[vlc-devel] [PATCH 4/4] win32: Send fullscreen to specific monitor

Gabriel Luci github at luci.ca
Thu Oct 24 15:40:51 CEST 2019


Well I just set a breakpoint at the start of WinVoutEventProc and it wasn't
hit, but maybe it's not being hit for other reasons. It seems inconsistent.
Spy++ does show WM_MOVE and WM_WINDOWPOSCHANGED. I'll add a case to the
switch block for WM_MOVE and see what happens.

On Thu, Oct 24, 2019 at 9:11 AM Steve Lhomme <robux4 at ycbcr.xyz> wrote:

> On 2019-10-24 15:01, Gabriel Luci wrote:
> > I think this depends on what value is going to be in whatever setting we
> > end up deciding to use. If we store the device name (like
> > "\\.\DISPLAY2"), then you have to enumerate the displays to figure out
> > which one has that name, since there isn't any GetMonitorByName
> > function. But once you find it, you already have the MONITORINFOEX
> > struct that you need to go fullscreen. That's why I did what I did.
> >
> > If you want to use the SetFullscreenOnMonitor, then the enumeration
> > would have to pass back just the handle to the monitor
> > and SetFullscreenOnMonitor would have to use the handle to go find the
> > monitor info again.
>
> I guess you could pass the MONITORINFOEX to SetFullscreenOnMonitor then.
> We should read it in the current code as well.
>
> > To Alexandre's point about switching monitors: This is a bug that exists
> > in VLC 3 too. I'm just testing this now in the current v4 code, and I
> > don't see any message at all sent to WinVoutEventProc when using
> > Win+Shift+Arrow. I would have expected to see WM_MOVE and figured we
> > could just check if we're in fullscreen mode and resize to the new
> > monitor. But as it is, I don't see VLC being notified at all that the
> > window moved. Any idea why?
>
> This is weird and unexpected. Did you log all the messages received at
> the beginning of WinVoutEventProc() ? You may also check with Spy++ (or
> Spy32++?)
>
> > If you move from a larger resolution monitor to a smaller one, you see
> > WM_SIZE because Windows resizes the window to the new display, but minus
> > the taskbar. But you see no notifications at all if you move from a
> > smaller monitor to a larger one.
> >
> > On Thu, Oct 24, 2019 at 7:32 AM Steve Lhomme <robux4 at ycbcr.xyz
> > <mailto:robux4 at ycbcr.xyz>> wrote:
> >
> >     Here's a patch that allows going fullscreen on a given HMONITOR.
> Thats'
> >     what this patch should use in the end and the logo+shift+arrow patch.
> >
> >     On 2019-10-24 13:17, Steve Lhomme wrote:
> >      > On 2019-10-24 13:14, Steve Lhomme wrote:
> >      >> On 2019-10-24 12:21, Alexandre Janniaux wrote:
> >      >>> When in fullscreen, the window can be moved to another
> >      >>> screen by using logo+shift+arrow key. When doing so on
> >      >>> screens which are not the same size, the window is not
> >      >>> resized and keep the size from the former screen. It should
> >      >>> probably keep the resize state somewhere and resize
> >      >>> adequatly while using vout_window_ReportSize() so that
> >      >>> video outputs can be resized too.
> >      >>>
> >      >>> I don't know if you want to add this to your patch too, as
> >      >>> the behaviour is already present without yours, but maybe
> >      >>> Windows dev here can provide advice for this?
> >      >>
> >      >> I should be able to fix it. We probably need to handle the
> WM_MOVE
> >      >> and, when in fullscreen, adapt the size of the window if the
> >     monitor
> >      >> changed.
> >      >>
> >      >> It's probably better to apply a fix after this change as it would
> >      >> impact how we apply the fullscreen based on the selected
> >     monitor. In
> >      >> this case it wouldn't come the core anymore but from the local
> >      >> detection of the monitor. That means the monitor from the core
> >     should
