[WIP] thread safety

[memoryleak47]

This adds the thread_safety::assert_window_thread-function, which panics if called from some other thread than the window-thread.
There is still a lot of room for experimentation in order to find out what functions should call thread_safety::assert_window_thread, or whether this mechanism is too restrictive or relaxed.
cc #226

[crumblingstatue]

This simple example still produces varying errors

    std::thread::spawn(|| {
        let mut rw = RenderWindow::new((640, 480), "Hi", Style::CLOSE, &Default::default());
    });
    println!("{}", Key::K.is_pressed());

This is likely because is_pressed() also inits some kind of windowy context (opens a connection to the X server on Linux at least), and it can run before the code in the thread::spawn above it, so it really should also set the window thread as well. Probably almost everything should set the window thread instead of just merely asserting it.

[crumblingstatue]

Do you have any objection to just getting rid of assert_window_thread entirely and just using set_window_thread everywhere?

diff --git a/src/thread_safety.rs b/src/thread_safety.rs
index 96c1e66..894cb77 100644
--- a/src/thread_safety.rs
+++ b/src/thread_safety.rs
@@ -12,20 +12,8 @@ pub(super) fn set_window_thread() {
         None => WINDOW_THREAD.set(current_id).unwrap(),
         Some(id) => {
             if *id != current_id {
-                panic!("Opening Windows in multiple threads is not supported.")
+                panic!("Graphics/windowing functionality is only supported on the same thread.")
             }
         }
     }
 }
-
-// asserts that the either current thread is WINDOW_THREAD, or that WINDOW_THREAD is unset
-pub(super) fn assert_window_thread(func_name: &str) {
-    if let Some(&id) = WINDOW_THREAD.get() {
-        if id != current().id() {
-            panic!(
-                "{} has to be called in the same thread as the RenderWindow!",
-                func_name
-            );
-        }
-    }
-}
diff --git a/src/window/keyboard.rs b/src/window/keyboard.rs
index 0da0417..b5c3552 100644
--- a/src/window/keyboard.rs
+++ b/src/window/keyboard.rs
@@ -127,7 +127,7 @@ impl Key {
     /// triggered.
     #[must_use]
     pub fn is_pressed(self) -> bool {
-        thread_safety::assert_window_thread("Key::is_pressed()");
+        thread_safety::set_window_thread();
 
         unsafe { ffi::sfKeyboard_isKeyPressed(self as _) }.to_bool()
     }
diff --git a/src/window/mouse.rs b/src/window/mouse.rs
index ca9aa57..b4b440c 100644
--- a/src/window/mouse.rs
+++ b/src/window/mouse.rs
@@ -77,7 +77,7 @@ impl Button {
     /// triggered.
     #[must_use]
     pub fn is_pressed(self) -> bool {
-        thread_safety::assert_window_thread("Button::is_pressed()");
+        thread_safety::set_window_thread();
 
         unsafe { ffi::sfMouse_isButtonPressed(self.raw()) }.to_bool()
     }
diff --git a/src/window/video_mode.rs b/src/window/video_mode.rs
index 6286ea4..9495f8f 100644
--- a/src/window/video_mode.rs
+++ b/src/window/video_mode.rs
@@ -79,7 +79,7 @@ impl VideoMode {
     /// return the urrent desktop video mode
     #[must_use]
     pub fn desktop_mode() -> Self {
-        thread_safety::assert_window_thread("VideoMode::desktop_mode()");
+        thread_safety::set_window_thread();
 
         unsafe { Self::from_raw(ffi::sfVideoMode_getDesktopMode()) }
     }

[memoryleak47]

Sure, go ahead!

[crumblingstatue]

I meant you could do it, unless you want me to make another pull request based on this, and close this one.

Although I guess I could submit a pull request to your fork, lol.

[crumblingstatue]

Thank you, I'll merge this as a good first step towards thread safety

[memoryleak47]

Merging this is maybe a little early;
As we are only certain that this problem arises when used with X11, I could image that a fair amount of non-X11 users which use multi-threading will now have their code broken.
Maybe we should add #[cfg(target_os = "linux")].

[crumblingstatue]

That would mean that the application they made would work on non-linux platforms, but panic on Linux. Shouldn't the panic be universal so they design their application in a way that works on all platforms?

SFML is meant to be cross-platform and work as similar as possible on all platforms.