Xdamage: get event on window content changes - c++

I am a beginner in xlib and I want to catch event on window content changes. For examples if some region changed. I've read that I need to use Xdamage for that and this is my code
Display* display = XOpenDisplay(NULL);
Window root = DefaultRootWindow(display);
XWindowAttributes attributes = {0};
XGetWindowAttributes(display, root, &attributes);
int width, height;
width = attributes.width;
height = attributes.height;
//XDAMAGE INIT
int damage_event, damage_error, test;
test = XDamageQueryExtension(display, &damage_event, &damage_error);
Damage damage = XDamageCreate(display, root, XDamageReportNonEmpty);
while (true){
XEvent event;
XNextEvent(display,&event);
//HERE I GET EVENT
XDamageSubtract(display,damage,None,None);
}
XCloseDisplay(display);
So, I have a working example in which I can get event and its type.The types, as I understand are xlib types - https://tronche.com/gui/x/xlib/events/types.html. But how can I find out from event that window was changed.

What you are actually receiving are DamageNotify events in your normal event loop that look like so:
typedef struct {
int type; /* event base */
unsigned long serial;
Bool send_event;
Display *display;
Drawable drawable;
Damage damage;
int level;
Bool more; /* more events will be delivered immediately */
Time timestamp;
XRectangle area;
XRectangle geometry;
} XDamageNotifyEvent;
according to the documentation from https://www.x.org/releases/current/doc/damageproto/damageproto.txt
The area member reports the damaged rectangular area as an XRectangle.
The drawable member normally reports the window that was damaged.
Note XNextEvent will always return an XEvent That is the static type returned by the function. It also has a dynamic type that depends on the ev->type member. - Check its type field to find out if you may cast it to an XDamageEvent:
XEvent ev;
/* We might need a pointer to an XDamageEvent... */
XDamageNotifyEvent *dev;
while (1) {
XNextEvent(display,&ev);
switch (ev.type){
...
case DamageNotify:
dev = (XDamageNotifyEvent*)&ev; /* if type matches, cast to XDamageNotify event */
printf ("Damage notification in window %d\n", dev->drawable);
printf ("Rectangle x=%d, y=%d, w=%d, h=%d\n",
dev->area.x, dev->area.y,
dev->area.width, dev->area.height);
break;
...
}

Late, but for anyone looking for the area that actually changed (ev->area are the window coordinates and size, for me at least):
auto ev = (XDamageNotifyEvent*)e;
auto region = XFixesCreateRegion(display, null, 0);
XDamageSubtract(display, ev->damage, None, region);
int count;
auto area = XFixesFetchRegion(display, region, &count);
if(area){
for(int i; i < count; i++){
auto rect = area[i];
// do something with rect
}
XFree(area);
}
XFixesDestroyRegion(display, region);

Related

Why isn't SDL_CreateWindow showing a window when called from another class?