> >      >> return just a
> >      >
> >      > the monitor *selection* (based on the screen from the core)
> >      >
> >      >> HMONITOR, which is always what we'll use when we use the one
> >     forced by
> >      >> the OS.
> >      >>
> >      >>> PS: just for giving context, I used your patch in a VR demo
> >      >>> so as to force a VR sidebyside output to the HMD screen and
> >      >>> a classic 360° pano output to the main screen.
> >      >>>
> >      >>> On Sat, Oct 12, 2019 at 11:58:31PM -0400, Gabriel Luci wrote:
> >      >>>> This reads the id parameter as the name of the monitor that
> >     should be
> >      >>>> used. If anything goes wrong, it will fall back to the monitor
> the
> >      >>>> window is already on (which is what it always did before).
> >      >>>>
> >      >>>> ---
> >      >>>>   modules/video_output/win32/window.c | 72
> >      >>>> +++++++++++++++++++++++++++++++++----
> >      >>>>   1 file changed, 65 insertions(+), 7 deletions(-)
> >      >>>>
> >      >>>> diff --git a/modules/video_output/win32/window.c
> >      >>>> b/modules/video_output/win32/window.c
> >      >>>> index 50e172b9b4..eefaed256e 100644
> >      >>>> --- a/modules/video_output/win32/window.c
> >      >>>> +++ b/modules/video_output/win32/window.c
> >      >>>> @@ -37,6 +37,7 @@
> >      >>>>   #include <vlc_actions.h>
> >      >>>>
> >      >>>>   #include
> >     <shellapi.h>                                         /*
> >      >>>> ExtractIcon */
> >      >>>> +#include <vlc_charset.h>
> /*
> >      >>>> FromWide */
> >      >>>>
> >      >>>>   #define RECTWidth(r)   (LONG)((r).right - (r).left)
> >      >>>>   #define RECTHeight(r)  (LONG)((r).bottom - (r).top)
> >      >>>> @@ -153,9 +154,37 @@ static void SetTitle(vout_window_t *wnd,
> >     const
> >      >>>> char *title)
> >      >>>>       PostMessage( sys->hwnd, WM_VLC_CHANGE_TEXT, 0, 0 );
> >      >>>>   }
> >      >>>>
> >      >>>> +typedef struct enum_monitors_callback_params_t {
> >      >>>> +    bool is_monitor_found;
> >      >>>> +    const char *psz_name_to_find;
> >      >>>> +    MONITORINFOEX *found_monitor_info;
> >      >>>> +} enum_monitors_callback_params_t;
> >      >>>> +
> >      >>>> +static BOOL CALLBACK EnumDisplayMonitorsCallback(HMONITOR
> >     hMonitor,
> >      >>>> HDC hdcMonitor, LPRECT lprcMonitor, LPARAM dwData)
> >      >>>> +{
> >      >>>> +    /* this is called for each active monitor */
> >      >>>> +
> >      >>>> +    VLC_UNUSED(hdcMonitor);
> >      >>>> +    VLC_UNUSED(lprcMonitor);
> >      >>>> +
> >      >>>> +    MONITORINFOEX mi;
> >      >>>> +    mi.cbSize = sizeof(MONITORINFOEX);
> >      >>>> +    if (!GetMonitorInfo(hMonitor, (LPMONITORINFO) &mi))
> >     return true;
> >      >>>> +
> >      >>>> +    enum_monitors_callback_params_t *params =
> >      >>>> (enum_monitors_callback_params_t*) dwData;
> >      >>>> +    char *psz_monitor_name = FromWide(mi.szDevice);
> >      >>>> +    if (!strcmp(psz_monitor_name, params->psz_name_to_find))
> >      >>>> +    {
> >      >>>> +        /* the monitor name matches what we're looking for */
> >      >>>> +        params->is_monitor_found = true;
> >      >>>> +        params->found_monitor_info = &mi;
> >      >>>> +    }
> >      >>>> +    free(psz_monitor_name);
> >      >>>> +    return !params->is_monitor_found;
> >      >>>> +}
> >      >>>> +
> >      >>>>   static void SetFullscreen(vout_window_t *wnd, const char *id)
> >      >>>>   {
> >      >>>> -    VLC_UNUSED(id);
> >      >>>>       msg_Dbg(wnd, "entering fullscreen mode");
> >      >>>>       vout_window_sys_t *sys = wnd->sys;
> >      >>>>
> >      >>>> @@ -167,12 +196,41 @@ static void SetFullscreen(vout_window_t
> >     *wnd,
> >      >>>> const char *id)
> >      >>>>       /* Change window style, no borders and no title bar */
> >      >>>>       SetWindowLong(sys->hwnd, GWL_STYLE, WS_CLIPCHILDREN |
