June 2004 - Wine-Devel - list.winehq.org

Re: Change REFIID to IID* for stack declarations in oaidl_p.c to avoid initializers being discarded

by Mike Hearn

On Sun, 27 Jun 2004 21:13:17 +0200, Marcus Meissner wrote: >> + It is usually generated directly by MIDL, however this file has >> + been tweaked since then to account for slight differences in the way >> + gcc and MSVC++ compile it. In particular, in some functions REFIIDs >> + declared on the stack have been converted to plain IID* in order to eliminate >> + the constness of the REFIID type, ensuring that the zero initializer is not >> + discarded. >> + >> + Therefore, please do not regenerate this file. >> +*/ > > The comment is not understandable. > > Shouldn't it read '"REFIID" should not be used, because the value is not "const"'? I don't understand that wording of the comment either unfortunately :( It's not that it shouldn't be used, it's more that in the original generated code (I think) it was written like this: REFIID riid; (REFIID ) riid = 0; I'm not sure why MIDL did this, but as REFIID is simply #define REFIID const IID* const changing those instances to IID* should "fix" it without any warnings about assigning to a readonly variable. REFIID can still be used elsewhere though. thanks -mike

June 27, 2004

Re: Tooltips: Start of Balloon Support

by Filip Navara

Dmitry Timoshkov wrote: >"Robert Shearman" <R.J.Shearman(a)warwick.ac.uk> wrote: > >>This patch marks the start of balloon tip support. Note that it also has >>the same bug as native due to regions not being handled the same as on >>Windows. >> >> > >Do you mean a bug in FrameRgn? If yes, ReactOS might have that bug >fixed in their implementation. Could you have a look? > If that's the case, I'll try to fix it. I will propably have time to look at this issue tomorrow. Regards, Filip

June 27, 2004

Re: Change REFIID to IID* for stack declarations in oaidl_p.c to avoid initializers being discarded

by Marcus Meissner

On Sun, Jun 27, 2004 at 05:23:27PM +0100, Mike Hearn wrote: > This can cause a subtle marshalling failure in some cases > (inter-thread IDispatch, native MSI, etc) > > > Mike Hearn <mh(a)codeweavers.com> > Change REFIID to IID* for stack declarations in oaidl_p.c to avoid > initializers being discarded > Index: oaidl_p.c > =================================================================== > RCS file: /home/wine/wine/dlls/oleaut32/oaidl_p.c,v > retrieving revision 1.2 > diff -u -p -w -r1.2 oaidl_p.c > --- oaidl_p.c 5 Sep 2003 23:08:33 -0000 1.2 > +++ oaidl_p.c 27 Jun 2004 16:17:04 -0000 > @@ -1,5 +1,14 @@ > -/* this ALWAYS GENERATED file contains the proxy stub code */ > +/* This file contains the proxy/stub code for core COM interfaces. > > + It is usually generated directly by MIDL, however this file has > + been tweaked since then to account for slight differences in the way > + gcc and MSVC++ compile it. In particular, in some functions REFIIDs > + declared on the stack have been converted to plain IID* in order to eliminate > + the constness of the REFIID type, ensuring that the zero initializer is not > + discarded. > + > + Therefore, please do not regenerate this file. > +*/ The comment is not understandable. Shouldn't it read '"REFIID" should not be used, because the value is not "const"'? Ciao, Marcus

June 27, 2004

Re: Wine/WineX and subfs

by Marcus Meissner

On Sun, Jun 27, 2004 at 06:04:32PM +0200, Uwe Bonnes wrote: > Hallo, > > when trying to install programs from subfs mounted CDROMS, wine stumbled on > several occasions (GetVolumenInformation?), running on Suse 9.1. I had not > yet time to investigate deeper. However the announcement for WineX 3.3.2 > tells explicit about changes for CDROMs and subfs for Suse9.1. Anybody from > Transgaming reading hear, willing to submit a patch or explain the changes? The CD detection code is different between WineX and WineHQ. I tried both winefile and notepad (file->open) and both detect the subfs mounts correctly and fine. Ciao, Marcus