Rewriting this to try to provide some clarity and update the code with some things that have changed.
I am restructuring a project that used SDL2, and have encountered issues trying to create a blank window. I have attempted to structure the project similarly to the original by separating all functionality dealing with SDL_Window into its own class. If I move the call to SDL_CreateWindow into the same class as the event loop or move the event loop to the same class as the window, the window is created and shown as expected, however as it is now, the window appears to be created successfully (SDL_CreateWindow is not returning NULL) and the program doesn't seem to be hanging, but it does not display a window while the program is running.
The SDL_Window is created in the Graphics class and stored in a member variable:
Graphics::Graphics(const char* title, unsigned int w, unsigned int h, unsigned int flags, int& status) {
screen = SDL_CreateWindow(title,
SDL_WINDOWPOS_UNDEFINED, SDL_WINDOWPOS_UNDEFINED,
w, h,
flags);
status = 0;
if (screen == NULL)
status = 1;
}
Graphics is instantiated in the Window class and stored in a member variable.
Window::Window(const char* title, unsigned int w, unsigned int h, unsigned int flags, int& status) {
g = Graphics(title, w,h, flags, status);
}
Window is instantiated in main, and if the window is created successfully, it starts the event loop.
{
int status;
Window window("Mirari", 640,480, SDL_WINDOW_SHOWN, status);
if (status == 0) {
window.eventLoop();
} else {
SDL_LogError(SDL_LOG_CATEGORY_APPLICATION, "Couldn't create window and renderer: %s", SDL_GetError());
return 1;
}
}
The event loop itself to be thorough (update and draw are both currently empty functions).
void Window::eventLoop() {
SDL_Event ev;
bool running = true;
while (running) {
const int start_time = SDL_GetTicks();
while (SDL_PollEvent(&ev)) {
switch (ev.type) {
case SDL_QUIT:
running = false;
break;
default:
break;
}
}
//update();
//draw();
std::cout << "." << std::endl;
const int elapsed = SDL_GetTicks() - start_time;
if (elapsed < 1000 / FPS)
SDL_Delay(1000 / FPS - elapsed);
}
}
SDL is initialized with this static function and these flags.
void Window::init(unsigned int sdl_flags, IMG_InitFlags img_flags) {
SDL_Init(sdl_flags);
IMG_Init(img_flags);
TTF_Init();
}
...
Window::init(SDL_INIT_VIDEO | SDL_INIT_EVENTS | SDL_INIT_TIMER, IMG_INIT_PNG);
I know that a window can be created in a separate class because the first version of this project did that and it worked, I'm just not sure what has changed that is causing the window not to show up.
As said by some programmer dude, you design is not perfect and should be thought again.
Nevertheless, from what we can see on your code : If the Window constructor is called (and the SDL_Init was called before, which I assume so), then the windows should be created.
From there we only can guess what we can't see (as it's not part of what you are displaying) :
is the definition of SDL_WINDOWPOS_UNDEFINED, the same in both context ?
is the screen variable definition the same in both context ?
is the "screen" used in "update", or "draw" method, and, as uninitialized : it fails
... ?
As you probably are new to development, I suggest you adopt this habit very early : your code should check and report everything it does. A good program is easy to debug, as it says what's wrong
For instance, just after :
screen = SDL_CreateWindow(title, SDL_WINDOWPOS_UNDEFINED, SDL_WINDOWPOS_UNDEFINED,
w, h, flags);
you may want to write something like :
if(!screen)
{
std::cout << "Failed to create window\n";
return -1;
}
or better :
if(!screen)
{
throw std::exception("Failed to create window\n");
}
And so on.
For instance, in your function update, you may want to have something like :
if(!screen)
{
throw std::exception("Unable to update the display as it is uninitialized\n");
}
I assume your application would not end without any comment... but that's a guess

