February 2017 - Wine-devel - list.winehq.org

Re: [v2 PATCH] dxgi: Added support for DXGI_SWAP_CHAIN_FLAG_GDI_COMPATIBLE
by Marvin 08 Feb '17

08 Feb '17

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=28438 Your paranoid android. === w864 (64 bit device) === device.c:1497: Test failed: SetFullscreenState failed, hr 0x887a0022. device.c:1498: Test failed: Got window rect (0,0)-(400,200), expected (0,0)-(800,600). device.c:1498: Test failed: Got client rect (0,0)-(394,171), expected (0,0)-(800,600). device.c:1498: Test failed: Got monitor rect (0,0)-(1024,768), expected (0,0)-(800,600). device.c:1498: Test failed: Got windowed 0x1, expected 0. device.c:1498: Test failed: Got fullscreen 0, expected 0x1.

1 0

Re: [PATCH v2 1/6] wined3d: Fix querying texture-related limits on core profile.
by Henri Verbeet 08 Feb '17

08 Feb '17

On 7 February 2017 at 22:08, Matteo Bruni <mbruni(a)codeweavers.com> wrote: > diff --git a/dlls/wined3d/directx.c b/dlls/wined3d/directx.c > index c47052c..bfb018e 100644 > --- a/dlls/wined3d/directx.c > +++ b/dlls/wined3d/directx.c > @@ -3405,24 +3405,35 @@ static void wined3d_adapter_init_limits(struct wined3d_gl_info *gl_info) > } > if (gl_info->supported[ARB_MULTITEXTURE]) > { > - gl_info->gl_ops.gl.p_glGetIntegerv(GL_MAX_TEXTURE_UNITS_ARB, &gl_max); > - gl_info->limits.textures = min(MAX_TEXTURES, gl_max); > - TRACE("Max textures: %d.\n", gl_info->limits.textures); > + if (gl_info->supported[WINED3D_GL_LEGACY_CONTEXT]) > + { > + gl_info->gl_ops.gl.p_glGetIntegerv(GL_MAX_TEXTURE_UNITS_ARB, &gl_max); > + gl_info->limits.textures = min(MAX_TEXTURES, gl_max); > + TRACE("Max textures: %d.\n", gl_info->limits.textures); > > - if (gl_info->supported[ARB_FRAGMENT_PROGRAM]) > + if (gl_info->supported[ARB_FRAGMENT_PROGRAM]) > + { > + gl_info->gl_ops.gl.p_glGetIntegerv(GL_MAX_TEXTURE_COORDS_ARB, &gl_max); > + gl_info->limits.texture_coords = min(MAX_TEXTURES, gl_max); > + } > + else > + { > + gl_info->limits.texture_coords = max(gl_info->limits.texture_coords, gl_max); > + } > + TRACE("Max texture coords: %d.\n", gl_info->limits.texture_coords); > + } > + > + if (gl_info->supported[ARB_FRAGMENT_PROGRAM] || gl_info->supported[ARB_FRAGMENT_SHADER]) > { > GLint tmp; > - gl_info->gl_ops.gl.p_glGetIntegerv(GL_MAX_TEXTURE_COORDS_ARB, &gl_max); > - gl_info->limits.texture_coords = min(MAX_TEXTURES, gl_max); > + > gl_info->gl_ops.gl.p_glGetIntegerv(GL_MAX_TEXTURE_IMAGE_UNITS_ARB, &tmp); > gl_info->limits.fragment_samplers = min(MAX_FRAGMENT_SAMPLERS, tmp); > } > else > { > - gl_info->limits.texture_coords = max(gl_info->limits.texture_coords, gl_max); > gl_info->limits.fragment_samplers = max(gl_info->limits.fragment_samplers, gl_max); > } > - TRACE("Max texture coords: %d.\n", gl_info->limits.texture_coords); > TRACE("Max fragment samplers: %d.\n", gl_info->limits.fragment_samplers); > > if (gl_info->supported[ARB_VERTEX_SHADER]) This still doesn't really look logical to me.

