September 2009 - Wine-Devel

Re: [4/5] WineD3D: Add a create flag for addref in GetPrivateData
by Henri Verbeet Sept. 30, 2009

Sept. 30, 2009

2009/9/30 Stefan Dösinger <stefan(a)codeweavers.com>: > This is hard to implement in the client lib - it would require to remove the > AddRef from wined3d altogether, and pass the priv data flags to the client, > and then AddRef in d3d8, d3d9 and dxgi/d3d10. I can't use a Release in ddraw > because we're calling application code here which might break if the app's > AddRef() implementation is broken. Handling the AddRef in d3d8/9/10 is a lot > of work because those libs don't have any IDirect3D*Resource base class that > could do this in a central place. I'm not sure why you'd need an entire baseclass, an utility function (which is how it's done in wined3d!) should do just fine. > +const UINT WINED3DCREATE_GET_PRIV_DATA_NO_ADDREFF = 0x00000004; There should only be a single 'F' in ADDREF.

1 0

Re: [2/5] WineD3D: Start replacing dxVersion with create flags
by Henri Verbeet Sept. 30, 2009

Sept. 30, 2009

2009/9/30 Stefan Dösinger <stefan(a)codeweavers.com>: > Is there a real reason this (and the flag from the next patch) has to be set on the wined3d object? If there is, I think WINED3D_CREATE_BUFFER_NO_CONVERSION and WINED3D_CREATE_BUFFER_STATIC_FORMAT would be more appropriate names.

1 0

Re: [comctl32/tests] Fix a test failure on several platforms
by Paul Vriens Sept. 30, 2009

Sept. 30, 2009

On 09/30/2009 02:21 PM, Paul Vriens wrote: > Hi, > > Positioning the point test a bit to the left makes this test succeed on > more boxes. > > (Maybe we should add MCHT_TITLEYEAR to the possible outcomes but I'm not > too sure about that). > > Changelog > Fix a test failure on several platforms > Just checked some other test failures on an XP. This one has more issues but also shows MCHT_TITLEMONTH for the test changed in this patch. Now I'm pretty convinced the fix is not correct, but I stand corrected. -- Cheers, Paul.

1 0

Compile error on current Git
by Paul Vriens Sept. 30, 2009

Sept. 30, 2009

Hi, Anyone else seeing this: make[2]: Entering directory `/wine/wine-git/dlls/system.drv16' ccache gcc -m32 -m32 -c -I. -I. -I../../include -I../../include -D__WINESRC__ -D_REENTRANT -fPIC -Wall -pipe -fno-strict-aliasing -Wdeclaration-after-statement -Wwrite-strings -Wtype-limits -Wpointer-arith -Wcast-qual -Wlogical-op -o system.o system.c ../../tools/winegcc/winegcc -m32 -B../../tools/winebuild --sysroot=../.. -shared ./system.drv16.spec system.o -Wb,--subsystem,win16 -o system.drv16.so -lkernel32 ../../libs/port/libwine_port.a system.o: In function `call_timer_proc16': system.c:(.text+0x54a): undefined reference to `NtCurrentTeb' collect2: ld returned 1 exit status winegcc: ccache failed make[2]: *** [system.drv16.so] Error 2 make[2]: Leaving directory `/wine/wine-git/dlls/system.drv16' make[1]: *** [system.drv16] Error 2 (Fedora 11, gcc 4.4.1) -- Cheers, Paul.

2 2

Re: [PATCH 2/2] ws2_32/tests: test WSAAccept with CF_DEFER
by Paul Vriens Sept. 30, 2009

Sept. 30, 2009

On 09/25/2009 07:59 AM, Mike Kaplinskiy wrote: > --- > dlls/ws2_32/tests/sock.c | 46 +++++++++++++++++++++++++++++++++++++--------- > 1 files changed, 37 insertions(+), 9 deletions(-) > > > ------------------------------------------------------------------------ > > Hi Mike, This one has introduced a test failure on Win98 and WinMe. Could you have a look? Let me know if you want me to run extra tests on these boxes. -- Cheers, Paul.