Switching Between windowed and full screen in OpenGL/GLFW 3.2

I am in the process of learning OpenGL on Linux but I can't get mode switching working (windowed to full screen and back).
The window appears to be going into full screen but but not looking correct. To switch modes a new window is being created and old one destroyed.
void OpenGLWindow::FullScreen(bool fullScreen, int width, int height)
{
GLFWwindow *oldHandle = m_window;
m_fullscreen = fullScreen;
m_width = width;
m_height = height;
m_window = glfwCreateWindow(width, height, m_caption.c_str(),
fullScreen ? m_monitor : NULL, m_window);
if (m_window == NULL)
{
glfwTerminate();
throw std::runtime_error("Failed to recreate window.");
}
glfwDestroyWindow(oldHandle);
m_camera->Invalidate();
// Use entire window for rendering.
glViewport(0, 0, width, height);
glfwMakeContextCurrent(m_window);
glfwSwapInterval(1);
if (m_keyboardHandler) SetKeyboardHandler(m_keyboardHandler);
}
Initial Window
Full Screen (incorrect)
Return to Windowed
Updates to Question
I have updated the code to use your code and getting the same issue. On your suggestion I am now updating the camera, but again no avail :(
void OpenGLCamera::Invalidate()
{
RecalculateProjection(m_perspProjInfo->Width(), m_perspProjInfo->Height());
m_recalculateViewMatrix = true;
m_recalculatePerspectiveMatrix = true;
m_recalculateProjectionMatrix = true;
}
void OpenGLCamera::RecalculateProjection(int width, int height)
{
float aspectRatio = float(width) / height;
float frustumYScale = cotangent(degreesToRadians(
m_perspProjInfo->FieldOfView() / 2));
float frustumXScale = frustumYScale;
if (width > height)
{
// Shrink the x scale in eye-coordinate space, so that when geometry is
// projected to ndc-space, it is widened out to become square.
m_projectionMatrix[0][0] = frustumXScale / aspectRatio;
m_projectionMatrix[1][1] = frustumYScale;
}
else {
// Shrink the y scale in eye-coordinate space, so that when geometry is
// projected to ndc-space, it is widened out to become square.
m_projectionMatrix[0][0] = frustumXScale;
m_projectionMatrix[1][1] = frustumYScale * aspectRatio;
}
}
Rabbid : When I resize:
Rabbid : When I go to full screen:
In the following, I'll describe a small but handy class, which deals with resizing a GLFW window and handles switch fullscreen window on and off.
All the used GLFW functions are well documented in the GLFW documentation.
#include <GL/gl.h>
#include <GLFW/glfw3.h>
#include <array>
#include <stdexcept>
class OpenGLWindow
{
private:
std::array< int, 2 > _wndPos {0, 0};
std::array< int, 2 > _wndSize {0, 0};
std::array< int, 2 > _vpSize {0, 0};
bool _updateViewport = true;
GLFWwindow * _wnd = nullptr;
GLFWmonitor * _monitor = nullptr;
void Resize( int cx, int cy );
public:
void Init( int width, int height );
static void CallbackResize(GLFWwindow* window, int cx, int cy);
void MainLoop ( void );
bool IsFullscreen( void );
void SetFullScreen( bool fullscreen );
};
When creating the window, then the user function pointer (glfwSetWindowUserPointer) is set to the window management class. And the resize callback is set by glfwSetWindowSizeCallback. After the window is created its current size and position can be get by glfwGetWindowPos and glfwGetWindowSize.
void OpenGLWindow::Init( int width, int height )
{
_wnd = glfwCreateWindow( width, height, "OGL window", nullptr, nullptr );
if ( _wnd == nullptr )
{
glfwTerminate();
throw std::runtime_error( "error initializing window" );
}
glfwMakeContextCurrent( _wnd );
glfwSetWindowUserPointer( _wnd, this );
glfwSetWindowSizeCallback( _wnd, OpenGLWindow::CallbackResize );
_monitor = glfwGetPrimaryMonitor();
glfwGetWindowSize( _wnd, &_wndSize[0], &_wndSize[1] );
glfwGetWindowPos( _wnd, &_wndPos[0], &_wndPos[1] );
_updateViewport = true;
}
When the resize notification occurs, then the pointer to the window management class can be get by glfwGetWindowUserPointer:
static void OpenGLWindow::CallbackResize(GLFWwindow* window, int cx, int cy)
{
void *ptr = glfwGetWindowUserPointer( window );
if ( OpenGLWindow *wndPtr = static_cast<OpenGLWindow*>( ptr ) )
