (I believe something like it slipped through recently -- not good.) It used to be that documentation/porting contained lots of reasons why mentioning any OS name in places like this is a bad idea. I suggest a suitable configure test or a porting wrapper for poll. Morten

2 1

Re: PATCH: PeekMessage - reducing load
by Francois Gouget 18 May '01

18 May '01

On Sat, 19 May 2001, Ove Kaaven wrote: > > On Fri, 18 May 2001, Marcus Meissner wrote: > > > winword.exe apparently uses a while (PeekMessage()) loop. > > > > Both wineserver and 1 wine thread take up 90% CPU even if idle. > > > > This adds a timeout of 10 milliseconds so we are still interactive, > > but do not take up as much CPU. (We basically wait 1 timeslice if 100HZ > > scheduling). > > I question this... an app could be doing something important in such a > loop, i.e. while (!PeekMessage()) do_a_piece_of_important_stuff(), and > this would severely slow down things like that. In particular, I expect a > lot of games to use such loops. If you need CPU time reduction, at least > make it configurable. Me too. There are (badly written) applications that suffer greatly from this. Furthermore I did some tests on Win 95 and Word 97 only does that with some files, not all. Also, Word 2000 no longer behaves that badly. Now, if it were a per application setting it could make sense. -- Francois Gouget fgouget(a)free.fr http://fgouget.free.fr/ Any sufficiently advanced Operating System is indistinguishable from Linux

1 0

Re: Fix in Variant's coerce function
by Alexandre Julliard 18 May '01

18 May '01

Francois Gouget <fgouget(a)free.fr> writes: > So I'm in favor of uncommenting the calls in Coerce, and if these > cases are really being used then either we do a full-fledged > implementation or we write pseudo-stubs for these functions. > > Do we have a know app that crashes with this patch? It won't ever get to crashing, since it doesn't even link! Try a make checklink and you'll see. Unfortunately because of the way we link libraries you can actually create a .so that contains undefined symbols; it will only fail at load-time. This is why we have make checklink to spot this kind of problems at compile-time. -- Alexandre Julliard julliard(a)winehq.com

2 1

Re: Fix in Variant's coerce function
by Alexandre Julliard 18 May '01

18 May '01

Francois Gouget <fgouget(a)free.fr> writes: > It seemed wrong, as did the 32 suffixes, and leaving all the > xxxFromDisp functions commented out. It may be a good idea to actually implement the functions before uncommenting the calls ;-) -- Alexandre Julliard julliard(a)winehq.com

2 1

Re: Question on exposing non-windows functions from dlls...
by Alexandre Julliard 18 May '01

18 May '01

Mike Bond <mbond(a)cox.rr.com> writes: > If you can point me to some way of implementing _exec[lv][ep] using the > standard win32 API I would be more than willing to. Unfortunately that > was one arguable shortcoming of win32 process management in that it did not > have any interface that allowed loading and running an image in-process. You cannot do this with the Win32 API, that's true. And since the native msvcrt is implemented on top of the Win32 API, this means that msvcrt exec functions cannot possibly behave exactly like the Unix ones. What you must do is find out the expected behavior of the msvcrt exec, and implement the same behavior; you should then be able to do this using the Win32 API, since this is how the native one does it. -- Alexandre Julliard julliard(a)winehq.com

1 0

Re: Networking with Mandrake 8.0 and SuSE 7.1
by Steve Fox 18 May '01

18 May '01

Unfortunately I can't really try the 2.4 kernel because Mandrake didn't get the PCMCIA token ring patch in there. I failed miserably to patch it in the last time I tried :( I plan to examine all the patches that Mandrake applies to their 2.2.19 kernel and back out some of them. Steve Fox IBM Linux Technology Center http://www.ibm.com/linux/ltc http://k-lug.com

