- Wine-Devel - list.winehq.org

Re: Small fixes in controls/menu.c
by gerard patel May 20, 2001

May 20, 2001

At 03:35 AM 18/05/2001 +0200, you wrote: >Application Super Memo 8 has two modes of functioning. One mulit-document >(using MDI) and the other single-document. <snip> >Second Problem: >Also occurs at switching between modes. >The menu is displayed at improper place (about 10-20 pixels upper than it >should). >I think the problem is in the function MENU_InitTracking in code : <snip one of my patches> My patch is certainly not positioning the menu correctly in the case of your app, and your patch fixes the app I did this patch for all right, but I am wondering if calling SetWindowPos could not generate infinite loops in some cases. Is Super Memo 8 available on the web as a demo app (freely downloable) ? Also vs your ignored patch, I don't know - only Alexandre Julliard is judge and executioner of all that is posted to wine-patches - but in such cases check always if your patch applies; if it does not, either because you did a mistake or the cvs code was changed too much between the creation of your sandbox and the moment he looked at your patch, it could be the reason. Gerard

2 1

Desktop mode
by lawson_whitney＠juno.com May 19, 2001

May 19, 2001

Good day! In case anyone is bored and wants a little bug, Dmitry's little About.exe program will answer neither the keyboard nor the mouse in desktop mode. In any other mode it works fine. Sorry, I didn't notice when it happened, but it has been a while. Lawson ---cut here ________________________________________________________________ GET INTERNET ACCESS FROM JUNO! Juno offers FREE or PREMIUM Internet access for less! Join Juno today! For your FREE software, visit: http://dl.www.juno.com/get/tagj.

1 0

Re: Patch for a bug in server/select.c
by Alexandre Julliard May 19, 2001

May 19, 2001

Bang Jun-Young <bjy(a)mogua.org> writes: > According to the manpage, NetBSD implementation of poll(2) is different > from historical one in that a given fd may not cause poll to return with > an error. I'm not sure if this is the case, however. I'm not sure I understand what they mean in the manpage. Could you write a small test program to determine the exact behavior WRT invalid fds? Particularly interesting would be to determine if poll actually sleeps waiting for something, or if it returns POLLNVAL at once. -- Alexandre Julliard julliard(a)winehq.com

2 1

More breakages
by gerard patel May 19, 2001

May 19, 2001

The patch moving queue management to the server has made about 10 casualties in my toy test suite (btw, why is this patch not appearing on any archive of wine-cvs either Integrita or the winehq one ? I am not suscribed to wine-cvs so I don't know if anyone has received it at all) The simplest problem is : --- timer.c.orig Sat May 19 01:24:04 2001 +++ timer.c Sat May 19 10:04:57 2001 @@ -104,7 +104,7 @@ TIMER * pTimer; HWINDOWPROC winproc = 0; - if (GetWindowThreadProcessId( hwnd, NULL ) != GetCurrentThreadId()) + if (hwnd && (GetWindowThreadProcessId( hwnd, NULL ) != GetCurrentThreadId())) { SetLastError( ERROR_INVALID_WINDOW_HANDLE ); return 0; This fixes ForteFreeAgent dying immediately with a 'no windows timer', Bcb4, and also funny menus problems. I checked GetWindowThreadProcessId under NT and it fails for hwnd=0 so I think that this is the proper fix. But there are other strange crashes that I have not yet investigated, also clicking in the vertical scrollbar in Acrobat Reader scrolls too much. Gerard

2 1

queue management problem
by eric pouech May 19, 2001

May 19, 2001

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 May 19, 2001

May 19, 2001

> +#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 May 18, 2001

May 18, 2001

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 May 18, 2001

May 18, 2001

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 May 18, 2001

May 18, 2001

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 May 18, 2001

May 18, 2001

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