August 2007 - Wine-Devel - list.winehq.org

Re: EVE Regression?

by Stefan Dösinger

Am Mittwoch, 1. August 2007 09:50 schrieb Stephan Rose: > Starting with wine 0.9.42 I continuously get the following error trying > to start EVE: > > fixme:d3d_draw:drawPrimitive Using software emulation because not all > material properties could be tracked > > And I just get a black screen for the login screen and it'll just sit > there forever like that. > > The error keeps being reported over and over until I close the window. > > Everything worked very well with 0.9.41 Do a regression test and report a bug: http://wiki.winehq.org/GitWine, 6.1. Regression testing Eve worked for me yesterday when I tried it. The fixme written is not an error, it is a sign of the proper fix for the lines in the starmap. EVE uses a lighting setup which cannot be set up in opengl, so we have to fall back to a slow rendering path, doing a sort of software emulation. I have added the rather noisy fixme to get a warning when other games show the same behavior. The problem can be fixed properly with a fixed function pipeline replacement shader, but this would be a lot of work. Specifically for eve, the lighting setup is equal to disabling lighting entirely. We could make use of that, but it would be a really nasty application specific hack.

Aug. 1, 2007

Re: dinput - Allow the use of the standard Joystick GUID when calling CreateDevice

by Peter Dons Tychsen

On Tue, 2007-07-31 at 20:05 -0600, Vitaliy Margolen wrote: > Peter Dons Tychsen wrote: > > On Tue, 2007-07-31 at 06:34 -0600, Vitaliy Margolen wrote: > >> Peter Dons Tychsen wrote: > >>> On Mon, 2007-07-30 at 21:20 +0200, Peter Dons Tychsen wrote: > >>>> On Sun, 2007-07-29 at 17:43 -0600, Vitaliy Margolen wrote: > >>>>> Peter Dons Tychsen wrote: > >>>>>> OK. Please review this diff: > >>>>>> > >>>>>> I will re-submit it if you like it. > >>>>>> > >>>>>> /Pedro > >>>>> Alright looks good now (you might want to remove the extra white space your > >>>>> patch adds - git complains about those). Just left to make the same change > >>>>> to the other joystick interface :) > >>>>> > >>>>> BTW don't forget to cc: the wine-devel list when you replying. > >>>>> > >>>>> Vitaliy. > >>>> OK. I still feel that trying to sync joystick_linuxinput.c and > >>>> joystick_linux.c is out of scope for this patch. But as my steadiness > >>>> has limits i have done some of it anyway... :). I have kept the changes > >>>> modest. > >>>> > >>>> Please take a look at the new patch. > >>>> > >>>> On another note i will agree with you that it would be practical if the > >>>> two files had similar (or maybe even shared) code. Redundant code is > >>>> never very practical as it always ends up as this very example. The two > >>>> implementations which were supposed to identical always drift apart. If > >>>> things are supposed to do the same, keep them together to begin with. > >>>> > >>>> Thanks, > >>>> > >>> Maybe i would be more successful if i actually attached the > >>> patch.... :-) > >>> > >>> Here it is.... > >> Just few small things left: > >> > >>> -#undef MAX_JOYSTICKS > >> As I mentioned earlier you should move this instead of removing it completely. > >> > >> > >>> static HRESULT joydev_create_deviceW(IDirectInputImpl *dinput, REFGUID rguid, REFIID riid, LPDIRECTINPUTDEVICEW* pdev) > >>> { > >>> - int i; > >>> - > >>> - find_joydevs(); > >> You removed a really important function call here. > >> > >> > >>> +static HRESULT joydev_create_deviceA(IDirectInputImpl *dinput, REFGUID rguid, REFIID riid, LPDIRECTINPUTDEVICEA* pdev) > >>> +{ > >>> + unsigned short index; > >>> + > >>> + if ((index = get_joystick_index(rguid)) < MAX_JOYDEV) { > >>> + if ((riid == NULL) || > >> If you reformat the whole function, please make it 4-space indented. > >> > >> Vitaliy. > > > > OK. Here is the new patch: > > > Looks good to me. The problem is, it doesn't work for me (USB joystick). I > think you forgot to look at how joystick_linuxinput.c handles that data3 > field. It's not just 0 based index... > > Vitaliy. Oh no! I told you i should bot have merged the changes over to joystick_linuxinput as they were too different... :-) OK, so how do i force my system to use the other implementation (so i can test it) ? I have a Logitech WingMan Extreme (old version). /pedro