June 27, 2004

Re: FILE_GetNtStatus Converting errno 38 to STATUS_UNSUCCESSFUL

by Jürgen Leibner

"Jürgen Leibner" <juergen.leibner(a)t-online.de> schrieb im Newsbeitrag news:cbhj09$kct$1(a)sea.gmane.org... Hi there, > i'm trying to install a Software for Schools using wine-20040505-SuSELinux90.i586.rpm. Now, after updating to wine-20040615-SuSELinux90.i586.rpm, the Error doesn't apear anymore! The setup.exe does its job correct. After this the MS-Installer is invoked to do the rest of the job with an msi-file. He crashes at once. So i think i have to contact the programmer of the software to ask him what the installer wants to do at crashtime. For this i will dump the windump output to explain the point of crashing. Thanks to all, Jürgen

June 27, 2004

Wine/WineX and subfs

by Uwe Bonnes

Hallo, when trying to install programs from subfs mounted CDROMS, wine stumbled on several occasions (GetVolumenInformation?), running on Suse 9.1. I had not yet time to investigate deeper. However the announcement for WineX 3.3.2 tells explicit about changes for CDROMs and subfs for Suse9.1. Anybody from Transgaming reading hear, willing to submit a patch or explain the changes? Bye -- Uwe Bonnes bon(a)elektron.ikp.physik.tu-darmstadt.de Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt --------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------

June 27, 2004

Re: Change Desktopsize for _INS5576._MP in config

by Uwe Bonnes

>>>>> "Uwe" == Uwe Bonnes <bon(a)elektron.ikp.physik.tu-darmstadt.de> writes: Uwe> Changelog: documentation/samples/config: Uwe> [AppDefaults\\_INS5576._MP\\x11drv] Change Desktop size to 800x600 Uwe> With a desktop siye of 640y480, Sage PC Kaufmann Setup aborts, Uwe> asking for higher resolution. Looking at http://frankscorner.org/modules.php?op=modload&name=Sections&file=index&req… the small desktopsize used until now for _INS5576._MP seems to cause problems with the photoshop installer too... -- Uwe Bonnes bon(a)elektron.ikp.physik.tu-darmstadt.de Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt --------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------

June 27, 2004

Re: Tooltips: Start of Balloon Support

by Dmitry Timoshkov

"Robert Shearman" <R.J.Shearman(a)warwick.ac.uk> wrote: > This patch marks the start of balloon tip support. Note that it also has > the same bug as native due to regions not being handled the same as on > Windows. Do you mean a bug in FrameRgn? If yes, ReactOS might have that bug fixed in their implementation. Could you have a look? -- Dmitry.

June 27, 2004

Suppressing NTDLL Relay messages

by Uwe Bonnes

Hallo, as the include and exclude lists for relay logging are initialized later than our ntdll is loader, relay logs contain all ntdll calls, even if the user didn't request them. Does anybody have an idea how to make the ntdll relay behave better? Bye -- Uwe Bonnes bon(a)elektron.ikp.physik.tu-darmstadt.de Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt --------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------

June 27, 2004

Re: dbghelp crashes

by Eric Pouech

Jon Griffiths a écrit : > Hi, > > Some overly long stabs cause crashes when debug tracing is enabled. > > Cheers, > Jon > > +dlls/dbghelp/stabs.c > Prevent the debug buffer from overflowing on long stabs You shouldn't use debugstr_an with a fixed value, since all strings are '\0' terminated. Either fix dlls/ntdll/debugtools.c to handle properly those long string conditions (and remove the ugly call to abort()), or use min(256, strlen(ptr)) as the actual length of the string in debugstr_an calls. I'd prefer the first solution. A+

June 27, 2004

Re: wine/loader preloader.c

by Dmitry Timoshkov

"Mike Hearn" <mh(a)codeweavers.com> wrote: > On Sat, 2004-06-26 at 08:25 -0500, Mike McCormack wrote: > > I think a better fix is to define it if it's missing. > > This commit by Alexandre should fix the segfaults people have been > seeing on Fedora Core 2. > > Either of the two patches just sent are correct BTW, Red Hat 9/Fedora > Core 1 do not place AT_SYSINFO* headers in the ELF startup auxiliary > vector. Both patches are correct if you take into account compilation only, but if you compile Wine without defined AT_SYSINFO_EHDR, and then try to distribute a Wine binary then Mike's patch is definitely better. -- Dmitry.

