Community
Participate
Working Groups
Created attachment 261177 [details] Crash trace Attempted using this environment variable to mitigate some UI problems that popped upon upgrading up to ubuntu 16.04. The IDE crashes with a trace, after bringing up the splash screen and the initial window. Using java 8u77 from Oracle, x86_64 host, ubuntu 16.04.
Moving to SWT: I haven't seen reports like this error before. The version of GTK3 that ships with 16.04 is known to cause issues. Eclipse Neon, which ships officially in June, is known to work very well on 16.04. The milestone builds are very stable: https://www.eclipse.org/downloads/index-developer.php
Thanks for the bug report! I just tried this on Ubuntu 16.04 with 4.5.2 as well as 4.6 (recent I-build). I'm able to launch and run eclipse both with and without using SWT_GTK3=0, not sure if something went wrong during the upgrading of Ubuntu that is causing this issue, I tried on a fresh install of Ubuntu 16.04. Would it be possible for you to try on a fresh install of Ubuntu to confirm? Also, can you attach the hs_err_*pid* log file that gets generated when eclipse crashes?
I did a fresh re-install, and I have no problems anymore, with or without the environment variable. Thanks.
Will have to do some work to bring up the old environment, and look for or generate the hs_err* files.
Looking at the attached log the problem is caused by /usr/lib/browser-plugin-freshplayer-pepperflash/libfreshwrapper-flashplayer.so which is compiled against Gtk2(IIRC) and due to Eclipse using webkit 1.x api plugins are loaded in process thus resulting in gtk3 and gtk2 loaded in same process leading to the crash. We have hit many issues wiht browser plugins and until a move to webkit 2.x api is completed to run them out of process eclipse is open to browser plugins taking it down.
(In reply to Antoun Kanawati from comment #3) > I did a fresh re-install, and I have no problems anymore, with or without > the environment variable. > > Thanks. Thanks for the response Antoun! I propose to close this bug as WORKSFORME now, can be reopened if necessary.
(In reply to Alexander Kurtakov from comment #5) > Looking at the attached log the problem is caused by > /usr/lib/browser-plugin-freshplayer-pepperflash/libfreshwrapper-flashplayer. > so which is compiled against Gtk2(IIRC) and due to Eclipse using webkit 1.x > api plugins are loaded in process thus resulting in gtk3 and gtk2 loaded in > same process leading to the crash. We have hit many issues wiht browser > plugins and until a move to webkit 2.x api is completed to run them out of > process eclipse is open to browser plugins taking it down. Alex, I'm probably missing something here but I don't seem to understand how loading the freshplayerplugin could be the cause for this crash. There cannot be a mix up of GTK3 and GTK2 libraries here as eclipse is loading the GTK2 libraries (SWT_GTK3=0) and libfreshwrapper-flashplayer.so is also dependent on the same, where/how would GTK3 be coming into the picture here? I have the same doubt regarding bug 492045 as well as the crash log seems to indicate eclipse running in GTK2 mode.
I think this is probably a "normal crash" in libfreshwrapper-flashplayer.so, not caused by a mix-up of GTK versions.
I can reproduce the crash: Ubuntu 16.04 64-bit Eclipse Neon.1 Install package: browser-plugin-freshplayer-pepperflash SWT_GTK3=0 Start a new workspace or open the welcome screen.
Whoever experiences crashes caused by some browser plugings installed has single option - use GTK3 and enforce webkit2 (via SWT_WEBKIT2=1 env variable). Reason for that is that Webkit2 runs plugins out of process so plugin crashes do not bring the whole webkit down and as a result swt too. Webkit1 runs browser plugins in-process so there is nothing we can do and there is no Webkit2 for gtk2. Other option is to uninstall the plugin in question. Resolving the bug as wontfix due to the previous.