2 2

Re: ddraw: set dwBackBufferCount=0 in CreateSurface for backbuffer
by Stefan Dösinger Sept. 29, 2009

Sept. 29, 2009

Am 29.09.2009 um 15:09 schrieb Markus Stockhausen: > Sorry for the incomplete topic of my first mail. > > <patch.diff> The patch looks ok to me

1 0

Re: wininet(1/3): Only initialize SSL_CTX once
by Hans Leidekker Sept. 29, 2009

Sept. 29, 2009

Hi Juan, > These patches collectively fix bug 18364, which was a crash in OpenSSL > in a multithreaded app (utorrent.exe.) The bug reporter sees long delays between torrents starting up after applying these patches, which makes we wonder if the ssl lock effectively serializes secure https requests? Would patch 1 and 3 not be sufficient to fix the utorrent crash? -Hans

2 1

Re: set caps.dwBackBufferCount=0 for backbuffer surface
by Paul Vriens Sept. 29, 2009

Sept. 29, 2009

On 09/29/2009 03:01 PM, Markus Stockhausen wrote: > > > ------------------------------------------------------------------------ > > Hi Markus, Please stick to the coding style of the file, or at least the surrounding code (implementation and test). -- Cheers, Paul.

2 2

Re: comctl32: Send the parent notification after the Wine has set the tooltip position (bug #14336)
by Nikolay Sivov Sept. 29, 2009

Sept. 29, 2009

Could you please add a test for that?

2 1

Re: wininet: Handle finding ftp files in subdirectories and using wildcards.
by Detlef Riekenberg Sept. 29, 2009

Sept. 29, 2009

On Sa, 2009-09-26 at 22:15 -0500, Mike Ruprecht wrote: > + if (lpszSearchPath) > + HeapFree(GetProcessHeap(), 0, lpszSearchPath); > + Please remove the unneeded if(lpszSearchPath) before HeapFree. -- By by ... Detlef

1 0

Re: ole32/tests: guard ReleaseStgMedium with SUCCEEDED
by Alexandre Julliard Sept. 29, 2009

Sept. 29, 2009

<Joerg-Cyril.Hoehle(a)t-systems.com> writes: > @@ -1401,7 +1405,7 @@ static void test_getdatahere(void) > hr = IDataObject_GetDataHere(get, &fmt, &med); > ok(hr == S_OK, "got %08x\n", hr); > ok(med.tymed == TYMED_HGLOBAL, "got %x\n", med.tymed); > - ReleaseStgMedium(&med); > + if(SUCCEEDED(hr)) ReleaseStgMedium(&med); I don't think that's correct for the GetDataHere case, it's allocated by the caller. -- Alexandre Julliard julliard(a)winehq.org

1 0

Re: server(4/5): Setting a process's affinity sets all of its threads' affinities too
by Alexandre Julliard Sept. 29, 2009

Sept. 29, 2009

Juan Lang <juan.lang(a)gmail.com> writes: > +static void set_process_affinity( struct process *process, affinity_t affinity ) > +{ > + struct thread *thread; > + > + process->affinity = affinity; > + > + suspend_process( process ); > + LIST_FOR_EACH_ENTRY( thread, &process->thread_list, struct thread, proc_entry ) > + { > + set_thread_affinity( thread, affinity ); > + } > + resume_process( process ); Why are you suspending the process? -- Alexandre Julliard julliard(a)winehq.org

1 0

help with dwBackBufferCount in backbuffer surfaces needed
by Markus Stockhausen Sept. 29, 2009

Sept. 29, 2009

Hi, currently I'm investigating bug 1660 and hopefully I found the reason for the misbehaviour. The only problem is, that I do not know what results I have to expect in windows. So it would be very helpful if someone can execute the attached testcase in a native environment so that I can fix the error condition and send in the final patch. The idea is that backbuffer surfaces have dwBackBufferCount=0 in Windows but on Wine they get the same count as the primary surface. The test will fail in Windows and Wine to get the results in both cases. The result will show like this: dsurface.c:1264: Test failed: backbuffer surface has dwBackBufferCount==2 Thanks for your help in advance. Markus

4 4

Re: [PATCH 10/10] add tests for SHOpenRegStream2A and SHCreateMemStream
by Alexandre Julliard Sept. 29, 2009