> >      >>>> WS_VISIBLE);
> >      >>>>
> >      >>>> -    /* Retrieve current window position so fullscreen will
> happen
> >      >>>> -     * on the right screen */
> >      >>>> -    HMONITOR hmon = MonitorFromWindow(sys->hwnd,
> >      >>>> MONITOR_DEFAULTTONEAREST);
> >      >>>> -    MONITORINFO mi;
> >      >>>> -    mi.cbSize = sizeof(MONITORINFO);
> >      >>>> -    if (GetMonitorInfo(hmon, &mi))
> >      >>>> +    /* Figure out which monitor we should be using */
> >      >>>> +    MONITORINFOEX mi;
> >      >>>> +    bool is_monitor_found = false;
> >      >>>> +
> >      >>>> +    if (id)
> >      >>>> +    {
> >      >>>> +        /* We're being told to use a specific monitor, so see
> >     if we
> >      >>>> can find an
> >      >>>> +         * active monitor with the same name */
> >      >>>> +        enum_monitors_callback_params_t params = {
> >      >>>> +            .is_monitor_found = false,
> >      >>>> +            .psz_name_to_find = id
> >      >>>> +        };
> >      >>>> +        EnumDisplayMonitors(NULL, NULL,
> >      >>>> &EnumDisplayMonitorsCallback, (LPARAM) &params);
> >      >>>> +
> >      >>>> +        if (params.is_monitor_found)
> >      >>>> +        {
> >      >>>> +            is_monitor_found = true;
> >      >>>> +            mi = *(params.found_monitor_info);
> >      >>>> +        }
> >      >>>> +        else
> >      >>>> +        {
> >      >>>> +            msg_Dbg(wnd, "display device with name \"%s\" was
> >      >>>> requested, but no such device exists. Falling back to
> >     primary", id);
> >      >>>> +        }
> >      >>>> +    }
> >      >>>> +
> >      >>>> +    if (!is_monitor_found)
> >      >>>> +    {
> >      >>>> +        /* fall back to using the screen the window is
> currently
> >      >>>> mostly on */
> >      >>>> +        HMONITOR hmon;
> >      >>>> +        hmon = MonitorFromWindow(sys->hwnd,
> >     MONITOR_DEFAULTTONEAREST);
> >      >>>> +        mi.cbSize = sizeof(MONITORINFOEX);
> >      >>>> +        is_monitor_found = GetMonitorInfo(hmon,
> >     (LPMONITORINFO) &mi);
> >      >>>> +    }
> >      >>>> +
> >      >>>> +    if (is_monitor_found)
> >      >>>>           SetWindowPos(sys->hwnd, 0,
> >      >>>>                        mi.rcMonitor.left,
> >      >>>>                        mi.rcMonitor.top,
> >      >>>> --
> >      >>>> 2.11.0
> >      >>>>
> >      >>>> _______________________________________________
> >      >>>> vlc-devel mailing list
> >      >>>> To unsubscribe or modify your subscription options:
> >      >>>> https://mailman.videolan.org/listinfo/vlc-devel
> >      >>> _______________________________________________
> >      >>> vlc-devel mailing list
> >      >>> To unsubscribe or modify your subscription options:
> >      >>> https://mailman.videolan.org/listinfo/vlc-devel
> >      >>>
> >      >> _______________________________________________
> >      >> vlc-devel mailing list
> >      >> To unsubscribe or modify your subscription options:
> >      >> https://mailman.videolan.org/listinfo/vlc-devel
> >      > _______________________________________________
> >      > vlc-devel mailing list
> >      > To unsubscribe or modify your subscription options:
> >      > https://mailman.videolan.org/listinfo/vlc-devel
> >     _______________________________________________
> >     vlc-devel mailing list
> >     To unsubscribe or modify your subscription options:
> >     https://mailman.videolan.org/listinfo/vlc-devel
> >
> >
> > _______________________________________________
> > vlc-devel mailing list
> > To unsubscribe or modify your subscription options:
> > https://mailman.videolan.org/listinfo/vlc-devel
> >
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/vlc-devel/attachments/20191024/b8edb500/attachment-0001.html>


More information about the vlc-devel mailing list