Aug. 1, 2007

Re: WGL: fix wglChoosePixelFormatARB behavior

by Roderick Colenbrander

Ignore this patch. I'll write a better version. Roderick On Wednesday 01 August 2007 19:54, Roderick Colenbrander wrote: > Hi, > > According to the specs wglChoosePixelFormatARB returns the number of > matching pixelformats. This number can be different than the size of the > return buffer which is given by nFormats. > > This bug wasn't bad before but that was because we only offered 1 > pixelformat for onscreen and offscreen rendering. > > Regards, > Roderick Colenbrander

Aug. 1, 2007

Re: mshtml #2: Added HTMLWindow's IDispatch methods implementation.

by Jacek Caban

Hi Robert, Thanks for your review. Robert Shearman wrote: > Jacek Caban wrote: >> static HRESULT WINAPI HTMLWindow2_item(IHTMLWindow2 *iface, VARIANT >> *pvarIndex, VARIANT *pvarResult) >> diff --git a/dlls/mshtml/main.c b/dlls/mshtml/main.c >> index 3f8131c..41f32eb 100644 >> --- a/dlls/mshtml/main.c >> +++ b/dlls/mshtml/main.c >> @@ -51,10 +51,52 @@ DWORD mshtml_tls = 0; >> >> static HINSTANCE shdoclc = NULL; >> >> +static ITypeLib *typelib; >> +static ITypeInfo *typeinfos[LAST_tid]; >> + >> +static const REFIID tid_ids[] = { >> + &IID_IHTMLWindow2 >> +}; >> + >> +HRESULT get_typeinfo(enum tid_t tid, ITypeInfo **typeinfo) >> +{ >> + HRESULT hres; >> + >> + if(!typelib) { >> + hres = LoadRegTypeLib(&LIBID_MSHTML, 4, 0, >> LOCALE_SYSTEM_DEFAULT, &typelib); >> > > This isn't thread-safe. I don't consider it really important here ATM. I haven't seen any app that would benefit from thread safe MSHTML. Anyway I will fix it. >> + if(FAILED(hres)) { >> + ERR("LoadRegTypeLib failed: %08x\n", hres); >> + return hres; >> + } >> + } >> + >> + if(!typeinfos[tid]) { >> + hres = ITypeLib_GetTypeInfoOfGuid(typelib, tid_ids[tid], >> typeinfos+tid); >> + if(FAILED(hres)) { >> + ERR("GetTypeInfoOfGuid failed: %08x\n", hres); >> + return hres; >> + } >> + } >> > > Why not just use the IID instead of an enumeration here? I know you're > trying to cache the typeinfo here, but the current implementation of > typelibs makes ITypeLib_GetTypeInfoOfGuid very cheap. It's surely not cheaper than a simple table element check. It's a part of code that is extremely profiled in Gecko (that is its counterpart in Gecko). We don't have to take that much care of it, but I think such simple caching is wort the efforts. >> + >> + *typeinfo = typeinfos[tid]; >> + return S_OK; >> +} >> + >> static void thread_detach(void) >> { >> - thread_data_t *thread_data = get_thread_data(FALSE); >> + thread_data_t *thread_data; >> + >> + if(typelib) { >> + unsigned i; >> + >> + for(i=0; i < sizeof(typeinfos)/sizeof(*typeinfos); i++) >> + if(typeinfos[i]) >> + ITypeInfo_Release(typeinfos[i]); >> + >> + ITypeLib_Release(typelib); >> + } >> > > Why are you doing this from DLL_THREAD_DETACH and not DLL_PROCESS_DETACH? Ops, that's my fatal mistake. Thanks, Jacek

Aug. 1, 2007

Re: New winetest skipping test for nonexistent dll's

by Jacek Caban

Hi Paul, Paul Vriens wrote: > + > + dll = LoadLibrary(dllname); > + if (!dll) { I think it would be better to call LoadLibraryEx(dllname, LOAD_LIBRARY_AS_DATA_FILE) here to avoid executing DllMain code. Jacek

Aug. 1, 2007

Re: Add winhttp.dll.