2 2

Re: [v2 PATCH] dxgi: Added support for DXGI_SWAP_CHAIN_FLAG_GDI_COMPATIBLE
by Marvin 08 Feb '17

08 Feb '17

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=28427 Your paranoid android. === build (build) === Patch failed to apply

1 0

Re: [PATCH] compobj.dll16: Implement CoGetClassObject().
by Huw Davies 08 Feb '17

08 Feb '17

On Sat, Feb 04, 2017 at 07:47:32PM -0600, Zebediah Figura wrote: > Fixes bug https://bugs.winehq.org/show_bug.cgi?id=41209 > > Signed-off-by: Zebediah Figura <z.figura12(a)gmail.com> > --- > dlls/compobj.dll16/compobj.c | 242 +++++++++++++++++++++++++++++++--- > dlls/compobj.dll16/compobj.dll16.spec | 2 +- > 2 files changed, 222 insertions(+), 22 deletions(-) > Hi Zebediah, This patch could be split into a series of smaller patches. For example, implement CoGetClassObject16() first, then CoCreateInstance16() as a seperate patch, and then CoFreeAllLibraries(). I've inlined some comments below. > diff --git a/dlls/compobj.dll16/compobj.c b/dlls/compobj.dll16/compobj.c > index b934a06..1b100c3 100644 > --- a/dlls/compobj.dll16/compobj.c > +++ b/dlls/compobj.dll16/compobj.c > @@ -30,6 +30,8 @@ > #include <string.h> > #include <assert.h> > > +#define COBJMACROS > + > #include "windef.h" > #include "winbase.h" > #include "winuser.h" > @@ -42,6 +44,7 @@ > #include "wtypes.h" > #include "wine/unicode.h" > #include "wine/winbase16.h" > +#include "wine/list.h" > > #include "wine/debug.h" > > @@ -98,6 +101,115 @@ static inline IMalloc16Impl *impl_from_IMalloc16(IMalloc16 *iface) > return CONTAINING_RECORD(iface, IMalloc16Impl, IMalloc16_iface); > } > > +typedef HRESULT (CALLBACK *DllGetClassObjectFunc)(REFCLSID clsid, REFIID iid, LPVOID *ppv); > +typedef HRESULT (WINAPI *DllCanUnloadNowFunc)(void); > + > +typedef struct tagOpenDll > +{ > + LONG refs; > + LPSTR library_name; > + HMODULE16 library; > + DllGetClassObjectFunc DllGetClassObject; > + DllCanUnloadNowFunc DllCanUnloadNow; > + struct list entry; > +} OpenDll; Why not call the typedef open_dll and drop the tag name - it's not used. Or better yet, ditch the typedef and just use struct open_dll throughout. > +static struct list openDllList = LIST_INIT(openDllList); Similarly rename to open_dll_list. > + > +/***************************************************************************** > + * This section contains OpenDllList implementation > + */ > + > +static OpenDll *COMPOBJ16_DllList_Get(LPCSTR library_name) > +{ > + OpenDll *ptr; > + LIST_FOR_EACH_ENTRY(ptr, &openDllList, OpenDll, entry) > + { > + if (!strcasecmp(library_name, ptr->library_name) && > + ++(ptr->refs) != 1) /* entry is being destroy if == 1 */ This reference handling looks odd - at the very least you're leaving a previously zero ref count at one. Also, how about naming this function (and others similarly below) dll_list_get()? > +/* pass FALSE for free_entry to release a reference without destroying the > + * entry if it reaches zero or TRUE otherwise */ > +static void COMPOBJ16_DllList_ReleaseRef(OpenDll *entry) > +{ > + if (--(entry->refs) == 0) > + { > + list_remove(&entry->entry); > + > + TRACE("freeing %x\n", entry->library); > + FreeLibrary16(entry->library); > + > + HeapFree(GetProcessHeap(), 0, entry->library_name); > + HeapFree(GetProcessHeap(), 0, entry); > + } > +} > + > /****************************************************************************** > * IMalloc16_QueryInterface [COMPOBJ.500] > */ > @@ -279,13 +391,34 @@ HRESULT WINAPI CoCreateStandardMalloc16(DWORD dwMemContext, > return S_OK; > } > > +/*********************************************************************** > + * CoFreeAllLibraries [COMPOBJ.12] > + */ > +void WINAPI CoFreeAllLibraries16(void) > +{ > + OpenDll *entry, *next; > + > + TRACE("()\n"); > + LIST_FOR_EACH_ENTRY_SAFE(entry, next, &openDllList, OpenDll, entry) > + { > + list_remove(&entry->entry); > + > + TRACE("freeing %x\n", entry->library); > + FreeLibrary16(entry->library); > + > + HeapFree(GetProcessHeap(), 0, entry->library_name); > + HeapFree(GetProcessHeap(), 0, entry); > + } > +} > + This code and the code in COMPOBJ16_DllList_ReleaseRef are essentially the same, it should be possible to avoid this duplication. Huw.