2 1

Re: Networking with Mandrake 8.0 and SuSE 7.1
by gerard patel 18 May '01

18 May '01

At 10:14 AM 17/05/2001 -0500, you wrote: >I'm running Lotus Notes 5.04a under WINE with one of the January releases >and its been working great for months. However, any release since February >does not seem to do reliable networking on my Mandrake 8.0 system. With the other replies to this post, I'd say that it would be interesting to try to configure your Mandrake system with kernel 2.4 (I suspect that you choosed 2.2 kernel). Gerard

1 0

Re: Networking with Mandrake 8.0 and SuSE 7.1
by Steve Fox 18 May '01

18 May '01

In my case I'm using Mandrake's 2.2.19 kernel, but apparently the straight unpatched 2.2.19 works ok. I guess I'll have to go look through the patches Mandrake applied to the kernel. Thanks for the response. Steve Fox IBM Linux Technology Center http://www.ibm.com/linux/ltc http://k-lug.com

1 0

Re: Networking with Mandrake 8.0 and SuSE 7.1
by r.maurizzi＠gvs.it 18 May '01

18 May '01

> I'm running Lotus Notes 5.04a under WINE with one of the January releases > and its been working great for months. However, any release since February > does not seem to do reliable networking on my Mandrake 8.0 system. (...) > On a related note, another cow-orker who is using Mandrake Cooker > (development distro) and his own compiled kernel does not have this > problem. I am on a token ring network, which I don't think should matter. I had the exact problem with SuSE Linux: starting from Kernel SuSE-2.2.18, wine doesn't work anymore, but if I use a self compiled "stock" kernel, it works again. It's a pity I know little of both Wine and Kernel tcp/ip, and I don't have time to check kernels and diffs... :-( I don't know about Mandrake, but SuSE distributes a tar.gz with all the patches they apply to their kernels. If Mandrake does something similar, I guess a few "apply a part of the patches - compile - test" cycle should help spot the problem(s). If anyone on the list has ideas on tests I could do to gather more info, I'll be glad to help (I still have SuSE stock kernels installed, so I can test differences in behaviours between different versions rather quickly) Ciao, Rob!

2 1

Documentation: lists interfer with the TOC
by Francois Gouget 18 May '01

18 May '01

If you look at the documentation table of contents you'll see things like: Anonymous unions/structs Unicode 21. C library 22. 23. The 21, 22 and 23 are not supposed to be here. What I noticed is that they appear each time I have an orderedlist or an itemizedlist in the section. If I put them in a para instead of directly at the level of the section then I don't get the extra numbers anymore. So should I go through all the doc to make things look like that? <para> <orderedlist> ... </orderedlist> </para> Hmmm, I don't see this problem in the postscript version and DocBook doesn't say that the orderedlist elements must be in a para tag. So it looks like a bug in the HTML generation. Does anyone know how to go fix this? -- Francois Gouget fgouget(a)free.fr http://fgouget.free.fr/ I haven't lost my mind, it's backed up on tape around here somewhere...

2 3

Re: where do I go to find out who needs help
by Andreas Mohr 17 May '01

17 May '01

On Wed, May 09, 2001 at 05:09:20AM -0700, Sean Callahan wrote: > Hi, > I want to do some work for this project who do I mail > Regards, > seanie(a)www.com Well, I'd say that depends on what you're able to do ;-) At the moment I could imagine e.g. the following items: - web server maintenance/update work - improvement of documentation (this badly needs fixing !!) - dlls/shlwapi/reg.c: implementation of user-specific registry stuff (e.g. needed by IE 5) Multiple choice time :-) Andreas Mohr

