September 2018 - Wine-Devel

Re: [PATCH v2 5/8] shell32/autocomplete: Send messages directly to the edit control's procedure

by Alexandre Julliard

Gabriel Ivăncescu <gabrielopcode(a)gmail.com> writes: > On Fri, Sep 21, 2018 at 4:57 PM, Alexandre Julliard <julliard(a)winehq.org> wrote: >> Huw Davies <huw(a)codeweavers.com> writes: >> >>> On Fri, Sep 21, 2018 at 03:38:39PM +0300, Gabriel Ivăncescu wrote: >>>> On Fri, Sep 21, 2018 at 3:02 PM, Huw Davies <huw(a)codeweavers.com> wrote: >>>> > >>>> > As I hinted at above, unless you can type very much faster than me >>>> > then I don't consider this much of an issue. >>>> >>>> Well it's not much of an issue but it's still a slight improvement for >>>> the wineserver to have to process less in its queue, and all it needs >>>> is change a few lines (4 in total) from SendMessage to CallWindowProc. >>>> >>>> Are you sure you don't want me to do that? >>> >>> Yes. >>> >>> It may only be 4 lines, but it's a reasonably fundamental change to >>> how things work which will require time to review and add to the >>> possibility of regressions. So if it's not actually improving >>> things in a noticable manner then it should stay out. >> >> Not to mention that SendMessage doesn't do a wineserver call for the >> thread-local case, so it's not even improving anything. >> >> -- >> Alexandre Julliard >> julliard(a)winehq.org > > Yeah but it's not the SendMessage that's the wineserver call I'm > talking about, but the fact that it will go through this subclassed > window procedure again (it starts at the top), which ends up calling > GetPropW to get the This pointer just to forward it to the edit > control. As I see, GetPropW uses a wineserver call. If it turns out that GetPropW is a bottleneck (very unlikely), it can be optimized in various ways. Please don't obfuscate the code in an attempt to make things faster where it doesn't matter. -- Alexandre Julliard julliard(a)winehq.org

Sept. 21, 2018

Re: [PATCH v2 4/8] shell32/autocomplete: Don't autocomplete at all on most control characters

by Gabriel Ivăncescu

On Fri, Sep 21, 2018 at 3:40 PM, Huw Davies <huw(a)codeweavers.com> wrote: > On Fri, Sep 21, 2018 at 03:34:52PM +0300, Gabriel Ivăncescu wrote: >> On Fri, Sep 21, 2018 at 2:57 PM, Huw Davies <huw(a)codeweavers.com> wrote: >> > >> > Multiline edit controls will take all of the above. Do they work with >> > autocomplete? >> >> If they handle the same messages, then they should, there's nothing >> (at least in our) code that specifically asks for a specific type of >> edit control. So maybe I should just keep it like my example, to be on >> the safe side. > > I was hoping that you might test it with a multiline edit ;-) > > Huw. Well as expected, with Wine's edit control it seems they do work indeed with autocomplete, in the sense that a tab character does end up printed on a multi-line edit control (same with newline of course) and then it does auto-suggest an entry with a tab, so autocomplete works. I couldn't get Windows XP to work yet to print a tab, but since our edit control already handles such characters like tab, I think it's better to allow it to autocomplete just in case. Or am I wrong?

Sept. 21, 2018

Re: [PATCH v2 5/8] shell32/autocomplete: Send messages directly to the edit control's procedure

