May 2001 - Wine-Devel - list.winehq.org

Anybody know int-31?
by Bob Goodwin May 21, 2001

May 21, 2001

I want to run an application that requires win95 but relys heavily on DOS interupt functions. Specifically it requires interupt 31 which is just a stub in Wine. It appears (from R.B.'s list) that there is no limit on calls to interupt_31-AX=0305 (request the address of (?:the:a) save/restore handler.) That implies more than one thread could have a save/restore pointer. When calling the save/restore handler function, the only parameter passed is a byte in AL to indicate whether to save the current context or restore a saved context. My question is, are there multiple copies of the save/restore function or at least multiple entry points? It seems that this would be the only way of having multiple unique context saves. It would be differrent if a unique `handle' was returned by int_31 instead of a call address. If only one context can be stored at a time, do I need a spin-lock within the save/restore function to keep other applications holding until their turn? Thanks. Bob Goodwin

1 0

Re: Networking with Mandrake 8.0 and SuSE 7.1
by Steve Fox May 21, 2001

May 21, 2001

This weekend I upgraded to Mandrake's 2.4.4 kernel from Cooker (since it fixed both my mouse and token ring) and suddenly this problem has gone away (using Notes 5.0.7 and wine-20010305). I'm still planning to go through Mandrake's 2.2.19 SRPM and find which patch messes things up (I can't use a Linus-only kernel because I need reiserfs). I will post again once I have narrowed down the list of suspects. I'm guessing it has to be a common patch since SuSE seems to have the same symptoms. Steve Fox IBM Linux Technology Center http://www.ibm.com/linux/ltc http://k-lug.com

1 0

Re: Small fixes in controls/menu.c
by Alexandre Julliard May 20, 2001

May 20, 2001

gerard patel <gerard.patel(a)asi.fr> writes: > But does Windows send this message in this case ? From what > I can see in the traces, I don't think so. Maybe Windows does this differently then. But doing the NC calculation without sending a WM_NCCALCSIZE does not seem right. -- Alexandre Julliard julliard(a)winehq.com

1 0

Re: More breakages
by Alexandre Julliard May 20, 2001

May 20, 2001

gerard patel <gerard.patel(a)asi.fr> writes: > There is a nice FIXME in the new message handling around the > old timer handling code, so maybe fix is pending ? I guess you could say that... Hardware messages need to be integrated into the new queue handling, I'm working on it but it will probably take a few days. -- Alexandre Julliard julliard(a)winehq.com

1 0

Re: More breakages
by gerard patel May 20, 2001

May 20, 2001

At 10:28 AM 20/05/2001 +0200, I wrote: >(one more time following up to myself) >Now searching for the Acrobat Reader funny scrolling... I guess that this one is probably because of a still incomplete implementation; what happens is : - I click in the vertical scrollbar - the text begins to scroll (for some reason this particular text is very slow to scroll) - the scroll timer expires and scrolls again the page The WM_LBUTTONUP (supposed to end this loop) is not received until 20 pages later. If I test with a text that scrolls much faster, the problem is much less severe (only one page too much). It seems to be a message order problem; before Wine had - hardware messages (mouse, keyboard) - timer Now it is - timer - hardware messages There is a nice FIXME in the new message handling around the old timer handling code, so maybe fix is pending ? Gerard

1 0

Re: More breakages
by gerard patel May 20, 2001

May 20, 2001

At 02:49 PM 20/05/2001 +0200, you wrote: >That sounds correct, even for all mailing lists, not just wine-cvs. The >hypermail archives at winehq.com is updated simultaneously with the list >mail being sent out to all subscribers. Ah, I was not understanding how the moderation is working (just read Andi's explanations on cemw). So the extraordinary delays I see sometimes on my posts to wine mailing lists are probably caused by my ISP beta testing IP on avian carriers. >(This is not the case with the >integrita archives, though, since integrita is just one of the >subscribers, subject to the whims of the MTA.) Did you use Reload? It's hard to remember what I did past 2 in the morning. I think that the following morning I looked only at Integrita; they seem to have not taken in account anything since. I'll take a look more accurately next time. Thanks for the explanations. Gerard

1 0

What about XP?
by Uhmmmm May 20, 2001

May 20, 2001

Just wondering: Is there anything being done to update wine for windoze XP?

