Xlib clipboard/selection requests do not play well with the powersave FPS-cap

[Meriipu]

Godot version:
Godot Engine v3.2..2.beta.custom_build.aa57bb047
aa57bb0

Edit: kind of a regression (see #39422 (comment) ). Introduced in ac14efc.

Or instead of decoupling, make the part that deals with selections and clipboards more efficient somehow? Xlib is not very enjoyable to deal with.

OS/device including version:
Linux

Issue description:

• Firefox with autoscroll enabled (pressing and holding middle click and dragging the mouse up causes page to scroll)
• press ctrl+c in the editor in the client. Does not need to have text selected
• Try to hold+drag middle in firefox, then after a few seconds release it
  For me, the page in firefox keeps scrolling for around a second before stopping scrolling.
  This issue persists until I select some text in firefox. After that, scrolling stops immediately upon release again.

Steps to reproduce:

1. Have the editor in a project and Firefox (my version is 76) open
2. Press ctrl+c in the editor
3. Press and hold middleclick in firefox, move the mouse a bit, release middleclick
   The reticle lags behind for a second or so, long after the middleclick was released
   In order to "fix it", select some text in Firefox. This fixes it until the above steps are repeated again.

[Meriipu]

I tried commenting out this line and it makes the sluggishness moderately less sluggish, but only moderately.
https://github.com/godotengine/godot/blob/master/editor/editor_node.cpp#L442

[Calinou]

This may be unrelated, but I noticed that pasting the clipboard can take a very long time in other applications, if the current clipboard data comes from Godot. (Tested on Godot 3.2.1 running KDE on Fedora 31.)

[Meriipu]

It very likely is related. I called this in a terminal

sleep 10; echo 'abc'|xclip

then caused the issue. It fixes itself when xclip is done.

It is kind of a regression, it was introduced apparently in ac14efc

I say kind of and apparently because this commit just reduces the framerate when inactive from what I can tell, so the issue was likely always there, it is just that with a reduced framerate it becomes more apparent due to how the xclipboard works [in conjunction with?] the engine?

[Calinou]

It is kind of a regression, it was introduced apparently in ac14efc

I would really prefer keeping the FPS cap when the window is unfocused (think of the laptops). We should try to find another fix for this, such as running a loop dedicated to the clipboard that's not affected by any FPS limit.

[Meriipu]

This section of code gets executed multiple times (around 8) each time a paste is attempted, either via ctrl+v or via middle-click (even autoscrolling middleclick).

godot/platform/linuxbsd/display_server_x11.cpp

Lines 2858 to 2918 in d3c10e8

	case SelectionRequest: {
	XSelectionRequestEvent *req;
	XEvent e, respond;
	e = event;
	req = &(e.xselectionrequest);
	if (req->target == XInternAtom(x11_display, "UTF8_STRING", 0) ||
	req->target == XInternAtom(x11_display, "COMPOUND_TEXT", 0) ||
	req->target == XInternAtom(x11_display, "TEXT", 0) ||
	req->target == XA_STRING ||
	req->target == XInternAtom(x11_display, "text/plain;charset=utf-8", 0) ||
	req->target == XInternAtom(x11_display, "text/plain", 0)) {
	CharString clip = clipboard_get().utf8();
	XChangeProperty(x11_display,
	req->requestor,
	req->property,
	req->target,
	8,
	PropModeReplace,
	(unsigned char *)clip.get_data(),
	clip.length());
	respond.xselection.property = req->property;
	} else if (req->target == XInternAtom(x11_display, "TARGETS", 0)) {
	Atom data[7];
	data[0] = XInternAtom(x11_display, "TARGETS", 0);
	data[1] = XInternAtom(x11_display, "UTF8_STRING", 0);
	data[2] = XInternAtom(x11_display, "COMPOUND_TEXT", 0);
	data[3] = XInternAtom(x11_display, "TEXT", 0);
	data[4] = XA_STRING;
	data[5] = XInternAtom(x11_display, "text/plain;charset=utf-8", 0);
	data[6] = XInternAtom(x11_display, "text/plain", 0);
	XChangeProperty(x11_display,
	req->requestor,
	req->property,
	XA_ATOM,
	32,
	PropModeReplace,
	(unsigned char *)&data,
	sizeof(data) / sizeof(data[0]));
	respond.xselection.property = req->property;
	} else {
	char *targetname = XGetAtomName(x11_display, req->target);
	printf("No Target '%s'\n", targetname);
	if (targetname) {
	XFree(targetname);
	}
	respond.xselection.property = None;
	}
	respond.xselection.type = SelectionNotify;
	respond.xselection.display = req->display;
	respond.xselection.requestor = req->requestor;
	respond.xselection.selection = req->selection;
	respond.xselection.target = req->target;
	respond.xselection.time = req->time;
	XSendEvent(x11_display, req->requestor, True, NoEventMask, &respond);
	XFlush(x11_display);
	} break;

[Meriipu]

but yes I think it takes some guru to fix this one