2 1

Re: [PATCH] kernel32/nls: Added Bangla locale data
by Nikolay Sivov 08 Feb '17

08 Feb '17

Wrong sublang here, please skip this one.

1 0

Re: [PATCH] winex11: Pass correct buffer length to RegQueryValueExW()
by Ken Thomases 08 Feb '17

08 Feb '17

On Feb 8, 2017, at 5:46 AM, Nikolay Sivov <nsivov(a)codeweavers.com> wrote: > > Signed-off-by: Nikolay Sivov <nsivov(a)codeweavers.com> > --- > > Similar change could also be applied to Mac driver. Yeah, I had a patch in my working tree since I saw your original submission, but then code freeze. :) -Ken

1 0

Re: oleaut32: Make the heap allocation functions static.
by Alexandre Julliard 08 Feb '17

08 Feb '17

Francois Gouget <fgouget(a)free.fr> writes: > Signed-off-by: Francois Gouget <fgouget(a)free.fr> > --- > > These were supposed to see more widespread usage but it has not happened > in the past 3 years. I blame Michael ;-) If you really want to change them, I'd suggest to do it the way it's done in most other places, as static inline functions in the header. -- Alexandre Julliard julliard(a)winehq.org

1 0

Re: [PATCH 4/6] wined3d: Implement compute pipeline UAV binding.
by Henri Verbeet 08 Feb '17

08 Feb '17

On 8 February 2017 at 11:54, Józef Kucia <jkucia(a)codeweavers.com> wrote: > diff --git a/dlls/wined3d/cs.c b/dlls/wined3d/cs.c > index 422f3f1..f2e10a8 100644 > --- a/dlls/wined3d/cs.c > +++ b/dlls/wined3d/cs.c > @@ -44,6 +44,7 @@ enum wined3d_cs_op > WINED3D_CS_OP_SET_TEXTURE, > WINED3D_CS_OP_SET_SHADER_RESOURCE_VIEW, > WINED3D_CS_OP_SET_UNORDERED_ACCESS_VIEW, > + WINED3D_CS_OP_SET_COMPUTE_UAV, It's a little undesirable to introduce multiple operations that are largely the same, although I don't necessarily see a good way to avoid it either right now. > @@ -1053,6 +1054,35 @@ static void wined3d_cs_exec_set_unordered_access_view(struct wined3d_cs *cs, con > device_invalidate_state(cs->device, STATE_UNORDERED_ACCESS_VIEW_BINDING); > } > > +void wined3d_cs_emit_set_compute_unordered_access_view(struct wined3d_cs *cs, unsigned int view_idx, > + struct wined3d_unordered_access_view *view) > +{ > + struct wined3d_cs_set_unordered_access_view *op; > + > + op = cs->ops->require_space(cs, sizeof(*op)); > + op->opcode = WINED3D_CS_OP_SET_COMPUTE_UAV; > + op->view_idx = view_idx; > + op->view = view; > + > + cs->ops->submit(cs); > +} > + > +static void wined3d_cs_exec_set_compute_unordered_access_view(struct wined3d_cs *cs, const void *data) > +{ > + const struct wined3d_cs_set_unordered_access_view *op = data; > + struct wined3d_unordered_access_view *prev; > + > + prev = cs->state.compute_unordered_access_view[op->view_idx]; > + cs->state.compute_unordered_access_view[op->view_idx] = op->view; > + > + if (op->view) > + InterlockedIncrement(&op->view->resource->bind_count); > + if (prev) > + InterlockedDecrement(&prev->resource->bind_count); > + > + device_invalidate_state(cs->device, STATE_COMPUTE_UNORDERED_ACCESS_VIEW_BINDING); > +} I should have noticed this when WINED3D_CS_OP_SET_UNORDERED_ACCESS_VIEW was introduced, but these are between wined3d_cs_exec_set_shader_resource_view() and wined3d_cs_emit_set_shader_resource_view(), please don't do that. More generally, all the other handlers have execution above emission, I think it's best to stick with that.