Sept. 29, 2009

Peter Oberndorfer <kumbayo84(a)arcor.de> writes: > +#define test_streamsize(stream, size) \ > +{ \ > + STATSTG stat; \ > + HRESULT result; \ > + result = IStream_Stat(stream, &stat, STATFLAG_NONAME); \ > + ok(result == S_OK, "IStream::Stat failed %08x\n", result); \ > + ok(stat.cbSize.QuadPart == size, "Stat size wrong\n"); \ > +} > + > +#define test_value_type(name, wantedtype) \ > +{ \ > + LONG retval; \ > + DWORD type; \ > + retval = RegQueryValueExA(hBaseKey, name, NULL, &type, NULL, NULL); \ > + ok(retval == ERROR_SUCCESS, "RegQueryValueExA should succeed, error %d\n", retval); \ > + ok(type == wantedtype, "type of key wrong. wanted %d, got %d\n", wantedtype, type); \ > +} > + > +#define test_value_not_exist(name) \ > +{ \ > + LONG retval; \ > + retval = RegQueryValueExA(hBaseKey, name, NULL, NULL, NULL, NULL); \ > + ok(retval == ERROR_FILE_NOT_FOUND, "RegQueryValueExA returned wrong error %d\n", retval); \ > +} > + > +/* older versions (Win98) seem broken, they do not delete the key on close when it is empty */ > +#define test_value_type_broken(name, wantedtype) \ > +{ \ > + LONG retval; \ > + DWORD type; \ > + retval = RegQueryValueExA(hBaseKey, name, NULL, &type, NULL, NULL); \ > + ok(retval == ERROR_SUCCESS, "RegQueryValueExA should succeed, error %d\n", retval); \ > + ok(type == wantedtype || broken(type != wantedtype), "type of key wrong. wanted %d, got %d\n", wantedtype, type); \ > +} > + > +#define test_value_not_exist_broken(name) \ > +{ \ > + LONG retval; \ > + retval = RegQueryValueExA(hBaseKey, name, NULL, NULL, NULL, NULL); \ > + ok((retval == ERROR_FILE_NOT_FOUND) || broken(retval == ERROR_SUCCESS), \ > + "RegQueryValueExA returned wrong error %d\n", retval); \ > +} Please avoid hiding the code inside macros. When you have only a couple of lines just spell things out explicitly. For larger things write an inline function. -- Alexandre Julliard julliard(a)winehq.org

1 0

Re: [PATCH 02/15] comctl32/monthcal: Make tests not depend from each other
by Paul Vriens Sept. 29, 2009

Sept. 29, 2009

On 09/27/2009 07:51 PM, Nikolay Sivov wrote: > Changelog: > - Inependent tests are always better and more scalable. > > > ------------------------------------------------------------------------ > > Hi Nikolay, This one seems (there were more for monthcal.c) to have introduced 2 test failures on a lot of boxes. The only reason I see (from the logs on test.winehq.org) is that the failing boxes don't run the English locale. Any idea? -- Cheers, Paul.

2 1

RFC: Sharing a wineprefix
by Paul Chitescu Sept. 29, 2009

Sept. 29, 2009

Hi people! I know this has been talked about again and again and no satisfactory solution (generic, safe, somehow simple) has been found for all. This proposal aims to solve a common scenario with minimum of changes in Wine, in particular remaining compatible with existing installations. It doesn't solve all problems but a quite pressing one: have a bunch of programs installed just once and used by several users. The idea is to have two categories of users: - The Administrator: - Is the owner of the WINEPREFIX, just like now - R/W access to the prefix - Can install and uninstall programs - Exclusive lock, no other user can run in the same prefix at the same time - All others are restricted users: - R/O access to the prefix (including the machine Registry) - They can only write to their profile (including the user Registry) - Can install programs only to their profile if the installer supports so - Cannot run wine in that prefix while the Administrator runs it - Multiple restricted users can run in the same prefix at once UNIX access rights and/or symlinks will be used to: - Make sure other users cannot write to the prefix - Each user has a Wine visible path to his profile directory - Each user can create a new profile directory - or, subject to policy, the Administrator may create (but no need to populate) them in advance Required changes I've identified so far: - The R/W lock to the prefix - wineserver -k as admin should terminate all users' instances - wineserver needs to be able to load Registry hives R/O - wineboot needs to be able to create just the user's profile and ignore pending renames for restricted users - Report to application the detected type of user: Administrator / Restricted - Some winecfg settings need to be migrated from HKCU to HKLM so the admin can enforce sane values or provide proper defaults or because they are logically per prefix - The user Registry hive and the temporary directory for the restricted users need to be stored in their profile directory Advantages: - Not much coding required, less chances to introduce bugs - Will not break existing installations - No trust is required between users, only to Administrator - Each user runs its own wineserver, there is no communication between them - Solves the most common requests from users - Required step to a more complete (but dangerous) extension Disadvantages: - Single admin user - The Administrator cannot run concurrently with other users - To install or remove programs all users need to shutdown their Wine - A restricted user cannot promote to Admin just to do one action (but of course he can su, wineserver -k others, run as Administrator) - Some programs (although not many, most need just to install) need to run as Administrator - After a Wine upgrade the Administrator needs to upgrade the prefix too - All users share the view of the dosdevices (can be moved to the profile dir if we really want to) - Users cannot install programs (except some) or fonts - Doesn't allow non-simultaneous usage of the prefix by different Admins Opinions? Comments? Votes? Although most changes are not complex there is much work and this needs some political decision to proceed to actual coding.

