Community
Participate
Working Groups
I'm working on Nebula webstart example. On OSX / cocoa, mousedown events are not fired on canvas (Gallery custom widget) : I cannot select on item on the gallery example tab. but they work for : OSX /cocoa without webstart Win32 both with and without webstart the jnlp file includes -XstartOnFirstThreat See : http://nricheton.homeip.net/public/nebula/nebula.jnlp
This demo seems very large (>18Mb). Can you attach a simple (SWT only, no nebula) webstart sample that shows the problem ? Silenio, do you we support webstart for cocoa ?
I won't be able to create a new webstart application before tomorrow. In the meantime the nebula app can still be used to view this problem. I know this is not the best , but at least this shows the problem. :)
Cocoa SWT isn't going to work with the AWT yet because the AWT has already installed its own NSApplication. I have some local changes related to SWT/AWT embedding that should address this -- we need to call the superclass NSApplication methods we override if the event goes to a window not in the SWT's control.
hoi scott. can you please commit your fix or provide a patch. My application is almost unusable due to this bug (started with JWS on mac/cocoa). Btw. StyledText is also affected :(
Created attachment 135165 [details] Display patch for AWT events This is a patch on Display that should help. If there is an NSApplication in place it overrides it on the fly, and if it doesn't find a Display because the wrong thread is running, it calls the superclass handler.
the patch works like a charm. Thanks. here in action: http://www.brainpot.de/jws/brainpot.jnlp
(In reply to comment #6) > the patch works like a charm. Thanks. here in action: > http://www.brainpot.de/jws/brainpot.jnlp Excellent! Kevin, Silenio, I think we should consider this for RC1 or RC2 -- AWT coexistence (in this case, with Web Start) is a pretty important feature.
The callbacks (applicationCallback2, etc) are static, so they should be created only once. Why we need to save the old impl (curr_sendEvent, etc). It seems that they should be used in Display.applicationProc() instead of calling super. Should they be put back when the display is disposed? In the isEmbedded case, should we call method_setImplementation() instead of class_addMethod()? What is the difference? Does AWT subclasses NSApplication as well? When we call super in Display.applicationProc(), are we bypassing AWT? Sorry, I could get some of the answers myself, but I have not been able to run the code yet.
The patch has some problems and we don't have time to address them all. I suggeset we defer to 3.5.1 or 3.6. I believe there are too many code changes for 3.5 RC3.
*** Bug 287379 has been marked as a duplicate of this bug. ***
Fixed > 20090828
Silenio, in Display.java I see: static int /*long*/ applicationProc(int /*long*/ id, int /*long*/ sel, int /*long*/ arg0) { //TODO optimize getting the display Display display = getCurrent (); if (display == null) return 0; if (display == null) { Did you mean to leave in the first "display == null" check? The next if test never executes in that case.
Good catch. Thanks!
It works fine ! Thanks.
Is this patch release in any of the build ? I have logged similar kind of issue my bug id is 288219. If the patch is not released please let me know where to get jar which has this patch. Your help in this issue is desparately needed. Thanks!
This build has the fix. http://download.eclipse.org/eclipse/downloads/drops/I20090901-0800/index.php
(In reply to comment #16) > This build has the fix. > > http://download.eclipse.org/eclipse/downloads/drops/I20090901-0800/index.php This link is dead (404). It would be great if someone could provide a working link to a version with this patch.
(In reply to comment #17) > This link is dead (404). It would be great if someone could provide a working > link to a version with this patch. Pick any of the recent Integration builds -- the fix has been checked in for some time now. Even a 3.6 M build will have it.
Bug was fixed but the patch wasn't used, so I'm cleaning up the review flag.