2 2

Networking with Mandrake 8.0 and SuSE 7.1
by Steve Fox 17 May '01

17 May '01

Repost from wine.users since no reply. Thanks for any help. ------------------------- I'm running Lotus Notes 5.04a under WINE with one of the January releases and its been working great for months. However, any release since February does not seem to do reliable networking on my Mandrake 8.0 system. I can occassionally actually get into my inbox and even open up a message, but I've never been able to read more than one. Some cow-orkers using this same setup on Redhat 7.1 do not have this problem, but others using SuSE 7.1 do too. Does anyone have any ideas why only networking would be affected and only on select distributions? It stumps me because all three mentioned distros use glibc-2.2 and gcc-2.96 (SuSE may use 2.95...not sure). On a related note, another cow-orker who is using Mandrake Cooker (development distro) and his own compiled kernel does not have this problem. I am on a token ring network, which I don't think should matter. -- Steve Fox IBM Linux Technology Center http://oss.software.ibm.com/developerworks/opensource/linux/ http://k-lug.com (Rochester Linux Users Group)

1 0

Re: Patch for a bug in server/select.c
by Alexandre Julliard 17 May '01

17 May '01

Bang Jun-Young <bjy(a)mogua.org> writes: > Suppose that > > ret = poll( pollfd, nb_users, diff ); > > has returned an error bit (either of POLLERR, POLLHUP, or POLLNVAL) > set in pollfd[].revents. Those error codes are defined in positive > numbers in poll.h, so in this case when > > poll_users[i]->ops->poll_event( poll_users[i], pollfd[i].revents ); > > is called, it tries to get an access to an already freed function and > causes a segfault. POLLERR and POLLHUP are legitimate results, and you definitely have to call the function in that case. POLLNVAL is not supposed to happen, unless there is a bug in the fd management. Or does NetBSD set POLLNVAL when the file descriptor is -1? This wouldn't be correct according to the Single Unix spec. -- Alexandre Julliard julliard(a)winehq.com

2 1

DLL separation and PROFILE functions
by Ian Pilcher 16 May '01

16 May '01

DLL separation has been mentioned quite a few times on this list (usually as the ultimate solution to a problem), but I don't think I've ever seen it actually defined. Based on my own misadventures in this area, however, I would hazard a guess that is essentially the conversion of all inter-DLL function calls from OS-based dynamic linking to the Wine DLL loading mechanism. Under this assumption, adding 'IMPORTS = ntdll' to a DLL's Makefile.in file (in order to use one of the PROFILE functions) seems like a definite step backwards. I think that the proper approach is to add the function in question to dlls/ntdll/ntdll.spec and make sure that the importing DLL lists ntdll.dll as an import in its own SPEC file. All of which is wonderful until you consider the following in ntdll.spec: ################## # Wine extensions # # All functions must be prefixed with '__wine_' (for internal functions) # or 'wine_' (for user-visible functions) to avoid namespace conflicts. This would seem to indicate that the PROFILE functions (in files/profile.c) need to have their names changed before they are exported. Is this an appropriate time to go crazy with grep and sed? (For example, PROFILE_EnumWineIniString would become __wine_EnumIniString.) -- ======================================================================== Ian Pilcher ian.pilcher(a)home.com ========================================================================

2 1

Re: Crash in MSG_PeekHardwareMsg
by Alexandre Julliard 14 May '01

14 May '01

Rein Klazes <rklazes(a)xs4all.nl> writes: > Dutch tax 2000 declaration setup program crashes at exactly the same > point. Cause of this seems to be some memory allocation/corruption > problem. At some point the memory used by the sysMsgQueue, is > allocated again by some HeapAlloc() and happily used by its requester. Does this help? Index: windows/queue.c =================================================================== RCS file: /opt/cvs-commit/wine/windows/queue.c,v retrieving revision 1.70 diff -u -r1.70 queue.c --- windows/queue.c 2001/04/04 00:19:55 1.70 +++ windows/queue.c 2001/05/14 18:55:22 @@ -598,7 +598,7 @@ /* Note: We dont need perQ data for the system message queue */ if (!(hmemSysMsgQueue = QUEUE_CreateMsgQueue( FALSE ))) return FALSE; - + FarSetOwner16( hmemSysMsgQueue, 0 ); sysMsgQueue = (MESSAGEQUEUE *) GlobalLock16( hmemSysMsgQueue ); return TRUE; } -- Alexandre Julliard julliard(a)winehq.com

