September 2018 - Wine-Devel

Re: [PATCH 2/2] user32/listbox: Optimize SetCount partly

by Gabriel Ivăncescu

On Wed, Sep 19, 2018 at 4:53 PM, Huw Davies <huw(a)codeweavers.com> wrote: > > I think implementing LBS_NODATA should come before adding LB_SETCOUNT > code that will need to be changed again. Also, adding a test to > show that LB_SETCOUNT fails if the style is LBS_OWNERDRAWFIXED, for > example, wouldn't hurt. > > Huw. > To be honest, we should just fail SetCount if LBS_NODATA is not set, but maybe later not now, if tests prove this is Windows' behavior, which I'll check in a bit. Even if Wine doesn't support it, it still "works" currently (but very slow and suboptimal, there's some bugs around about it), so I think it should be left with this wrong behavior for now. BTW I think you meant OWNERDRAWVARIABLE right? LBS_OWNERDRAWFIXED should work (and is actually required) for LBS_NODATA (and thus, LB_SETCOUNT), or I'm missing something. I guess for now this patchset should be rejected until LBS_NODATA is implemented which would require a totally different patchset anyway, or I find some other behavior with tests.

Sept. 19, 2018

Re: [PATCH 1/2] comctl32/listbox: Fix InitStorage heap extension

by Gabriel Ivăncescu

On Wed, Sep 19, 2018 at 5:06 PM, Huw Davies <huw(a)codeweavers.com> wrote: > > I dislike having the calculation split up like this. > > Further, I think we can insist that LB_ARRAY_GRANULARITY is a power of two, > (perhaps add a comment to that effect above its #define) which should > simplify things a bit. > > Huw. > Yeah, if it's a power of 2 I can use a bitwise AND operator to simplify it. I'll do that and add a comment.

Sept. 19, 2018

[PATCH 2/2] winhttp: Set last error in WinHttpQueryAuthSchemes when no auth scheme is found.

by Jacek Caban

Signed-off-by: Jacek Caban <jacek(a)codeweavers.com> --- dlls/winhttp/request.c | 1 + dlls/winhttp/tests/winhttp.c | 4 ++-- 2 files changed, 3 insertions(+), 2 deletions(-)

Sept. 19, 2018

[PATCH 1/2] winhttp: Modify index only if query_headers succeeded.

by Jacek Caban

Signed-off-by: Jacek Caban <jacek(a)codeweavers.com> --- dlls/winhttp/request.c | 3 +-- dlls/winhttp/tests/winhttp.c | 39 ++++++++++++++++++++++++++++++++++++++- 2 files changed, 39 insertions(+), 3 deletions(-)

Sept. 19, 2018

Re: [PATCH 3/4] msi/tests: Also test full range of properties for MsiSummaryInfoSetPropertyW.

by Francois Gouget

On Wed, 19 Sep 2018, Hans Leidekker wrote: [...] > This is causing new test failures. The testbot fails recognize them > however, probably because they come from custom.c instead of the test > module that was invoked, install.c. It does recognize them. They are even bolded orange since they are new failures and the summary at the top says Failures 4. https://testbot.winehq.org/JobDetails.pl?Key=42225#k1001 But yes, the TestBot is a bit confused because the errors don't come from a known test unit which is why it issues an extra error: msi:install contains a misplaced failure message for custom The problem is there is no 'custom' test unit: * You cannot run 'msi_test.exe custom'. * There is no "xxx:custom: 123 tests executed...' summary line. * Nothing ties 'custom' error messages to any real test unit such as 'install', 'action', etc. So when parsing the WineTest results there is no way to know where these messages are coming from. I don't know how to improve handling of these cases. * The simplest for me would be if they issued 'install' errors instead but that would be confusing since the error does not come from install.c. So that's out. * Maybe add a new type of line to tie custom messages to the install test unit? 0b40:install: Loading custom sub test unit. -- Francois Gouget <fgouget(a)codeweavers.com>

Sept. 19, 2018

Re: [PATCH 1/2] comctl32/listbox: Fix InitStorage heap extension

by Huw Davies

On 18 Sep 2018, at 21:10, Gabriel Ivăncescu <gabrielopcode(a)gmail.com> wrote: > > Signed-off-by: Gabriel Ivăncescu <gabrielopcode(a)gmail.com> > --- > > Only increase the item array if we actually have to. Previously, sending 1 to > nb_items repeatedly would always increase the array by LB_ARRAY_GRANULARITY, > even if there was plenty of space. > > dlls/comctl32/listbox.c | 18 +++++++++++------- > 1 file changed, 11 insertions(+), 7 deletions(-) > > diff --git a/dlls/comctl32/listbox.c b/dlls/comctl32/listbox.c > index 2137ef8..9cd89ff 100644 > --- a/dlls/comctl32/listbox.c > +++ b/dlls/comctl32/listbox.c > @@ -673,16 +673,20 @@ static LRESULT LISTBOX_InitStorage( LB_DESCR *descr, INT nb_items ) > { > LB_ITEMDATA *item; > > - nb_items += LB_ARRAY_GRANULARITY - 1; > - nb_items -= (nb_items % LB_ARRAY_GRANULARITY); > if (descr->items) { > - nb_items += HeapSize( GetProcessHeap(), 0, descr->items ) / sizeof(*item); > - item = HeapReAlloc( GetProcessHeap(), 0, descr->items, > - nb_items * sizeof(LB_ITEMDATA)); > + nb_items += descr->nb_items; > + if (nb_items > HeapSize(GetProcessHeap(), 0, descr->items) / sizeof(*item)) > + { > + UINT n = nb_items + LB_ARRAY_GRANULARITY - 1; > + item = HeapReAlloc(GetProcessHeap(), 0, descr->items, > + (n - n % LB_ARRAY_GRANULARITY) * sizeof(*item)); I dislike having the calculation split up like this. Further, I think we can insist that LB_ARRAY_GRANULARITY is a power of two, (perhaps add a comment to that effect above its #define) which should simplify things a bit. Huw.

