Community
Participate
Working Groups
We are going to do the migration after GEF 3.4 M4
*** Bug 214552 has been marked as a duplicate of this bug. ***
Still stuck trying to run tests: Failed to invoke suite():org.eclipse.swt.SWTError: No more handles [gtk_init_check() failed] http://emf.torolab.ibm.com/gef/gef/downloads/drops/3.4.0/N200801141138/buildlog.txt > Check out https://bugs.eclipse.org/bugs/show_bug.cgi?id=115822 > 1. The DISPLAY environment variable is set. > 2. The host/port information is correct (e.g. added to /etc/hosts). > 3. The X server is up. 3. X is up -- kdm is running and a working xorg.conf has been installed. Running `ps ax --forest`, I get in part: 1 ? S 0:01 init [5] 4015 ? Ss 0:00 /usr/bin/kdm 4025 tty7 Ss+ 0:01 \_ /usr/bin/X -br -nolisten tcp :0 vt7 -auth /var/run/xauth/A:0-5vtRiz 4043 ? S 0:00 \_ -:0 4052 ? S 0:00 \_ /usr/bin/kdm_greet Should that "-nolisten" be there? 1. DISPLAY is being set right before invoking the tests: export DISPLAY=":0.0" 2. What does that mean? Where in /etc/hosts do you define X server ports? $ cat /etc/hosts 127.0.0.1 localhost 9.26.101.101 emf.torolab.ibm.com emf # The following lines are desirable for IPv6 capable hosts # (added automatically by netbase upgrade) ::1 ip6-localhost ip6-loopback fe00::0 ip6-localnet ff00::0 ip6-mcastprefix ff02::1 ip6-allnodes ff02::2 ip6-allrouters ff02::3 ip6-allhosts I've even added -Dosgi.bundlefile.limit=100 and export CVS_RSH=ssh ulimit -c unlimited export JAVA_HIGH_ZIPFDS=500 export LANG=C and nothing (yet) works. Failure is happening running both ui-test and core-test ant targets, suggesting it's maybe not an x server issue, but a java / threading issue? I'm stumped. Any suggestions?
Boo-freakin'-yeah! JUnit tests now pass. X server works. I'm even switched the tests over to be "ui-test" instead of "core-test", and they still work! I've updated the wiki with lessons learned here: http://wiki.eclipse.org/EMF_Build_Server_Setup#Verify_X_server And I've also unearthed 637 new CHKPII errors. However, as 633 of are javadoc problems, I've added reference/api/*.html to the CHKPII skiplist. Also found invalid XML in topics_Samples.xml, which I've fixed. There are three issues remaining, all from the gef.doc plugin: http://emf.torolab.ibm.com/gef/downloads/drops/3.4.0/N200801150106/testresults/chkpii/org.eclipse.nls.htm.xml Failing files are: /about.html /guide/guide.html /guide/index.html However, CHKPII doesn't find these now that I've fixed the fact that the doc plugin wasn't being jarred. Anthony: do you want to fix these? Meanwhile... playing with different window managers and headless X server connections have not been entirely successful -- yet. In one case, changes in relengbuildgtk.sh caused the build's tests to not run at all [1]; in another, we had 4 failures [2]. [1]http://emf.torolab.ibm.com/gef/downloads/?showAll=1&hlbuild=N200801150223&sortBy=date&project=gef#N200801150223 (see build log; search for "Java Result: 13") [2]http://emf.torolab.ibm.com/gef/gef/downloads/drops/3.4.0/N200801150313/testresults/html/org.eclipse.draw2d.test_linux.gtk.html
(In reply to comment #4) > JUnit tests now pass. X server works. I'm even switched the tests over to be > "ui-test" instead of "core-test", and they still work! Good stuff > Anthony: do you want to fix these? Yes, we should raise a separate Bugzilla for these issues.
Guys: After much debugging, I can't figure that there's anything wrong with the build system in terms of its ability to run the tests. I've done/undone everything I changed since the last good build, and I still get 4 JUnit failures. I've even reverted the mapfile to its 1/15 2:06am state to see if by chance something broke along the way. Maybe it's because I've using a newer JDK, a newer kernel, or a newer basebuilder/Eclipse. I'm not sure. What's on your existing build server? java version "1.4.2_15" Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_15-b02) Java HotSpot(TM) Client VM (build 1.4.2_15-b02, mixed mode) Linux emf 2.6.15-26-server #1 SMP Fri Sep 8 21:00:37 UTC 2006 i686 GNU/Linux (I've tried three versions: "v20071108", "M4_34", "vI20070503-1400" -- all have been proven to work at one point or another, despite using different versions of the Equinox launcher. I've tried 3.4M4, I20080108, and I20080115. All work, except for the 4 JUnit failures.) I do see some changes that sound like they're related to the JUnit failures -- perhaps they only manifest on Linux (as you told me you couldn't reproduce them on Windows). Here's the 4 commits done by Pratik (cvs user pshah) in/around the time I was working on the build: http://www.eclipse.org/modeling/emf/searchcvs.php?q=file%3A%25gef%25+author%3Apshah+days%3A7&project=0&fullpath=Y While you figure out if there's in fact a linux gtk bug in there, I'll try to get the build running on my thinkpad and also run a build from within a full UI Eclipse instance. If you want to build and promote using your old system this week, you're more than welcome. I've got a number of other things to set up before the web will be ready anyway... Oh, and don't forget, there's also the 24 compiler warnings... do you get those with your old builder? If not, I wonder why not.
(In reply to comment #6) > [...] I still get 4 JUnit failures. [...] As discussed, I also got errors when I tried to change the build machine from an older RedHat to Ubuntu. There may be a GTK issue at work here. The latest code and JUnits work for me on Windows and RedHat, so I will test on Ubuntu when I get a chance. This is not a show stopper in my opinion. > If you want to build and promote using your old system this week, Will do for the January 16 weekly integration build > Oh, and don't forget, there's also the 24 compiler warnings... do you get those > with your old builder? If not, I wonder why not. Yes, there are a number of warnings.
Remaining items here are: * mapfile generation not working (but GEF doesn't use that so I'm not concerned here; obvious workaround if you want a "generated" mapfile is to set the map to use HEAD as the tag, commit to CVS, and build from that) * -addSDK flag not working -- the purpose here is to add GEF as a dependency to the build page for others to consume; but as GEF is build on emf.toro and the only projects that consume it are built on emft.eclipse, this wouldn't benefit them anyway * Bug 216984: 4 JUnit test failures And as I just sorted out Jar signing and packing (bug 173651), I think we're done here.
