Hi Omar,
On 24 January 2016 at 01:10, Omar Sandoval <osandov_AT_osandov.com> wrote:
> ---
> This has been discussed a couple of times [1][2], and I think it's about time
> that it gets fixed. I noticed this issue because xfce4-screenshooter -w, which
> is supposed to screenshot the active window, doesn't work with dwm. After
> digging through the source code, I found that it uses
> gdk_screen_get_active_window() [3]. According to the documentation, that
> function looks at _NET_ACTIVE_WINDOW, which dwm sets. Looking at the GDK source
> code, I found that gdk_x11_screen_supports_net_wm_hint() [4] returns false for
> _NET_ACTIVE_WINDOW on dwm regardless of the value of _NET_SUPPORTED because
> _NET_SUPPORTING_WM_CHECK is not set (see fetch_net_wm_check_window() [5]).
>
> tl;dr Not setting _NET_SUPPORTING_WM_CHECK breaks real applications, and dwm is
> already going through the trouble of setting several EWMH atoms, so we should
> fix this.
>
> 1: http://lists.suckless.org/dev/1411/24636.html
> 2: http://lists.suckless.org/dev/1505/26735.html
> 3: https://developer.gnome.org/gdk2/stable/GdkScreen.html#gdk-screen-get-active-window
> 4: https://git.gnome.org/browse/gtk+/tree/gdk/x11/gdkscreen-x11.c?h=3.19.7#n1635
> 5: https://git.gnome.org/browse/gtk+/tree/gdk/x11/gdkscreen-x11.c?h=3.19.7#n1580
>
> dwm.c | 12 +++++++++++-
> 1 file changed, 11 insertions(+), 1 deletion(-)
>
> diff --git a/dwm.c b/dwm.c
> index ff7e096f61fc..7c12796419ca 100644
> --- a/dwm.c
> +++ b/dwm.c
> _AT_@ -60,7 +60,7 @@
> /* enums */
> enum { CurNormal, CurResize, CurMove, CurLast }; /* cursor */
> enum { SchemeNorm, SchemeSel, SchemeLast }; /* color schemes */
> -enum { NetSupported, NetWMName, NetWMState,
> +enum { NetSupported, NetSupportingWMCheck, NetWMName, NetWMState,
> NetWMFullscreen, NetActiveWindow, NetWMWindowType,
> NetWMWindowTypeDialog, NetClientList, NetLast }; /* EWMH atoms */
> enum { WMProtocols, WMDelete, WMState, WMTakeFocus, WMLast }; /* default atoms */
> _AT_@ -1574,6 +1574,7 @@ setup(void)
> wmatom[WMTakeFocus] = XInternAtom(dpy, "WM_TAKE_FOCUS", False);
> netatom[NetActiveWindow] = XInternAtom(dpy, "_NET_ACTIVE_WINDOW", False);
> netatom[NetSupported] = XInternAtom(dpy, "_NET_SUPPORTED", False);
> + netatom[NetSupportingWMCheck] = XInternAtom(dpy, "_NET_SUPPORTING_WM_CHECK", False);
> netatom[NetWMName] = XInternAtom(dpy, "_NET_WM_NAME", False);
> netatom[NetWMState] = XInternAtom(dpy, "_NET_WM_STATE", False);
> netatom[NetWMFullscreen] = XInternAtom(dpy, "_NET_WM_STATE_FULLSCREEN", False);
> _AT_@ -1806,6 +1807,8 @@ updatebars(void)
> .background_pixmap = ParentRelative,
> .event_mask = ButtonPressMask|ExposureMask
> };
> + Atom utf8string;
> + utf8string = XInternAtom(dpy, "UTF8_STRING", False);
> for (m = mons; m; m = m->next) {
> if (m->barwin)
> continue;
> _AT_@ -1814,7 +1817,14 @@ updatebars(void)
> CWOverrideRedirect|CWBackPixmap|CWEventMask, &wa);
> XDefineCursor(dpy, m->barwin, cursor[CurNormal]->cursor);
> XMapRaised(dpy, m->barwin);
> + XChangeProperty(dpy, m->barwin, netatom[NetWMName], utf8string,
> + 8, PropModeReplace, (unsigned char *)"dwm", 3);
> }
> + /* it doesn't matter which bar we use for this, so just use the first one */
> + XChangeProperty(dpy, root, netatom[NetSupportingWMCheck], XA_WINDOW, 32,
> + PropModeReplace, (unsigned char *)&mons->barwin, 1);
> + XChangeProperty(dpy, mons->barwin, netatom[NetSupportingWMCheck], XA_WINDOW,
> + 32, PropModeReplace, (unsigned char *)&mons->barwin, 1);
> }
To me it looks like a rather odd interpretation of wm spec that is
used in the gdkscreen implementation. They query
_NET_SUPPORTING_WM_CHECK != None for any atom that is queried through
gdk_x11_screen_supports_net_wm_hint().
I think I can live with re-introducing this atom into dwm.
Nevertheless EWMH stinks. I did remove this handling back in 2007 btw.
-Anselm
Received on Sun Jan 24 2016 - 15:22:36 CET