1 0

Re: [PATCH v2 2/6] d3d11/tests: Add test for interleaved Draw() and Dispatch() calls.
by Henri Verbeet 08 Feb '17

08 Feb '17

On 8 February 2017 at 11:54, Józef Kucia <jkucia(a)codeweavers.com> wrote: > Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> > --- > > Version 2: Remove checks for backbuffer content. > > --- > dlls/d3d11/tests/d3d11.c | 182 +++++++++++++++++++++++++++++++++++++++++++++++ > 1 file changed, 182 insertions(+) > This is taking on the order of 10 seconds per draw/dispatch pair, or 30 seconds for the entire test, for me on Windows. That's still manageable, but I suspect that's longer than intended, and could at some point cause the test to timeout during automated runs.

1 0

Re: [PATCH 3/3] dxgi: Added support for DXGI_SWAP_CHAIN_FLAG_GDI_COMPATIBLE
by Henri Verbeet 08 Feb '17

08 Feb '17

On 8 February 2017 at 09:26, Nikolay Sivov <nsivov(a)codeweavers.com> wrote: > @@ -5275,7 +5275,7 @@ static HRESULT CDECL device_parent_sub_resource_created(struct wined3d_device_pa > } > > static HRESULT CDECL device_parent_create_swapchain_texture(struct wined3d_device_parent *device_parent, > - void *container_parent, const struct wined3d_resource_desc *wined3d_desc, > + void *container_parent, const struct wined3d_resource_desc *wined3d_desc, DWORD texture_flags, > struct wined3d_texture **wined3d_texture) > { > struct d3d_device *device = device_from_wined3d_device_parent(device_parent); > @@ -5299,7 +5299,7 @@ static HRESULT CDECL device_parent_create_swapchain_texture(struct wined3d_devic > desc.Usage = D3D10_USAGE_DEFAULT; > desc.BindFlags = D3D10_BIND_RENDER_TARGET; > desc.CPUAccessFlags = 0; > - desc.MiscFlags = 0; > + desc.MiscFlags = texture_flags & WINED3D_TEXTURE_CREATE_GET_DC ? D3D10_RESOURCE_MISC_GDI_COMPATIBLE: 0; Please also print a FIXME for other flags. > diff --git a/dlls/d3d8/device.c b/dlls/d3d8/device.c > index 805694b853..bf6653a42b 100644 > --- a/dlls/d3d8/device.c > +++ b/dlls/d3d8/device.c > @@ -3036,7 +3036,8 @@ static HRESULT CDECL device_parent_volume_created(struct wined3d_device_parent * > } > > static HRESULT CDECL device_parent_create_swapchain_texture(struct wined3d_device_parent *device_parent, > - void *container_parent, const struct wined3d_resource_desc *desc, struct wined3d_texture **texture) > + void *container_parent, const struct wined3d_resource_desc *desc, DWORD texture_flags, > + struct wined3d_texture **texture) > { > struct d3d8_device *device = device_from_device_parent(device_parent); > struct d3d8_surface *d3d_surface; Likewise, I think we should do a little more here. Either passing the texture flags along to wined3d_texture_create(), or just printing a FIXME in case they're non-zero. > diff --git a/dlls/d3d9/device.c b/dlls/d3d9/device.c > index e646cb76c7..923ced55b4 100644 > --- a/dlls/d3d9/device.c > +++ b/dlls/d3d9/device.c > @@ -3746,7 +3746,8 @@ static HRESULT CDECL device_parent_volume_created(struct wined3d_device_parent * > } > > static HRESULT CDECL device_parent_create_swapchain_texture(struct wined3d_device_parent *device_parent, > - void *container_parent, const struct wined3d_resource_desc *desc, struct wined3d_texture **texture) > + void *container_parent, const struct wined3d_resource_desc *desc, DWORD texture_flags, > + struct wined3d_texture **texture) > { > struct d3d9_device *device = device_from_device_parent(device_parent); > DWORD flags = WINED3D_TEXTURE_CREATE_MAPPABLE; I could live with adding WINED3D_TEXTURE_CREATE_GET_DC in device_parent_create_swapchain_texture() in d3d9, but ideally we'd set the appropriate flag on swapchain creation.

1 0