wndPtr->Resize( cx, cy );
}
Any change of the window size is notified and the new window size is stored (glfwGetWindowSize):
void OpenGLWindow::Resize( int cx, int cy )
{
_updateViewport = true;
}
When the window size has changed, then the viewport has to be suited to the window size (glViewport). This can be done in the main loop of the application:
void OpenGLWindow::MainLoop ( void )
{
while (!glfwWindowShouldClose(_wnd))
{
if ( _updateViewport )
{
glfwGetFramebufferSize( _wnd, &_vpSize[0], &_vpSize[1] );
glViewport( 0, 0, _vpSize[0], _vpSize[1] );
_updateViewport = false;
}
// ..... render the scene
glfwSwapBuffers(_wnd);
glfwPollEvents();
}
}
If the current window is in full screen mode, can be achieved by asking for the monitor that the window uses for full screen mode (glfwGetWindowMonitor):
bool OpenGLWindow::IsFullscreen( void )
{
return glfwGetWindowMonitor( _wnd ) != nullptr;
}
To switch the full screen mode on and off, glfwSetWindowMonitor has to be called, either with the monitor for the full screen mode, or with nullptr:
void OpenGLWindow::SetFullScreen( bool fullscreen )
{
if ( IsFullscreen() == fullscreen )
return;
if ( fullscreen )
{
// backup window position and window size
glfwGetWindowPos( _wnd, &_wndPos[0], &_wndPos[1] );
glfwGetWindowSize( _wnd, &_wndSize[0], &_wndSize[1] );
// get resolution of monitor
const GLFWvidmode * mode = glfwGetVideoMode(_monitor);
// switch to full screen
glfwSetWindowMonitor( _wnd, _monitor, 0, 0, mode->width, mode->height, 0 );
}
else
{
// restore last window size and position
glfwSetWindowMonitor( _wnd, nullptr, _wndPos[0], _wndPos[1], _wndSize[0], _wndSize[1], 0 );
}
_updateViewport = true;
}
I recommend you to not create a new Window with glfwCreateWindow when you just want to switch between windowed and fullscreen. Use glfwSetWindowMonitor instead.
When you create a window with fullscreen enabled, you have to pass arguments which are compatible with a video mode on the monitor. You can get the standard video mode on the primary monitor like this:
GLFWmonitor *monitor = glfwGetPrimaryMonitor();
const GLFWvidmode *mode = glfwGetVideoMode(monitor);
and to switch to fullscreen:
glfwSetWindowMonitor(window, monitor, 0, 0, mode->width, mode->height, mode->refreshRate);
Just pass a nullptr-mode and your own values of course:
glfwSetWindowMonitor(window, nullptr, 0, 0, windowWidth, windowHeight, windowRefreshRate);
And don't forget to resize the viewport and update the camera.
Are you resizing the viewport and updating the camera when the user resizes the window?
There are a couple of issues with your code:
Assuming that glfwCreateWindow will set the resolution to width * height in fullscreen mode is not correct. The GLFW documentation states (emphasis mine):
For full screen windows, the specified size becomes the resolution of the window's desired video mode. As long as a full screen window is not iconified, the supported video mode most closely matching the desired video mode is set for the specified monitor.
Assuming that the window size is specified in "pixels" is not correct either.Quoting the relevant part of the documentation again:
While the size of a window is measured in screen coordinates, OpenGL works with pixels. The size you pass into glViewport, for example, should be in pixels. On some machines screen coordinates and pixels are the same, but on others they will not be. There is a second set of functions to retrieve the size, in pixels, of the framebuffer of a window.
Issues 1 and 2 can be solved by simply calling glfwGetFramebufferSize after the window was created. This leaves us with issue 3:
You call glViewport without having a current GL context -
resulting in undefined behavior, and especially in not setting the viewport at all. Now that is actually an interesting one, because the initial viewport for the new context will be the full new window, so that your mistakes 1 and 2 have no direct effect. They still might have some effect later if your code relies on m_width and m_height containing useful values, though.