5 5

programs: add extrac32 tool
by Kirill K. Smirnov Sept. 29, 2009

Sept. 29, 2009

Hi, +int PASCAL wWinMain ( HINSTANCE hInstance, HINSTANCE prev, LPWSTR cmdline, int show ) +{ extrac32 is a console application, isn't it? Why not wmain() instead? And get rid of rather complex ParseCmdLine(). + if (!SetupIterateCabinetW( pszCabFile, 0, (PSP_FILE_CALLBACK) ExtCabCallback, 0 )) + WINE_FIXME( "Can not extract cab file:%s!!!\n", wine_dbgstr_w( pszCabFile ) ); This is an error, not a FIXME + if(!argv) + { + WINE_FIXME("Error!Can not get cmd line\n" ); + return 0; + } not a FIXME + if (argc < 2) + { + WINE_FIXME( "No options\n" ); + return 1; + } not a FIXME + if (fArgc != 1) + WINE_FIXME( "Invalid usage for /C flag - two filenames expected\n" ); not a FIXME -- Kirill

2 1

Re: [PATCH 06/12] d3d10: Implement ID3D10EffectType::GetMemberName().
by Henri Verbeet Sept. 28, 2009

Sept. 28, 2009

> static LPCSTR STDMETHODCALLTYPE d3d10_effect_type_GetMemberName(ID3D10EffectType *iface, UINT index) > { > - FIXME("iface %p, index %u stub!\n", iface, index); > + struct d3d10_effect_type *This = (struct d3d10_effect_type *)iface; > > - return NULL; > + TRACE("iface %p, index %u\n", iface, index); > + > + if(index >= This->member_count) return NULL; > + > + return This->members[index].name; > } You can't do this yet, it needs to work for types created by parse_fx10_type() as well. The same goes for patches 7-11. Patches 1-5 look ok to me, and can stand on their own.

2 1

Review Request: Patch to selectively bind to interfaces while still allowing broadcast packets
by Erich Hoover Sept. 28, 2009

Sept. 28, 2009

Attached is a patch I have been working to fix Bug #7929 (C&C 3 network does not work). This patch replaces interface-specific bind requests (on UDP datagrams only) with a global bind and stores the interface index. The patch then uses IP_PKTINFO to set the outgoing interface and filter incoming packets based on interface (checking against the stored interface index from the initial bind). Additionally, the patch peeks at packets in "select" and asynchronous I/O operations to ensure that spurious wakeups do not occur when a packet comes in from the wrong interface. I intend to split the patch into the following pieces: 1) UDP broadcast tests (used for making sure everything works) 2) The "bare essentials" (storing and retrieving the interface index, and setting/filtering the interface) 3) Ensure "select" does not wake up on packets with an interface mismatch 4) Ensure Async WSARecv does not wake up on packets with an interface mismatch 5) Ensure Async ReadFile does not wake up on packets with an interface mismatch I would greatly appreciate any comments people have on this patch. Thanks so much! Erich Hoover ehoover(a)mines.edu

1 0

Re: [PATCH 3/4] mshtml: Add helper function get_url_components
by Detlef Riekenberg Sept. 28, 2009

Sept. 28, 2009

On Fr, 2009-09-25 at 15:30 -0500, Andrew Eikum wrote: > +static HRESULT get_url_components(HTMLLocation *This, URL_COMPONENTSW > *url) > +{ > + const WCHAR *doc_url; > + if(!InternetCrackUrlW(doc_url, 0, 0, url)) { > + FIXME("InternetCrackUrlW failed\n"); Such a FIXME does not help to find the reason. Dumping doc_url helps, but might add personal data to the log -- By by ... Detlef

2 1

Patch for [Bug 14222] RpcBindingSetAuthInfo(Ex) fails with RPC_C_AUTHN_GSS_NEGOTIATE
by Stefan Kuhr Sept. 28, 2009

Sept. 28, 2009

Hello everyone, please find attached a patch for Bug 14222. IIRC I sent a similar patch a year ago or so and never received feedback why it was rejected. Austin encouraged me to send it again, so please if it is unacceptable, let me know why. The patch will make RPC clients work that are written for Windows 2000 (and later), where the negotiate provider performs an automatic fallback to NTLM, if Kerberos is not possible, both in workgroup and domain scenarios. Currently, Wine doesn't have a negotiate and a Kerberos provider. This patch allows clients requesting Negotiate to get the NTLM provider and will still work, even if one day negotiate and Kerberos providers will be added to Wine. Cheers, -- Stefan

1 0

Re: comctl32/tests: Fix monthcal test on NT4
by Greg Geldorp Sept. 28, 2009

Sept. 28, 2009

Hi Nikolay, > Ge van Geldorp wrote: > > Changelog: > > comctl32/tests: Fix monthcal test on NT4 > Hi, Ge. > > This one interfere with a mine, sent yesterday: > > http://www.winehq.org/pipermail/wine-patches/2009-September/078955.html > > Could you wait a bit? I hope this could fix some failures. Sure. I'll resend after your patch is in (if my fix is still needed). Ge.