3 2

CalcChildScroll
by Bill Medland 14 May '01

14 May '01

Does anyone have a problem with the following fix to windows/mdi.c/CalcChildScroll? It fixes a problem for one product but I am not in a position to test it thoroughly. When measuring the union of the children rectangles only the visible children should be included: .. } else if( pWnd->dwStyle & WS_VISIBLE) UnionRect( &childRect, &pWnd->rectWindow, &childRect ); This stops scrollbars appearing and disappearing while resizing apparently empty mdi clients etc. Bill Medland

1 0

Crash in MSG_PeekHardwareMsg
by Uwe Bonnes 14 May '01

14 May '01

Hallo, the archive viewer "start.exe" for the CDROM archive of german magazine CT crashes since some time: 085a8b90:Ret USER.114: DISPATCHMESSAGE() retval=0x00000000 ret=02ef:005c ds=033f 085a8b90:Call USER.108: GETMESSAGE(0x033fd1be,0x0000,0x0000,0x0000) ret=02ef:4055 ds=033f fixme:pthread_kill_other_threads_np ... Unhandled exception: page fault on read access to 0xffff0203 in 32-bit code (0x40a1718f). In 32-bit mode. 0x40a1718f (MSG_PeekHardwareMsg+0x4f [message.c:595] in libkeyboard.so): repe movsl (%esi),% es:(%edi) 597 nextqmsg = qmsg->nextMsg; Wine-dbg>bt Backtrace: =>0 0x40a1718f (MSG_PeekHardwareMsg+0x4f(msg=0x40d26d28, hwnd=0x0, first=0x0, last=0x0, remove=0x1) [message.c:595] in libkeyboard.so) (ebp=40d26cbc) 1 0x40a17faf (MSG_PeekMessage+0x30f(type=0x0, msg_out=0x40d26d84, hwnd=0x0, first=0x0, last=0x0, flags=0x1, peek=0x0) [message.c:1187] in libkeyboard.so) (ebp=40d26d44) 2 0x40a185d7 (GetMessage32_16+0x67(msg16_32=0x33fd1be, hWnd=0x0, first=0x0, last=0x0, wHaveParamHigh=0x0) [message.c:1468] in libkeyboard.so) (ebp=40d26da0) 3 0x40a186b0 (GetMessage16+0x30(msg=0x33fd1be, hwnd=0x0, first=0x0, last=0x0) [message.c:1491] in libkeyboard.so) (ebp=40d26dc0) 4 0x409c6e20 (USER_CallFrom16_p_word_lwww+0x30(proc=0x40a18680, args=0x404690a4) [user.spec.c:213] in libkeyboard.so) (ebp=40d26ddc) 5 0x4008d7ae (__wine_call_from_16_word+0x8e [asmrelay.s] in libntdll.so) (ebp=40d26e0c) 6 0x02ef:0x4055 (bp=d1ec) 7 0x02ef:0x4240 (bp=d200) 8 0x02e7:0x5a11 (bp=d26a) 9 0x02e7:0x0063 (bp=0000) Any ideas? Uwe Bonnes bon(a)elektron.ikp.physik.tu-darmstadt.de Free Software: If you contribute nothing, expect nothing --

2 1

Re: GDB remote debugging
by Ove Kaaven 14 May '01

14 May '01

On Mon, 14 May 2001 =?iso-8859-1?Q?G=E9rard_Patel_=3Cgerard.patel=40asi.fr=3E?=(a)s5.asi.fr wrote: > At 11:12 PM 09/05/2001 +0200, Ove Kaaven wrote: > >For any sufficiently disillusioned Wine hackers, here's my current > >experimental code for letting GDB attach to a Wine process. > <snip> > > Is there any way to debug an application from the beginning using this tool ? > (the reason I would like to try it is of course that the app crashes while starting) > If not, is it thinkable to change the code to wait for gdb to connect when > the initial Wine thread connects to the server ? I put that variable "create_suspended" in gdbremote.c for that purpose... just change it from 0 to 1... ought to be a command-line option, but that patch won't be included in wine in its current form anyway, so...