How to create a full screen window on the current monitor with GLFW

Creating a window with GLFW3 is done using glfwCreateWindow:
GLFWwindow* glfwCreateWindow ( int width,
int height,
const char *title,
GLFWmonitor *monitor,
GLFWwindow *share
)
If the monitor parameter is not NULL, the window is created in full screen mode on the given monitor. One can receive the primary monitor by calling glfwGetPrimaryMonitor, or chose one of the results of glfwGetMonitors. But how can I create a full screen window on the current monitor, i.e. the monitor the window is currently running in windowed mode? There seems to be no way to receive the currently used monitor. There is glfwGetWindowMonitor, but it only returns the monitor in full screen mode, NULL in windowed mode.
You can find the current monitor with glfwGetWindowPos/glfwGetWindowSize.
This function returns the monitor that contains the greater window area.
static int mini(int x, int y)
{
return x < y ? x : y;
}
static int maxi(int x, int y)
{
return x > y ? x : y;
}
GLFWmonitor* get_current_monitor(GLFWwindow *window)
{
int nmonitors, i;
int wx, wy, ww, wh;
int mx, my, mw, mh;
int overlap, bestoverlap;
GLFWmonitor *bestmonitor;
GLFWmonitor **monitors;
const GLFWvidmode *mode;
bestoverlap = 0;
bestmonitor = NULL;
glfwGetWindowPos(window, &wx, &wy);
glfwGetWindowSize(window, &ww, &wh);
monitors = glfwGetMonitors(&nmonitors);
for (i = 0; i < nmonitors; i++) {
mode = glfwGetVideoMode(monitors[i]);
glfwGetMonitorPos(monitors[i], &mx, &my);
mw = mode->width;
mh = mode->height;
overlap =
maxi(0, mini(wx + ww, mx + mw) - maxi(wx, mx)) *
maxi(0, mini(wy + wh, my + mh) - maxi(wy, my));
if (bestoverlap < overlap) {
bestoverlap = overlap;
bestmonitor = monitors[i];
}
}
return bestmonitor;
}
After discussion on IRC it seems that it is not possible to retrieve the currently active monitor (as in the monitor the window is currently drawn on) with GLFW. Therefore it is not possible to create a full screen window on the current monitor.
EDIT: Even though there is no GLFW functionality to directly achieve this, the answer of Shmo provides an elegant solution.
Here is Shmo's answer, ported over to LWJGL:
/** Determines the current monitor that the specified window is being displayed on.
* If the monitor could not be determined, the primary monitor will be returned.
*
* #param window The window to query
* #return The current monitor on which the window is being displayed, or the primary monitor if one could not be determined
* #author Shmo<br>
* Ported to LWJGL by Brian_Entei */
#NativeType("GLFWmonitor *")
public static final long glfwGetCurrentMonitor(long window) {
int[] wx = {0}, wy = {0}, ww = {0}, wh = {0};
int[] mx = {0}, my = {0}, mw = {0}, mh = {0};
int overlap, bestoverlap;
long bestmonitor;
PointerBuffer monitors;
GLFWVidMode mode;
bestoverlap = 0;
bestmonitor = glfwGetPrimaryMonitor();// (You could set this back to NULL, but I'd rather be guaranteed to get a valid monitor);
glfwGetWindowPos(window, wx, wy);
glfwGetWindowSize(window, ww, wh);
monitors = glfwGetMonitors();
while(monitors.hasRemaining()) {
long monitor = monitors.get();
mode = glfwGetVideoMode(monitor);
glfwGetMonitorPos(monitor, mx, my);
mw[0] = mode.width();
mh[0] = mode.height();
overlap =
Math.max(0, Math.min(wx[0] + ww[0], mx[0] + mw[0]) - Math.max(wx[0], mx[0])) *
Math.max(0, Math.min(wy[0] + wh[0], my[0] + mh[0]) - Math.max(wy[0], my[0]));
if (bestoverlap < overlap) {
bestoverlap = overlap;
bestmonitor = monitor;
}
}
return bestmonitor;
}

Xlib How Does This (Removing Window Decoration) Work?