June 27, 2004

Re: wine/loader preloader.c

by Mike Hearn

On Sat, 2004-06-26 at 08:25 -0500, Mike McCormack wrote: > I think a better fix is to define it if it's missing. This commit by Alexandre should fix the segfaults people have been seeing on Fedora Core 2. Either of the two patches just sent are correct BTW, Red Hat 9/Fedora Core 1 do not place AT_SYSINFO* headers in the ELF startup auxiliary vector. thanks -mike

June 26, 2004

Re: Wine causing a Segmentation Fault

by Vincent Béron

Le ven 25/06/2004 à 20:21, Mark A Fonnemann a écrit : > Mike Hearn <mike <at> navi.cx> writes: > > > > >On Tue, 22 Jun 2004 10:32:12 -0700, David Lee wrote: > >> I am using the CVS wine as of 6/22/04 and when I run the compiled binary > >> of wine, I get a Segmentation Fault. > > > > Does it work if you run wine-pthread directly? > > i didn't start this thread but i am equally (maybe more so?) interested to > see this Wine & Fedora problems (finally) resolved. i'm using FC2 and yes, > i can attest that using wine-pthread directly seemingly allows things to > work correctly. Could you try current CVS and see if it works (using straight wine)? Alexandre committed a fix for that yesterday. Vincent

June 26, 2004

Re: !LOSTWAGES! site misc updates

by Ivan

> Do you really think we need to switch from American to English spelling? A common solution is to spell the same word in A.E. in some places and in B.E. in others, and that's what my patch does. Apart fomr that, it's not only a spelling patch. Ivan.

June 26, 2004

Re: wine/loader preloader.c

by Saulius Krasuckas

On Sat, 26 Jun 2004, Michael Stefaniuc wrote: > On Red Hat Linux 9 there is no AT_SYSINFO_EHDR in /usr/include/elf.h > thus breaking compilation there. I confirm it: on Red Hat Linux 8 the definition is missing also.

June 26, 2004

Re: wine/loader preloader.c

by Michael Stefaniuc

Hello, On Fri, Jun 25, 2004 at 07:12:20PM -0500, Alexandre Julliard wrote: > ChangeSet ID: 12717 > CVSROOT: /opt/cvs-commit > Module name: wine > Changes by: julliard(a)wine.codeweavers.com 2004/06/25 19:12:20 > > Modified files: > loader : preloader.c > > Log message: > Remove the AT_SYSINFO and AT_SYSINFO_EHDR values if the sysinfo page > is in one of our reserved ranges. On Red Hat Linux 9 there is no AT_SYSINFO_EHDR in /usr/include/elf.h thus breaking compilation there. I've attached a trivial patch to fix this. > > Patch: http://cvs.winehq.org/patch.py?id=12717 > > Old revision New revision Changes Path > 1.6 1.7 +62 -20 wine/loader/preloader.c bye michael -- Michael Stefaniuc Tel.: +49-711-96437-199 System Administration Fax.: +49-711-96437-111 Red Hat GmbH Email: mstefani(a)redhat.com Hauptstaetterstr. 58 http://www.redhat.de/ D-70178 Stuttgart

June 26, 2004

Re: !LOSTWAGES! site misc updates

by Dimitrie O. Paun

On Fri, Jun 25, 2004 at 07:35:23PM +0200, Ivan wrote: > As in subject. Jeremy, please let me know if there's anything wrong in this patch. Do you really think we need to switch from American to English spelling? What if someone else is sending a patch to convert back to American? I think the spelling is fine the way it is now. -- Dimi.

