RFR: 8326712: Robot tests fail on XWayland
Phil Race
prr at openjdk.org
Fri Jun 28 07:57:52 UTC 2024
On Wed, 26 Jun 2024 11:25:37 GMT, Alexander Zvegintsev <azvegint at openjdk.org> wrote:
> Most of the headful test failures on XWayland are due to screen capture is not working.
>
> Wayland, unlike X11, does not allow arbitrary applications to capture the screen contents directly.
> Instead, screen capture functionality is managed by the compositor, which can enforce stricter controls and permissions, requiring explicit user approval for screen capture operations.
>
> This issue is already resolved in OpenJDK ([base issue](https://bugs.openjdk.org/browse/JDK-8280982), there are subsequent fixes) by using the [ScreenCast XDG portal](https://flatpak.github.io/xdg-desktop-portal/docs/doc-org.freedesktop.portal.ScreenCast.html).
>
> The XDG ScreenCast portal is utilized for capturing screen data as it provides a secure and standardized method for applications to access screen content.
> Additionally, it allows for the reuse of user permissions without requiring repeated confirmations, streamlining the user experience and enhancing convenience.
>
>
> <hr>
>
> So this changeset is a copy of the OpenJDK fixes with the addition of the JavaFX customization.
> For ease of review, you can skip the changes in the first two commits:
> - First commit is a direct copy of files from OpenJDK
> - Second commit removes the `gtk-` prefix before the `gtk_...` and `g_...` function calls (in AWT all gtk functions are dynamically loaded, we don't need that in JavaFX).
>
> properties added:
>
> - `javafx.robot.screenshotMethod`, accepts `gtk`(existing gtk method) and `dbusScreencast`(added by this changeset, used by default for Wayland)
> - `javafx.robot.screenshotDebug` prints debug info if it is set to `true`
>
> <hr>
>
> What are the remaining issues?
>
> 1.
>
> After applying this fix, system tests will pass except for the `SwingNodeJDialogTest` test.
>
> This interop test calls `java.awt.Robot#getPixelColor` which internally `gtk->g_main_context_iteration(NULL, TRUE);` causes a blocking of javafx gtk loop, so the test hangs.
> So a change is required on OpenJDK side to fix this issue.
>
> 2.
>
> Even after solving the `#1`, the `SwingNodeJDialogTest.testNodeRemovalBeforeShow` case is still failing.
>
> 3.
>
> Internally the ScreenCast session keeps open for [2s](https://github.com/openjdk/jdk/blob/d457609f700bbb1fed233f1a04501c995852e5ac/src/java.desktop/unix/classes/sun/awt/screencast/ScreencastHelper.java#L62).
> This is to reduce overhead in case of frequent consecutive screen captures.
>
> There is a crash when an AWT ScreenCast session overlaps with the FX ScreenCast session. E.g. `java.awt.Robot#getPixelColor()` and `jav...
modules/javafx.graphics/src/main/java/com/sun/glass/ui/gtk/screencast/ScreencastHelper.java line 70:
> 68: AccessController.doPrivileged((PrivilegedAction<Boolean>) () -> {
> 69: final String str =
> 70: System.getProperty("fx.robot.screenshotDebug", "false");
fx - > javafx
modules/javafx.graphics/src/main/java/com/sun/glass/ui/gtk/screencast/TokenStorage.java line 69:
> 67:
> 68: private static final String REL_NAME =
> 69: ".fx/robot/screencast-tokens.properties";
should be .javafx
However I think we can try to READ .awt (and .java) first and if it exists try to use that.
Perhaps WRITE .javafx ?
If we don't understand what is at .awt or .java then we would then READ .javafx and if it is not there or
doesn't work CREATE .javafx
The overall intent is that if a user authorised Java for AWT then we don't need to re-ask them for FX so long as we can parse what is there and it works
modules/javafx.graphics/src/main/native-glass/gtk/fp_pipewire.h line 26:
> 24: */
> 25:
> 26: #ifdef HEADLESS
There is no HEADLESS build for JavaFX. I see the same ifdef in other C files in this PR.
I guess its harmless so I'll let Kevin decide what we should do
tests/system/src/test/java/test/robot/javafx/embed/swing/SwingNodeJDialogTest.java line 37:
> 35: public class SwingNodeJDialogTest extends SwingNodeBase {
> 36:
> 37: @Test(timeout = 15000)
Is this in milliseconds ?
-------------
PR Review Comment: https://git.openjdk.org/jfx/pull/1490#discussion_r1655461505
PR Review Comment: https://git.openjdk.org/jfx/pull/1490#discussion_r1655467715
PR Review Comment: https://git.openjdk.org/jfx/pull/1490#discussion_r1655452918
PR Review Comment: https://git.openjdk.org/jfx/pull/1490#discussion_r1655458869
More information about the openjfx-dev
mailing list