2 1

Re: Small fixes in controls/menu.c
by Alexandre Julliard May 20, 2001

May 20, 2001

gerard patel <gerard.patel(a)asi.fr> writes: > (patch using NC_HandleNCCalcSize instead of SetWindowPos) > > IMO it's better and it solves my problem as well. I think the SetWindowPos one was better. If the app handles WM_NCCALCSIZE we shouldn't try to bypass it. -- Alexandre Julliard julliard(a)winehq.com

1 0

Re: wineps include fix.
by Ian Pilcher May 20, 2001

May 20, 2001

David Hammerton wrote: > > it should be > #include <freetype/ftsnames.h> > instead of > #include <freetype/ftnames.h> > David - It's definitely ftnames.h on my system (Red Hat 7.1). What version of FreeType do you have? -- ======================================================================== Ian Pilcher ian.pilcher(a)home.com ========================================================================

3 3

wineps include fix.
by David Hammerton May 20, 2001

May 20, 2001

hey mate.. just got latest CVS of wine (From winehq), it didnt compile, something about not finding "freetype/ftnames.h"... i couldnt find it on my system, in any deb packages, no one had it... so i went hunting, find you left out an 's'.. it should be #include <freetype/ftsnames.h> instead of #include <freetype/ftnames.h> i've cc'd this to "wine-devel"... so yer, if you could fix it, thatd be cool.. but whatever ;p cheers David -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ David Hammerton aka crazney Victoria, Australia http://crazney.net/ HOWTO maintainer and Linux Half-Life http://lhl.linuxgames.com/ email: http://crazney.net/contact.html -> Note: this email address expires after one week.

1 0

Re: More breakages
by gerard patel May 20, 2001

May 20, 2001

At 02:38 PM 19/05/2001 +0200, you wrote: >What do you mean? It's there... >http://www.winehq.com/hypermail/wine-cvs/2001/05/0095.html Yes :-/ I should have checked again, I was thinking that because wine-cvs input was generated automatically there could be no reason for data not to be posted to archives immediately. I checked yesterday and it seems that a delay of several hours can happen between a cvs commit and the updating of archives. Gerard

2 1

Re: More breakages
by gerard patel May 20, 2001

May 20, 2001

At 11:07 AM 19/05/2001 +0200, I wrote: (again following up to myself) >But there are other strange crashes that I have not yet investigated Now I have, this change in server/queue.c removes the 3 remaining hangings : --- queue.c.orig Sat May 19 21:06:18 2001 +++ queue.c Sun May 20 09:37:09 2001 @@ -695,7 +695,7 @@ } /* now check for WM_PAINT */ - if (queue->wake_bits & QS_PAINT) + if (((WM_PAINT >= req->get_first) && (WM_PAINT <= req->get_last)) && (queue->wake_bits & QS_PAINT)) { req->type = 0; req->win = 0; @@ -707,6 +707,9 @@ } /* now check for timer */ This check was done in previous code, but I am not completely sure that this is the correct place for this check, though. Now searching for the Acrobat Reader funny scrolling... Gerard

1 0

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

May 20, 2001

At 03:33 AM 20/05/2001 +0200, you wrote: >I think it shouldn't generate infinite loops, but my patch could work >improperly when app catches and doesn't pass to defwndproc WM_NCCALCSIZE >message. So I wrote next patch (attached to this e-mail). (patch using NC_HandleNCCalcSize instead of SetWindowPos) IMO it's better and it solves my problem as well. >Any comments,hints ? - you are supposed to send patches against current Wine, not an outdated version - you are supposed to { respect the indentation style of the file you patch; } >When WIN_FindWndPtr(hWnd) return a NULL pointer doesn't it means that >something bad happens ? Yes but I have a hunch that this WARN will be edited out by the Wine maintainer anyway. You'll notice that in similar cases there are never warnings in Wine code; this is not because Wine coders are lazy, it's because the Wine maintainer thinks it's a bad idea to add warnings everywhere. Gerard

1 0

Re: Small fixes in controls/menu.c
by gerard patel May 19, 2001

May 19, 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

Re: Networking with Mandrake 8.0 and SuSE 7.1
by Steve Fox May 18, 2001

May 18, 2001

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 May 18, 2001

May 18, 2001

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