by Gabriel Ivăncescu

On Fri, Sep 21, 2018 at 4:57 PM, Alexandre Julliard <julliard(a)winehq.org> wrote: > Huw Davies <huw(a)codeweavers.com> writes: > >> On Fri, Sep 21, 2018 at 03:38:39PM +0300, Gabriel Ivăncescu wrote: >>> On Fri, Sep 21, 2018 at 3:02 PM, Huw Davies <huw(a)codeweavers.com> wrote: >>> > >>> > As I hinted at above, unless you can type very much faster than me >>> > then I don't consider this much of an issue. >>> >>> Well it's not much of an issue but it's still a slight improvement for >>> the wineserver to have to process less in its queue, and all it needs >>> is change a few lines (4 in total) from SendMessage to CallWindowProc. >>> >>> Are you sure you don't want me to do that? >> >> Yes. >> >> It may only be 4 lines, but it's a reasonably fundamental change to >> how things work which will require time to review and add to the >> possibility of regressions. So if it's not actually improving >> things in a noticable manner then it should stay out. > > Not to mention that SendMessage doesn't do a wineserver call for the > thread-local case, so it's not even improving anything. > > -- > Alexandre Julliard > julliard(a)winehq.org Yeah but it's not the SendMessage that's the wineserver call I'm talking about, but the fact that it will go through this subclassed window procedure again (it starts at the top), which ends up calling GetPropW to get the This pointer just to forward it to the edit control. As I see, GetPropW uses a wineserver call. In contrast, CallWindowProc on the This->wpOrigEditProc would obviously bypass this (which is what the original would end up doing anyway, since we don't handle WM_GETTEXT and such in our subclassed proc).

Sept. 21, 2018

Re: [PATCH v4 3/3] comctl32/pager: Support comboboxex notification conversion.

by Nikolay Sivov

Signed-off-by: Nikolay Sivov <nsivov(a)codeweavers.com>

Sept. 21, 2018

Re: [PATCH v4 2/3] comctl32/pager: Support tooltip notification conversion.

by Nikolay Sivov

Signed-off-by: Nikolay Sivov <nsivov(a)codeweavers.com>

Sept. 21, 2018

Re: [PATCH v4 1/3] comctl32/pager: Support toolbar notification conversion.

by Nikolay Sivov

Signed-off-by: Nikolay Sivov <nsivov(a)codeweavers.com>

Sept. 21, 2018

Re: [PATCH v2 5/8] shell32/autocomplete: Send messages directly to the edit control's procedure

by Alexandre Julliard

Huw Davies <huw(a)codeweavers.com> writes: > On Fri, Sep 21, 2018 at 03:38:39PM +0300, Gabriel Ivăncescu wrote: >> On Fri, Sep 21, 2018 at 3:02 PM, Huw Davies <huw(a)codeweavers.com> wrote: >> > >> > As I hinted at above, unless you can type very much faster than me >> > then I don't consider this much of an issue. >> >> Well it's not much of an issue but it's still a slight improvement for >> the wineserver to have to process less in its queue, and all it needs >> is change a few lines (4 in total) from SendMessage to CallWindowProc. >> >> Are you sure you don't want me to do that? > > Yes. > > It may only be 4 lines, but it's a reasonably fundamental change to > how things work which will require time to review and add to the > possibility of regressions. So if it's not actually improving > things in a noticable manner then it should stay out. Not to mention that SendMessage doesn't do a wineserver call for the thread-local case, so it's not even improving anything. -- Alexandre Julliard julliard(a)winehq.org

Sept. 21, 2018

[PATCH] dwrite/layout: Test resulting line width before adding inline trimming run.