Sept. 19, 2018

Re: [PATCH] d2d1: Implement CreateBitmapFromDxgiSurface().

by Henri Verbeet

Signed-off-by: Henri Verbeet <hverbeet(a)codeweavers.com>

Sept. 19, 2018

Re: [PATCH 2/2] user32/listbox: Optimize SetCount partly

by Huw Davies

> On 19 Sep 2018, at 12:46, Gabriel Ivăncescu <gabrielopcode(a)gmail.com> wrote: > > On Wed, Sep 19, 2018 at 1:55 PM, Huw Davies <huw(a)codeweavers.com> wrote: >> >> I'm also concerned that we won't get WM_MEASUREITEMs in the >> LBS_OWNERDRAWVARIABLE case. >> >> Huw. > > I don't know if that's an issue since SetCount should theoretically be > an "instant" operation and these items are just added at the end as > empty. I think Windows doesn't even support SetCount without > LBS_NODATA (which is not supported by Wine yet, I think I'll try > working on it and implement it in an optimal way, i.e. with the > minimum amount of memory needed, since it's meant for very very large > lists). > > But, at least for now, I can fall back to the original (unoptimal) > behavior if OWNERDRAWVARIABLE is set, would that be acceptable? I think implementing LBS_NODATA should come before adding LB_SETCOUNT code that will need to be changed again. Also, adding a test to show that LB_SETCOUNT fails if the style is LBS_OWNERDRAWFIXED, for example, wouldn't hurt. Huw.

Sept. 19, 2018

[PATCH] d2d1: Implement CreateBitmapFromDxgiSurface().

by Nikolay Sivov

Signed-off-by: Nikolay Sivov <nsivov(a)codeweavers.com> --- dlls/d2d1/device.c | 28 ++++++++++++++++++++++++++-- 1 file changed, 26 insertions(+), 2 deletions(-) diff --git a/dlls/d2d1/device.c b/dlls/d2d1/device.c index d779fb0a97..d20377b5c2 100644 --- a/dlls/d2d1/device.c +++ b/dlls/d2d1/device.c @@ -1825,9 +1825,33 @@ static HRESULT STDMETHODCALLTYPE d2d_device_context_CreateColorContextFromWicCol static HRESULT STDMETHODCALLTYPE d2d_device_context_CreateBitmapFromDxgiSurface(ID2D1DeviceContext *iface, IDXGISurface *surface, const D2D1_BITMAP_PROPERTIES1 *desc, ID2D1Bitmap1 **bitmap) { - FIXME("iface %p, surface %p, desc %p, bitmap %p stub!\n", iface, surface, desc, bitmap); + struct d2d_device_context *context = impl_from_ID2D1DeviceContext(iface); + D2D1_BITMAP_PROPERTIES1 bitmap_desc; + struct d2d_bitmap *object; + HRESULT hr; - return E_NOTIMPL; + TRACE("iface %p, surface %p, desc %p, bitmap %p.\n", iface, surface, desc, bitmap); + + if (!desc) + { + DXGI_SURFACE_DESC surface_desc; + + if (FAILED(hr = IDXGISurface_GetDesc(surface, &surface_desc))) + { + WARN("Failed to get surface desc, hr %#x.\n", hr); + return hr; + } + + memset(&bitmap_desc, 0, sizeof(bitmap_desc)); + bitmap_desc.pixelFormat.format = surface_desc.Format; + bitmap_desc.bitmapOptions = D2D1_BITMAP_OPTIONS_TARGET | D2D1_BITMAP_OPTIONS_CANNOT_DRAW; + desc = &bitmap_desc; + } + + if (SUCCEEDED(hr = d2d_bitmap_create_shared(context, &IID_IDXGISurface, surface, desc, &object))) + *bitmap = &object->ID2D1Bitmap1_iface; + + return hr; } static HRESULT STDMETHODCALLTYPE d2d_device_context_CreateEffect(ID2D1DeviceContext *iface, -- 2.18.0

Sept. 19, 2018

Re: [PATCH 2/2] user32/listbox: Optimize SetCount partly

by Gabriel Ivăncescu

On Wed, Sep 19, 2018 at 1:55 PM, Huw Davies <huw(a)codeweavers.com> wrote: > > I'm also concerned that we won't get WM_MEASUREITEMs in the > LBS_OWNERDRAWVARIABLE case. > > Huw. I don't know if that's an issue since SetCount should theoretically be an "instant" operation and these items are just added at the end as empty. I think Windows doesn't even support SetCount without LBS_NODATA (which is not supported by Wine yet, I think I'll try working on it and implement it in an optimal way, i.e. with the minimum amount of memory needed, since it's meant for very very large lists). But, at least for now, I can fall back to the original (unoptimal) behavior if OWNERDRAWVARIABLE is set, would that be acceptable?

Sept. 19, 2018