[rfc] [icedtea-web] Fix EDT hanging on OpenJDK 11

[Jiri Vanek]

On 9/11/18 2:46 PM, Laurent Bourgès wrote:
> Jiri & Mario,
> 
> Here is my final patch (after merging with backports):
> http://cr.openjdk.java.net/~lbourges/itw/patch_EDT.log.3
> 
> Laurent


ok. I have the patch ready to push. The private class added, and one import removed.
Here is the changelogused:

2018-09-11  Laurent Bourgès <bourges.laurent at gmail.com>

	Fixed EDT hanging
	* AUTHORS: added Laurent
	* netx/net/sourceforge/jnlp/GuiLaunchHandler.java: invokeLater and invokeAndWait moved from 
SwingUtilities local wrapper of SwingUtils.
	Simplified splasshcreen loading
	* netx/net/sourceforge/jnlp/JNLPSplashScreen.java: added name to the dialog
	* netx/net/sourceforge/jnlp/Launcher.java: added log entry for main class loading. Highlighted 
usage of SwingUtilities on top SwingUtils
	for this particular space
	* netx/net/sourceforge/jnlp/about/AboutDialog.java: added name.  Highlighted usage of 
SwingUtilities on top SwingUtils
	for this particular space
	* netx/net/sourceforge/jnlp/cache/DefaultDownloadIndicator.java: reworked to work in 
SwingUtils.invokeAndWait. Added name.
	* netx/net/sourceforge/jnlp/controlpanel/AdvancedProxySettingsDialog.java: added name
	* netx/net/sourceforge/jnlp/controlpanel/CacheViewer.java: added name
	* netx/net/sourceforge/jnlp/controlpanel/ControlPanel.java: added top level setup for enabling 
SwingUtils debugging abilities
	added log entry for main class loading. Highlighted usage of SwingUtilities on top SwingUtils for 
this particular space
	* netx/net/sourceforge/jnlp/runtime/Boot.java: added top level setup for enabling SwingUtils 
debugging abilities
	CertificateViwer moved to invokeAndWait. added log entry for main class loading. Highlighted usage 
of SwingUtilities on top SwingUtils
	for this particular space.
	* netx/net/sourceforge/jnlp/runtime/HtmlBoot.java:invokeLater and invokeAndWait moved from 
SwingUtilities local wrapper of SwingUtils
	* netx/net/sourceforge/jnlp/runtime/JNLPRuntime.java: removed import of import javax.swing.JWindow
	* netx/net/sourceforge/jnlp/runtime/JNLPSecurityManager.java: direct call to JWindow().getOwner(), 
moved to  SwingUtils.getOrCreateWindowOwner.
	Its existence doubtful
	* netx/net/sourceforge/jnlp/security/SecurityDialog.java: installPanel moved to invokeAndWait
	* netx/net/sourceforge/jnlp/security/SecurityDialogs.java: used SwingUtils instead of 
SwingUtilities, added name
	* netx/net/sourceforge/jnlp/security/dialogs/ViwableDialog.java: swing ops moved to invokeAndWait
	* netx/net/sourceforge/jnlp/security/viewer/CertificateViewer.java: calling Swingutils.setup, 
astill al the same
	* netx/net/sourceforge/jnlp/splashscreen/impls/defaultsplashscreen2012/BasePainter.java: still the same
	* netx/net/sourceforge/jnlp/util/BasicExceptionDialog.java: same - name, SwingUtils x 
SwingUtilities, and proepr usage of invokes
	* netx/net/sourceforge/jnlp/util/FileUtils.java: SwingUtils x SwingUtilities
	* netx/net/sourceforge/jnlp/util/logging/JavaConsole.java: observable.notifyObservers moved to 
invokeLater and
	and invoked only if observable.hasChanged
	* netx/net/sourceforge/swing/SwingUtils.java: new class. Wrapper around most commonly used 
SwingUtilities calls. Add debugging bridge
	property of icedtea-web.edt.debug set to true, can enable it on runtime.
	* netx/net/sourceforge/swing/ThreadCheckingRepaintManager.java: Based on 
http://weblogs.java.net/blog/alexfromsun/archive/2006/02/debugging_swing.html
	debugging repaint manager
	* netx/net/sourceforge/swing/TracingEventQueue.java: new class to debug EDT - showing very long 
event processing (hanging or EDT blocked ?)



I will wait with push until Mario tell his opinion too.

thank you very much for !
   J.
> 
> Le mar. 11 sept. 2018 à 08:48, Laurent Bourgès <bourges.laurent at gmail.com 
> <mailto:bourges.laurent at gmail.com>> a écrit :
> 
>     Jiri,
> 
> 
>          >     Please note http://icedtea.classpath.org/hg/icedtea-web/rev/5290684409aa it will
>         interact with the
>          >     patch for bleeding edge.  Maybe it is worthy to backport it to 1.7 and to adapt your
>         patch on
>          >     top of it?
>          >
>          >
>          > Please backport to 1.7.
>         done. Plus few more.
>         Head and 1.7 are now very very similar. All except rust launchers.
> 
> 
>     Thanks a lot, I merged and it works well.
> 
> 
>          > I merged so the patch should not have these changes after backport:
> 
>         So your patch now do not apply correctly, but the two rejected hunks can be happily
>         ignored/removed.
>         Feel free to provide updated version, but not necessary.
> 
> 
>     I agree you should just ignore changes to JNLPRuntime.
>     I can post an updated patch for tracability ?
> 
> 
>          > I tested again and both 'javaws -viewer' and itw-settings dialogs are OK.
>          >
> 
>         Nothing more on my nitpicking list. Unless you or Mario will bring some news tomorrow, I
>         will apply
>         it to 1.7 and will look forward for patch for head, and other patches.
> 
> 
>     Excellent: my first patch is accepted.
> 
>         Note, that I'm on vacation 13-23.9, and I will be a bit delayed with replies. Lets guess a week
>         delayed :)
> 
> 
>     Enjoy holidays, I will have more time to improve shell scripts.
>     Probably I will setup an ITW mirror on github following your instructions (license...).
> 
> 
>         I will write an changelog on your behalf. If you wish to write it on your own, feel free.
> 
> 
>     I can review the update to the change log...
> 
>     Laurent
> 
> 
> 
> -- 
> -- 
> Laurent Bourgès


-- 
Jiri Vanek
Senior QE engineer, OpenJDK QE lead, Mgr.
Red Hat Czech
jvanek at redhat.com    M: +420775390109