1 0

Re: Problem with tooltips
by gerard patel 14 May '01

14 May '01

At 09:56 PM 13/05/2001 +0200, you wrote: <snip tooltip file dialog problem> >Does anyone know how to fix this? http://www.winehq.com/hypermail/wine-devel/2001/05/0035.html (or replace winehq by 130.225.13.222 if the Corel Dns is still down :-/) Gerard

1 0

Re: Crash in MSG_PeekHardwareMsg
by gerard patel 14 May '01

14 May '01

At 08:56 PM 12/05/2001 +0200, you wrote: >085a8b90:Ret USER.114: DISPATCHMESSAGE() retval=0x00000000 ret=02ef:005c ds=033f >085a8b90:Call USER.108: GETMESSAGE(0x033fd1be,0x0000,0x0000,0x0000) ret=02ef:4055 ds=033f >fixme:pthread_kill_other_threads_np >... >Unhandled exception: page fault on read access to 0xffff0203 > in 32-bit code (0x40a1718f). >In 32-bit mode. >0x40a1718f (MSG_PeekHardwareMsg+0x4f [message.c:595] in libkeyboard.so): repe movsl (%esi),% >es:(%edi) >597 nextqmsg = qmsg->nextMsg; <snip> >Any ideas? There is not much to go on in this information. I can't see this problem with any of my 16 bits apps, but maybe none of them use this kind of GetMessage without specifying a window. I'd suggest first that you compile message.c with optimization level O0 : it seems clear that the displayed instruction is not the one responsible for the crash; the crash happens really in *msg = qmsg->msg. Maybe having no optimization could give more accurate information. When you have done that, rerun and enter info reg and info local when the crash occurs. Also a trace with +relay could be interesting but I think that you know that since you post this tip on cemw often enough :-) Gerard

1 0

Could someone check this on windoze
by Uhmmmm 13 May '01

13 May '01

Could someone please doublecheck this. On windoze, does the bitmap in a SS_BITMAP style STATIC control change if you send it WM_SETTEXT or call SetWindowText() on it like it does in WINE? I don't think it does, but I'd like it if someone could check before I start changing things.

2 2

Problem with tooltips
by Marcus Meissner 13 May '01

13 May '01

Hi, Start any program that uses the common dialog file open dialog (Win95 style) using -managed. Then move the mouse over any of the toolbar icons. Watch a VERY HUGE window manager window pop up displaying the tooltip text in the upper left corner. Reason for this appears to be some changed flags that lets WIN_WindowNeedsWMBorder fail on tooltips. Does anyone know how to fix this? Adding WS_EX_TOOLWINDOW to the tooltip CreateWindow is probably wrong :/ Ciao, Marcus

1 0

RE: WINEPS: Read metrics from TrueType fonts
by Nerijus Baliunas 13 May '01

13 May '01

> This patch enables the PostScript driver to read font metrics directly > from TrueType font files with FreeType 2. It looks for font files in > directories listed in the [TrueType Font Directories] section of the > wine config file. For example: > > [TrueType Font Directories] > "dir0" = "/usr/X11R6/lib/X11/fonts/TrueType" > "dir1" = "/usr/share/fonts/default/TrueType" > etc. > > It does not embed the fonts into its output. Ghostscript users can > install their TrueType fonts in Ghostscript (with their Fontmap file or > the GS_FONTPATH environment variable); users printing to PostScript > printers will have to convert their fonts to Type 1 or Type 42 format > and use an external program to add them to the print jobs. Hello, Why don't you add this text to documentation/printing? Regards, Nerijus

1 0

Wine logo
by Ian Pilcher 12 May '01

12 May '01

Inspiration for the graphically (and accurately) inclined: http://www.sutterhome.com/interactive/in1034.html -- ======================================================================== Ian Pilcher ian.pilcher(a)home.com ========================================================================

1 0

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

wine-devel