How does the following code remove window borders?
//note the struct is declared elsewhere, is here just for clarity.
//code is from [http://tonyobryan.com/index.php?article=9][1]
typedef struct Hints
{
unsigned long flags;
unsigned long functions;
unsigned long decorations;
long inputMode;
unsigned long status;
} Hints;
//code to remove decoration
Hints hints;
Atom property;
hints.flags = 2;
hints.decorations = 0;
property = XInternAtom(display, "_MOTIF_WM_HINTS", true);
XChangeProperty(display,window,property,property,32,PropModeReplace,(unsigned char *)&hints,5);
XMapWindow(display, window);
So far I have gathered that an Atom is a sort of identifier similar to Window and Display but I can't figure out where the Hints structure or the "_MOTIF_WM_HINTS" came from. Can anyone explain all of this code for me? Thanks in advance, ell.
It's hard to come by any sort of "official"-looking standard or such, but the
_MOTIF_WM_HINTS property does seem to come from the... Motif toolkit (ok, you
guessed that :-) ).
See the MotifZone site.
Warning: what follows is incomplete, but should shed some light I hope.
Documentation for the XmNmwmDecorations, XmNmwmFunctions and XmNmwmInputMode
functions of the toolkit indicates that that property is a bitmask of various
values used to control the appearance, functions (resize, move, ...) and input
mode that the window manager should provide/give to the window.
See man vendorshell, or Oreilly Motif reference books, Vol6a chapter 16.
Properties are a part of the whole X11 thing. A window can have any number of
properties defined on it. Properties have a name, but setting/getting
properties is done through an "atom" (identifier of sorts) to avoid sending
the whole string on the wire for every get/set call.
See Properties and Atoms
The currently window manager (if any) can react to window property
changes by setting the appropriate event filter and acting on PropertyNotify
events, or simply inspecting the properties the window has when it gets mapped (or moved, or whatever). If the window manager knows of the _MOTIF_WM_HINT property, it'll
interpret those and (hopefully) do what you want.
(Note: I'm not entierly sure if that privilege is devolved to the window
manager, or if other windows can listen to those "PropertyNotify" events. Not
sure that's actually relevant to your question.)
So the code you have works just fine as long as your window manager knows
about the _MOTIF_WM_HINTS property.
You start by getting the atom
(identifier/shortcut) for it with XInternAtom, and setting its value via XChangeProperty()
before the window is actually drawn via MapWindow() (not sure if that would
work if you do it after the MapWindow(), that could depend on your window
manager).
[Edit: setting the .decorations part to zero clears all the decoration bits, so this requests that the window manager leave your borders the hell alone, essentially.]
I can't come up with somewhere "official" with the definition of that struct. It's defined in lib/X11/MwmUtils.h of the openmotif-2.3.3 distribution. Not sure how the .flags entry is used (can't find the code for the life of me :-/ ) but I suspect it's used to indicate which of the {decoration, function, inputMode} "sub-properties" you're acting on. (Don't take my word for that.)
As a side note, using _MOTIF_WM_HINTS might not be your best option right
now. Have you looked at the Extended Window Manager hints
specification and other information/standards/drafts over at freedesktop? I'll wager most "modern" window managers/desktop environments/kitchen sinks will tend to adhere to that rather than keeping backwards compatibility with Motif. All depends on what you're coding for I guess.
Thanks for reading :-)
Have a look at the _MOTIF_WM_HINTS property. I changed your code to a working state:
#include <X11/Xlib.h>
#include <X11/Xatom.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#define MAPIT
#undef REPEAT
void waitxevt(Display* d, int type)
{
XEvent e; /* XEvent holder */
do { XNextEvent(d, &e); } while (e.type != type);
}
void frame(Display* d, Window w, int e)
{
Atom window_type;
Atom motif_hints;
long value;
#ifdef MAPIT
XUnmapWindow(d, w);
waitxevt(d, UnmapNotify);
#endif
window_type = XInternAtom(d, "_NET_WM_WINDOW_TYPE", False);
if (e) value = XInternAtom(d, "_NET_WM_WINDOW_TYPE_NORMAL", False);
else value = XInternAtom(d, "_NET_WM_WINDOW_TYPE_DOCK", False);
XChangeProperty(d, w, window_type, XA_ATOM, 32, PropModeReplace, (unsigned char *) &value, 1);
long hints[5] = {e ? 0 : 2, 0, 0, 0, 0};
motif_hints = XInternAtom(d, "_MOTIF_WM_HINTS", False);
XChangeProperty(d, w, motif_hints, motif_hints, 32, PropModeReplace, (unsigned char *)&hints, 5);
#ifdef MAPIT
XMapWindow(d, w);
waitxevt(d, MapNotify);
#endif
}
int main(void) {
Display* d;
Window w;
XEvent e;
const char* msg = "Hello, World!";
int s;
GC gracxt;
int frmenb = 0;
d = XOpenDisplay(NULL);
if (d == NULL) {
fprintf(stderr, "Cannot open display\n");
exit(1);
}
s = DefaultScreen(d);
gracxt = XDefaultGC(d, s);
w = XCreateSimpleWindow(d, RootWindow(d, s), 10, 10, 640, 480, 5,
BlackPixel(d, s), WhitePixel(d, s));
XSelectInput(d, w, ExposureMask|KeyPressMask|StructureNotifyMask);
XMapWindow(d, w);
waitxevt(d, MapNotify);
while (1) {
XNextEvent(d, &e);
if (e.type == Expose) XDrawString(d, w, gracxt, 10, 50, msg, strlen(msg));
if (e.type == KeyPress) {
frame(d, w, frmenb);
frmenb = frmenb == 0 ? 1 : 0;
#ifdef REPEAT
frame(d, w, frmenb);
frmenb = !frmenb;
#endif
}
}
XCloseDisplay(d);
return 0;
}
The only thing changed is to include the hints.

