Community
Participate
Working Groups
The Kepler M5 parallel package hangs for me (Mac 64) when it comes up. I'm trying to rightmouse-new-sync c/c++ project. The busy-cursor (circle-quarters-icon) won't go away but i can still use menu items to create new>...>sync project and it does go away eventually. Trouble shutting it down too i think... Greg reports hangs on mac as well. Chris Navarro reports: I see the same thing with Ubuntu (64) when I try the same action sequence you did except my busy-cursor never really goes away. I can go through the action a second time while the busy cursor is still showing (right-mouse, new sync c/c++ project) and this time the wizard opens, but the original process is still running. Even after I close Eclipse (File > Exit), it doesn't shut down cleanly because that action is still hanging onto the main thread. When I try to re-open eclipse with the same workspace it says it's locked and in use until I forcefully kill the process.
Trying Kepler M4 parallel package with recent PTP 7.0 build on top of that. It has the same problem (busy-cursor stuck on new > sync project) Kepler M4 parallel package alone does not. No stuck cursor.
I get the same behavior as Chris with Juno SR2 RC3 using 32-bit opensuse linux. Investigating...
Oddly, I can't reproduce this problem in my development environment, which makes it hard to debug. I can trace the RSE initialization, which does occur at the same point, but it completes without any problems. Also, I wasn't able to find the git repository for RSE (org.eclipse.rse), which may contain a recent change that triggers the problem.
Created attachment 226887 [details] stacktrace from 4.3M5 that was force-quite (.metadata/.log file)
it may be worth hooking a monitoring tool to the hung jvm, I tried jconsole, and found 23-30 threads that may be worth investigating further -
Some history of how this seems to be getting resolved: (ptp-dev mailing list thread http://dev.eclipse.org/mhonarc/lists/ptp-dev/msg07167.html) Christopher Navarro: I did notice that if I go through the top level menus (File > New > Other > Remote sync c/c++ project), it's taking a long time with "Initializing remote services: RSE"... Greg Watson: PTP will try to initialize all the remote services it knows about, even if they are not used. So if both RSE and the PTP RSE adapter are installed, then you'll see this message. As far as I know there should be no functional changes in RSE between SR1 and SR2, but that doesn't mean there weren't. Chris Recoskie: I believe Dave Dykstal moved some of the init code in RSE for Juno SR2 RC2 (maybe it was RC3?) in order to prevent hangs we were getting with Eclipse 4.2.2. I'm not sure if this is related. We haven't encountered any issues with it here. Greg to David Dykstal: David, We have a serious regression in Juno SR2 RC3 which is causing a number of UI issues when users create new synchronized project. It seems to be related to the RSE initialization code, though it's difficult for us to debug because we're not seeing it in the development environment, only when run from the package. Do you know if there were any changes to RSE initialization in RC3 and what theses were? David Dykstal: There was a change that delayed the running of an initialization job until the workbench was fully started. That job is now scheduled in the plugin org.eclipse.rse.ui. Is it possible that you depend on this job running but do not bring up any RSE UI? Greg: Yes, this is possible. It looks like we're testing RSECorePlugin.isInitComplete(RSECorePlugin.INIT_ALL) and this is never returning true. Is your change likely to have changed the behavior of this method? David: We're have a fix and are attempting a rebuild
Fixed with resolution of 397995.