by Jacek Caban

Hi Vijay, Vijay Kiran Kamuju wrote: > Hi Jacek, > > I think you can use the header I sent a while back to wine-patches for this. > http://www.winehq.org/pipermail/wine-patches/2005-September/020901.html > The original patch: > http://www.winehq.org/pipermail/wine-patches/attachments/20050923/790dfab2/… > This might require some re-work. > It was rejected at that time, when I wanted to add winhttp.dll > It might be of some help to you. > Thanks, that seems useful. Though I have no plans to work on it ATM. Stub dll is enough for the app I'm working on. BTW this dll looks like a big code duplication with wininet. Jacek

Aug. 1, 2007

New winetest skipping test for nonexistent dll's

by Paul Vriens

Hi, I've been busy trying to come up with a solution and attached you will find 2 patches. One for 'winetest that will make sure we have specific output for dll's that are not on the system. I have made the output as close to normal test output as possible, so we don't have to patch dissect (the tool that dissects the reports). One for 'gather', the tool that translates the reports into a nice html page. There will be (in contrast with the other skip's) no visibility in the group reports yet. I'll have to figure out how to do that nicely. For now I'm assuming that if one system doesn't have a dll, all the others (same platform) don't have it at well (I know bad assumption and it will be dealt with in next revisions). (I do have another version of 'gather' which deals with the group reports but shows the whole test for all systems (same platform of course) if one or more systems have a missing dll). If we pursue this then we probably also need to change old tests as that will mean we can strip a lot of code from the tests. Comments/remarks etc.. are welcome. Cheers, Paul. diff --git a/programs/winetest/main.c b/programs/winetest/main.c index f743ffc..3c4846d 100644 --- a/programs/winetest/main.c +++ b/programs/winetest/main.c @@ -433,6 +433,24 @@ extract_test_proc (HMODULE hModule, LPCTSTR lpszType, LPTSTR lpszName, LONG_PTR lParam) { const char *tempdir = (const char *)lParam; + char dllname[MAX_PATH]; + HMODULE dll; + + /* Check if the main dll is present on this system */ + strcpy(dllname, lpszName); + *strstr(dllname, "_TEST.EXE") = 0; + + dll = LoadLibrary(dllname); + if (!dll) { + /* Here we add some information to the results file + * that indicates that the dll is not present + * on this platform. + */ + report (R_WARNING, "dll : %s is not present", dllname); + return TRUE; + } + FreeLibrary(dll); + get_subtests( tempdir, &wine_tests[nr_of_files], lpszName ); nr_of_tests += wine_tests[nr_of_files].subtest_count; nr_of_files++; Index: gather =================================================================== RCS file: /home/wine/tools/winetest/gather,v retrieving revision 1.20 diff -u -r1.20 gather --- gather 13 Mar 2007 16:10:46 -0000 1.20 +++ gather 1 Aug 2007 16:27:19 -0000 @@ -96,6 +96,9 @@ my ($digest, $unit, $test, $count, $todo, $error, $skipped, $source, $rev) = split; my $testname = "$unit:$test"; + if ($test =~ /_dll_missing/) { + $testref->{dll}->{$unit} = "missing"; + } $testref->{results}->{$testname} = [$count, $todo, $error, $skipped]; $alltests{$testname} = "http://cvs.winehq.org/cvsweb/wine/$source" . ($rev ne "-"?"#rev$rev":"") unless exists $alltests{$testname}; @@ -199,7 +202,18 @@ my ($test, $testname, $groupname) = @_; if (!exists $test->{results}->{$testname}) { - print OUT " <td class=\"note\">.</td>\n"; + my ($unit1, $test1) = split(/:/, $testname); + my $dll = $test->{dll}->{$unit1}; + if ((defined $dll) && ($dll eq "missing")) { + print OUT <<"EOF"; + <td class="skip_pass"><a + title="No tests run as $unit1.dll is not present on this system" + onMouseOver="No tests run as $unit1.dll is not present on this system" + >0</a></td> +EOF + } else { + print OUT " <td class=\"note\">.</td>\n"; + } } else { my $file = "$test->{dir}/$testname.txt"; my ($count, $todo, $error, $skipped) = @{$test->{results}->{$testname}}; @@ -223,6 +237,8 @@ } foreach my $testname (sort keys %alltests) { + next unless $testname !~ /_dll_missing/; + print OUT <<"EOF"; <tr> <td class="test"> @@ -290,6 +306,7 @@ <tbody onDblClick="clone();"> EOF foreach my $testname (sort keys %alltests) { # skip identical + next unless $testname !~ /_dll_missing/; my $digest = $group->{digests}->{$testname}; next unless defined $digest && $digest eq "differ"; print OUT <<"EOF";