June 26, 2004

Re: Wine causing a Segmentation Fault

by Mark A Fonnemann

Mike Hearn <mike <at> navi.cx> writes: > >On Tue, 22 Jun 2004 10:32:12 -0700, David Lee wrote: >> I am using the CVS wine as of 6/22/04 and when I run the compiled binary >> of wine, I get a Segmentation Fault. > > Does it work if you run wine-pthread directly? i didn't start this thread but i am equally (maybe more so?) interested to see this Wine & Fedora problems (finally) resolved. i'm using FC2 and yes, i can attest that using wine-pthread directly seemingly allows things to work correctly. i.e. [root(a)markf78 windows]# wine notepad.exe & [1] 8068 [root(a)markf78 windows]# [1]+ Segmentation fault wine notepad.exe [root(a)markf78 windows]# wine-pthread notepad.exe & [1] 8070 [root(a)markf78 windows]# [1]+ Done wine-pthread notepad.exe [root(a)markf78 windows]# this lingering problem between Fedora and Wine is in bugzilla here: http://bugs.winehq.org/show_bug.cgi?id=2137 mark. :-)

June 26, 2004

Chasing Galactic Civilizations Unhandled Exception

by Mike Kost

After sucking it up and realizing I'd have to learn how to navigate winedbg, I started playing around with the Galactic Civilizations unhandled exception. Some documentation and a few trips through winedbg later, it appears that Gal Civ is throwing a divide-by-zero -----------<snip>--------------- First chance exception: divide by zero in 32-bit code (0x40f2cda8) -----------<snip>--------------- I did a 'bt' next: -----------<snip>--------------- =>1 0x40f2cda8 send_mouse_event+0xc8(hwnd=0x10024, flags=0x8001, posX=0x1fb0000, posY=0x1df, data=0x0, time=0xef3d) [mouse.c:135] in x11drv (0x406cfbb0) 2 0x40f2e010 X11DRV_MotionNotify+0x60(hwnd=0x10024, event=0x406cfc4c) [mouse.c:613] in x11drv (0x406cfbdc) 3 0x40f24e90 .L79+0xa in x11drv (0x406cfc3c) ... stuff ... -----------<snip>--------------- and then did an 'info local' and got: -----------<snip>--------------- send_mouse_event: struct HWND__* hwnd = 0x00010024 (parameter) long unsigned int flags = 0x00008001 (parameter) long unsigned int posX = 0x01fb0000 (parameter) long unsigned int posY = 0x000001df (parameter) long unsigned int data = 0x00000000 (parameter) long unsigned int time = 0x0000ef3d (parameter) unsigned int state = 0x0000004c (local in register ECX) unsigned int state = 0x0000004c (local in register ECX) struct HWND__* hwnd = 0x01faffff (local in register EAX) long unsigned int data = 0x01faffff (local in register EAX) long unsigned int time = 0x0000ef3d (local in register EDI) struct tagINPUT input = 0x40f61028 (local) int width = 0x00000000 (local in register ESI) int height = 0x00000000 (local) -----------<snip>--------------- After digging around in dlls/x11drv/mouse.c (see the bt output), I found: -----------<snip>--------------- int width = GetSystemMetrics( SM_CXSCREEN ); int height = GetSystemMetrics( SM_CYSCREEN ); /* Relative mouse movements seem not to be scaled as absolute ones */ posX = (((long)posX << 16) + width-1) / width; posY = (((long)posY << 16) + height-1) / height; -----------<snip>--------------- The posX and poxY calculations are the only divides in the send_mouse_event function (pointed to by the bt as the stack location when the divide-by-zero occurred). The 'info local' confirms that width and height are indeed 0. My problem is that I do not think GetSystemMetrics should be returning 0. The display is quite chorked up by the time the unhandled exception occurs, so I'm pretty sure something bad has already happened; I just don't know what. If any x11drv experienced folk can provide pointers or insight, I'd greatly appreciate it. Mike Kost mike -at- tashcorp.net

June 25, 2004

Re: problem with linkage of so under RHEL3

by Bill Medland

On June 25, 2004 12:59 pm, Alexandre Julliard wrote: > Bill Medland <billmedland(a)mercuryspeed.com> writes: > > So for future reference is there an easy way to see what functions a so > > provides that don't show up in nm? > > Try nm -D. Thanks. > > > Then I guess the question will be a matter of why the dlopen doesn't see > > the existing functions when underneath Wine. > > Just a wild guess, but you could try adding RTLD_GLOBAL to the dlopen > call where Wine opens libodbc.so. Yep. That solves the link problem. I'll submit it later. Thanks. (Now to figure out why it doesn't run.) -- Bill Medland mailto:billmedland(a)mercuryspeed.com http://webhome.idirect.com/~kbmed