by Nikolay Sivov

Signed-off-by: Nikolay Sivov <nsivov(a)codeweavers.com> --- This fixes out-of-bounds read in cluster array, for example when trimming is applied to single cluster layout. dlls/dwrite/layout.c | 3 +++ dlls/dwrite/tests/layout.c | 43 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 46 insertions(+) diff --git a/dlls/dwrite/layout.c b/dlls/dwrite/layout.c index 10d591457c..5deae455b1 100644 --- a/dlls/dwrite/layout.c +++ b/dlls/dwrite/layout.c @@ -1926,6 +1926,9 @@ static void layout_add_line(struct dwrite_textlayout *layout, UINT32 first_clust if (FAILED(hr)) return; + if (get_cluster_range_width(layout, start, i) + sign_metrics.width > layout->metrics.layoutWidth) + append_trimming_run = FALSE; + if (append_trimming_run) { struct layout_effective_inline *trimming_sign; diff --git a/dlls/dwrite/tests/layout.c b/dlls/dwrite/tests/layout.c index 652f6b78ac..c77a0ea51b 100644 --- a/dlls/dwrite/tests/layout.c +++ b/dlls/dwrite/tests/layout.c @@ -1636,6 +1636,11 @@ static const struct drawcall_entry draw_seq[] = { { DRAW_LAST_KIND } }; +static const struct drawcall_entry draw_trimmed_seq[] = { + { DRAW_GLYPHRUN, {'a',0}, {'e','n','-','u','s',0}, 1 }, + { DRAW_LAST_KIND } +}; + static const struct drawcall_entry draw_seq2[] = { { DRAW_GLYPHRUN, {'s',0}, {'r','u',0}, 1 }, { DRAW_GLYPHRUN, {'t',0}, {'r','u',0}, 1 }, @@ -2386,6 +2391,44 @@ todo_wine IDWriteTextLayout_Release(layout); + /* Single cluster layout, trigger trimming. */ + hr = IDWriteFactory_CreateTextLayout(factory, str6W, 1, format, 1000.0f, 200.0f, &layout); + ok(hr == S_OK, "Failed to create layout, hr %#x.\n", hr); + + count = 0; + memset(metrics, 0, sizeof(metrics)); + hr = IDWriteTextLayout_GetClusterMetrics(layout, metrics, 1, &count); + ok(hr == S_OK, "Failed to get cluster metrics, hr %#x.\n", hr); + ok(count == 1, "Unexpected cluster count %u.\n", count); + + hr = IDWriteTextLayout_SetMaxWidth(layout, metrics[0].width / 2.0f); + ok(hr == S_OK, "Failed to set layout width, hr %#x.\n", hr); + + trimming_options.granularity = DWRITE_TRIMMING_GRANULARITY_CHARACTER; + trimming_options.delimiter = 0; + trimming_options.delimiterCount = 0; + hr = IDWriteTextLayout_SetTrimming(layout, &trimming_options, trimm); + ok(hr == S_OK, "Failed to set trimming options, hr %#x.\n", hr); + + count = 0; + memset(metrics, 0, sizeof(metrics)); + hr = IDWriteTextLayout_GetClusterMetrics(layout, metrics, 1, &count); + ok(hr == S_OK, "Failed to get cluster metrics, hr %#x.\n", hr); + ok(count == 1, "Unexpected cluster count %u.\n", count); + + hr = IDWriteTextLayout_GetLineMetrics(layout, &line, 1, &count); + ok(hr == S_OK, "Failed to get line metrics, hr %#x.\n", hr); + ok(count == 1, "Unexpected line count %u.\n", count); + ok(line.length == 1, "Unexpected line length %u.\n", line.length); + ok(line.isTrimmed, "Unexpected trimming flag %x.\n", line.isTrimmed); + + flush_sequence(sequences, RENDERER_ID); + hr = IDWriteTextLayout_Draw(layout, NULL, &testrenderer, 0.0f, 0.0f); + ok(hr == S_OK, "Draw() failed, hr %#x.\n", hr); + ok_sequence(sequences, RENDERER_ID, draw_trimmed_seq, "Trimmed draw test", FALSE); + + IDWriteTextLayout_Release(layout); + IDWriteInlineObject_Release(trimm); IDWriteTextFormat_Release(format); IDWriteFactory_Release(factory); -- 2.18.0

Sept. 21, 2018

Re: [PATCH] msvcp90: Remove unnecessary check for __arm__.

by Piotr Caban

Signed-off-by: Piotr Caban <piotr(a)codeweavers.com>

Sept. 21, 2018

Re: [PATCH v2 5/8] shell32/autocomplete: Send messages directly to the edit control's procedure

by Huw Davies

On Fri, Sep 21, 2018 at 03:38:39PM +0300, Gabriel Ivăncescu wrote: > On Fri, Sep 21, 2018 at 3:02 PM, Huw Davies <huw(a)codeweavers.com> wrote: > > > > As I hinted at above, unless you can type very much faster than me > > then I don't consider this much of an issue. > > Well it's not much of an issue but it's still a slight improvement for > the wineserver to have to process less in its queue, and all it needs > is change a few lines (4 in total) from SendMessage to CallWindowProc. > > Are you sure you don't want me to do that? Yes. It may only be 4 lines, but it's a reasonably fundamental change to how things work which will require time to review and add to the possibility of regressions. So if it's not actually improving things in a noticable manner then it should stay out. Huw.

Sept. 21, 2018