Handle "new top level window" events in Xlib/Xt

So I'm in a situation where I need to know when a top level window gets created. I'm working at the Xlib/Xt level and on a Window Manager that doesn't support the EWMH specification. My idea is to hook into the root window's SubstructureNotify events. But things are not as simple as just that.
The problem is that not every CreateNotify event corresponds to the creation of a [b]top level[/b] window. So what I think I need to do is test the window I get from the event somehow to confirm that it is a top level window. I've got close, but some spurious windows still make it through my net. For example, in a GTK application if you have a dropdown box and you click it, a new window is created that I can't figure out how to catch and ignore. Such a window is troublesomely indistinguishable from a typical top level application window.
Here's what I have so far:
// I am omiting (tons of) cleanup code and where I set the display and toplevel variables.
Display* display;
Widget toplevel;
bool has_name(Window window)
{
XTextProperty data = XTextProperty ();
return (!XGetWMName (display, window, &data));
}
bool has_client_leader(Window window)
{
unsigned long nitems = 0;
unsigned char* data = 0;
Atom actual_type;
int actual_format;
unsigned long bytes;
// WM_CLIENT_LEADER is an interned Atom for the WM_CLIENT_LEADER property
int status = XGetWindowProperty (display, window, WM_CLIENT_LEADER, 0L, (~0L), False,
AnyPropertyType, &actual_type, &actual_format, &nitems, &bytes, &data);
if (status != Success || acutal_type == None) return false;
Window* leader = reinterpret_cast<Window*> (data);
return (*leader != 0);
}
bool has_class(Window window)
{
XClassHint data = XClassHint ();
return (!GetClassHint (display, window, &data));
}
void handle_event(Widget widget, XtPointer, XEvent* event, Boolean*)
{
if (event->type != CreateNotify) return;
Window w = event->xcreatewindow.window;
// confirm window has a name
if (!has_name (w)) return;
// confirm window is a client window
Window client = XmuClientWindow (display, w);
if (!client || client != w) return;
// confirm window has a client leader that is not 0x0
if (!has_client_leader (client)) return;
// confirm window has a class
if (!has_class (client)) return;
// The window has passed all our checks!
// Go on to do stuff with the window ...
}
int main(int argc, char* argv[])
{
// ...
// Setting up the event handler for SubstructureNotify on root window
Window root_window = XDefaultRootWindow (display);
Widget dummy = XtCreateWidget ("dummy", coreWidgetClass, toplevel, 0, 0);
XtRegisterDrawable (display, root_window, dummy);
XSelectInput (display, root_window, SubstructureNotifyMask);
XtAddRawEventHandler (dummy, SubstructureNotifyMask, False, handle_event, 0);
// ...
}
A long shot, but does anyone have any ideas I could try? I can't think of much else I can really do here.
I assume you're familiar with the ICCCM and its long-winded discussion.
Have you checked the WM_TRANSIENT_FOR property?