June 25, 2004

Re: FILE_GetNtStatus Converting errno 38 to STATUS_UNSUCCESSFUL

by Alexandre Julliard

Mike Hearn <mike(a)navi.cx> writes: > It just means there was some information loss in converting error codes > internally inside Wine ... you'll have to look elsewhere to figure out > *why* it failed. There is a table of errno codes on MSDN you can use or > look in winerror.h to find out what 38 is. Actually it's a Unix errno, 38 is ENOSYS. It would be interesting to find out what function causes that, a relay trace is probably the best way. -- Alexandre Julliard julliard(a)winehq.org

June 25, 2004

Re: problem with linkage of so under RHEL3

by Alexandre Julliard

Bill Medland <billmedland(a)mercuryspeed.com> writes: > So for future reference is there an easy way to see what functions a so > provides that don't show up in nm? Try nm -D. > Then I guess the question will be a matter of why the dlopen doesn't see the > existing functions when underneath Wine. Just a wild guess, but you could try adding RTLD_GLOBAL to the dlopen call where Wine opens libodbc.so. -- Alexandre Julliard julliard(a)winehq.org

June 25, 2004

Re: FW: FW: WM_NEXTDLGCTL changes the default button ID and does not rest ore default control identifier

by Alexandre Julliard

Krishna Murthy <Krishna.Murthy(a)guptaworldwide.com> writes: > Alexandre, > Is these changes looks good? Do I have to submit this patch to > 'wine-patches(a)winehq.org' ? Well, I'm not really happy with it, that whole bSetDefID thing is fairly ugly IMO. I'd really appreciate if you could write a regression test that illustrates the behavior of both DM_SETDEFID and WM_NEXTDLGCTL so that the code can be restructured in a more readable way. -- Alexandre Julliard julliard(a)winehq.org

June 25, 2004

Re: FILE_GetNtStatus Converting errno 38 to STATUS_UNSUCCESSFUL

by Mike Hearn

On Fri, 25 Jun 2004 18:16:14 +0200, Jürgen Leibner wrote: > # fixme:ntdll:FILE_GetNtStatus Converting errno 38 to STATUS_UNSUCCESSFUL It just means there was some information loss in converting error codes internally inside Wine ... you'll have to look elsewhere to figure out *why* it failed. There is a table of errno codes on MSDN you can use or look in winerror.h to find out what 38 is. thanks -mike

June 25, 2004

Re: problem with linkage of so under RHEL3

by Bill Medland

On June 24, 2004 05:46 pm, Bill Medland wrote: > Guys > > Command line application opens a dll using LoadLibrary. > That Dll links to odbc > odbc uses unixODBC to link to the new Oracle 10g Beta ODBC client. > That client so access rather a lot of Oracle libraries > > Relocation error occurs > > /usr/local/bin/wine-pthread: relocation error: > /opt/oracle/product/10.1.0/client_1/lib/libsqora.so: undefined symbol: > SQLGetPrivateProfileString Can anyone help me on this please. I am basically constrained by my lack of deep knowledge about Linux and would appreciate guidance. I am quite prepared to do my best to help fix the problem with Wine but cannot do it alone. First, can someone explain why the Linux equivalent does work. unittest links to mylib.so using dlopen (..., RTLD_NOW) mylib.so is linked to libodbc.so At the request of mylib via a SQLDriverConnect libodbc.so opens a link to $ORACLE_HOME/lib/libsqora.so. I don't know whether it is RTLD_NOW or RTLD_LAZY because it's down in libtools somewhere Now here is where I get confused. ldd -r libsqora.so shows 7 unresolved functions, including SQLGetPrivateProfileString It also does not show a dependency on libodbc itself. SQLGetPrivateProfileString does exist in libodbc. nm libodbc.so does not show it. Presumably the link works because since it was the libodbc that linked to libsqora.so SQLGetPrivateProfileString was already defined when the dlopen occurred. So for future reference is there an easy way to see what functions a so provides that don't show up in nm? Then I guess the question will be a matter of why the dlopen doesn't see the existing functions when underneath Wine. -- Bill Medland mailto:billmedland(a)mercuryspeed.com http://webhome.idirect.com/~kbmed

June 25, 2004