Aug. 1, 2007

Re: Patch: VarWeekdayName

by Jiří Paleček

On Tue, 31 Jul 2007 17:38:18 +0200, Juan Lang <juan.lang(a)gmail.com> wrote: > Hi Jiri, > >> I want to ask, is anything wrong with this patch? > > Your indenting is haphazard. Please match the indenting in the rest > of the file, which in this case is no tabs, 2-space indenting. Also, > curly braces go on a new line, so e.g.: > + if (!size) { > should be > + if (!size) > + { > instead. Thank you. Hope it will be better this time. Also, I'll send a test if this is OK. Changelog Implement the VarWeekdayName function in oleaut32.dll (fixes 5884) Regards Jiri Palecek

Aug. 1, 2007

Re: RPC questions

by Robert Shearman

Kai Blin wrote: > Hi folks, > > I will be getting into working on RPC soonish, and some of that I'd like to > use for Wine. However, I don't know my way around that code so far, what > would be a good starting point to tackle this? > > My goal is to set up a daemon that will take care of remote named pipes over > smb and all of that stuff, while all that Wine needs to do is to access this > daemon talking DCE-RPC. As we probably don't need to link anything, we will > be able to use that daemon for other jobs, like doing netapi32 calls, etc. > > Where would I get started for this? The code for RPC transports (like SMB and TCP) is in dlls/rpcrt4/rpc_transport.c. It should be fairly easy to plug in an SMB-only transport for testing purposes by reimplementing the existing named pipe functions in that file, but switching between the two depending on whether the network address is local or remote is more challenging. -- Rob Shearman

Aug. 1, 2007

Re: mshtml #2: Added HTMLWindow's IDispatch methods implementation.

by Robert Shearman

Jacek Caban wrote: > static HRESULT WINAPI HTMLWindow2_item(IHTMLWindow2 *iface, VARIANT *pvarIndex, VARIANT *pvarResult) > diff --git a/dlls/mshtml/main.c b/dlls/mshtml/main.c > index 3f8131c..41f32eb 100644 > --- a/dlls/mshtml/main.c > +++ b/dlls/mshtml/main.c > @@ -51,10 +51,52 @@ DWORD mshtml_tls = 0; > > static HINSTANCE shdoclc = NULL; > > +static ITypeLib *typelib; > +static ITypeInfo *typeinfos[LAST_tid]; > + > +static const REFIID tid_ids[] = { > + &IID_IHTMLWindow2 > +}; > + > +HRESULT get_typeinfo(enum tid_t tid, ITypeInfo **typeinfo) > +{ > + HRESULT hres; > + > + if(!typelib) { > + hres = LoadRegTypeLib(&LIBID_MSHTML, 4, 0, LOCALE_SYSTEM_DEFAULT, &typelib); > This isn't thread-safe. > + if(FAILED(hres)) { > + ERR("LoadRegTypeLib failed: %08x\n", hres); > + return hres; > + } > + } > + > + if(!typeinfos[tid]) { > + hres = ITypeLib_GetTypeInfoOfGuid(typelib, tid_ids[tid], typeinfos+tid); > + if(FAILED(hres)) { > + ERR("GetTypeInfoOfGuid failed: %08x\n", hres); > + return hres; > + } > + } > Why not just use the IID instead of an enumeration here? I know you're trying to cache the typeinfo here, but the current implementation of typelibs makes ITypeLib_GetTypeInfoOfGuid very cheap. > + > + *typeinfo = typeinfos[tid]; > + return S_OK; > +} > + > static void thread_detach(void) > { > - thread_data_t *thread_data = get_thread_data(FALSE); > + thread_data_t *thread_data; > + > + if(typelib) { > + unsigned i; > + > + for(i=0; i < sizeof(typeinfos)/sizeof(*typeinfos); i++) > + if(typeinfos[i]) > + ITypeInfo_Release(typeinfos[i]); > + > + ITypeLib_Release(typelib); > + } > Why are you doing this from DLL_THREAD_DETACH and not DLL_PROCESS_DETACH? -- Rob Shearman

Aug. 1, 2007