2 1

Re: d3d9: Don't release the parent device before destroying its children
by Henri Verbeet Sept. 28, 2009

Sept. 28, 2009

2009/9/28 Allan Tong <actong88(a)gmail.com>: > I assume this was done to avoid referencing freed memory, but I'm not > sure if there was any other reason for this. > Not referencing potentially freed memory was the only reason. > Note that this patch only changes d3d9. > Could you write one for d3d8 as well? Otherwise I'll send one tomorrow.

2 1

Re: ole32/tests: fix clipboard tests on Windows
by Alexandre Julliard Sept. 28, 2009

Sept. 28, 2009

"Ge van Geldorp" <ggeldorp(a)vmware.com> writes: > @@ -417,18 +417,23 @@ static HRESULT DataObjectImpl_CreateComplex(LPDATAOBJECT *lplpdataobj) > InitFormatEtc(obj->fmtetc[1], cf_stream, TYMED_ISTREAM); > InitFormatEtc(obj->fmtetc[2], cf_storage, TYMED_ISTORAGE); > InitFormatEtc(obj->fmtetc[3], cf_another, TYMED_ISTORAGE|TYMED_ISTREAM|TYMED_HGLOBAL); > - memset(&dm, 0, sizeof(dm)); > - dm.dmSize = sizeof(dm); > - dm.dmDriverExtra = 0; > - lstrcpyW(dm.dmDeviceName, device_name); > - obj->fmtetc[3].ptd = HeapAlloc(GetProcessHeap(), 0, FIELD_OFFSET(DVTARGETDEVICE, tdData) + sizeof(device_name) + dm.dmSize + dm.dmDriverExtra); > - obj->fmtetc[3].ptd->tdSize = FIELD_OFFSET(DVTARGETDEVICE, tdData) + sizeof(device_name) + dm.dmSize + dm.dmDriverExtra; > - obj->fmtetc[3].ptd->tdDriverNameOffset = FIELD_OFFSET(DVTARGETDEVICE, tdData); > - obj->fmtetc[3].ptd->tdDeviceNameOffset = 0; > - obj->fmtetc[3].ptd->tdPortNameOffset = 0; > - obj->fmtetc[3].ptd->tdExtDevmodeOffset = obj->fmtetc[3].ptd->tdDriverNameOffset + sizeof(device_name); > - lstrcpyW((WCHAR*)obj->fmtetc[3].ptd->tdData, device_name); > - memcpy(obj->fmtetc[3].ptd->tdData + sizeof(device_name), &dm, dm.dmSize + dm.dmDriverExtra); > + if (! broken(TRUE)) /* Windows messes up ptd when doing OleGetClipboard() */ > + { If that case is really unconditionally broken on Windows it doesn't make sense to test it at all. -- Alexandre Julliard julliard(a)winehq.org

1 0

Re: [PATCH 12/12] d3d10/tests: Add test_effect_constant_buffer_type test.
by Henri Verbeet Sept. 28, 2009

Sept. 28, 2009

2009/9/28 Rico Schüller <kgbricola(a)web.de>: > + ID3D10Effect* effect; I'd prefer it if that was written as "ID3D10Effect *effect;" > + ok(SUCCEEDED(hr), "ID3D10EffectType::GetDesc failed!\n"); Please print the actual error code as well. > + ok(type_desc.Elements == 0, "Elements is %i, expected 0\n", type_desc.Elements); > + ok(type_desc.Members == 2, "Members is %i, expected 2\n", type_desc.Members); > + ok(type_desc.Rows == 0, "Rows is %i, expected 0\n", type_desc.Rows); > + ok(type_desc.Columns == 0, "Columns is %i, expected 0\n", type_desc.Columns); These are unsigned variables. > + string = type->lpVtbl->GetMemberName(type, 0); > + ok(strcmp(string, "f1") == 0, "GetMemberName(0) is \"%s\", expected \"f1\"\n", string); You should probably use wine_dbgstr_a() for this.

1 0