This Bugzilla instance is a read-only archive of historic NetBeans bug reports. To report a bug in NetBeans please follow the project's instructions for reporting issues.
Summary: | "Unexpected Failure: Suite is not finished" (testDebugging: unknown) during commit validation | ||
---|---|---|---|
Product: | obsolete | Reporter: | Jesse Glick <jglick> |
Component: | xtest | Assignee: | issues@obsolete <issues> |
Status: | RESOLVED INVALID | ||
Severity: | blocker | Keywords: | TEST |
Priority: | P2 | ||
Version: | 3.x | ||
Hardware: | PC | ||
OS: | Linux | ||
Issue Type: | DEFECT | Exception Reporter: | |
Attachments: | Complete results (.tar.gz) |
Description
Jesse Glick
2003-06-05 17:46:04 UTC
Created attachment 10608 [details]
Complete results (.tar.gz)
I just noticed this in the build log: executor: Executing module core, testtype unit at Jun 5, 2003 12:36:00 PM Output is redirected to /space/src/nb_all/xtest/instance/results/testrun_030605-123600/logs/core_unit.log Buildfile: /space/src/nb_all/xtest/lib/harness.xml Executed successfully. Executing module core/windows, testtype unit at Jun 5, 2003 12:37:09 PM Output is redirected to /space/src/nb_all/xtest/instance/results/testrun_030605-123600/logs/core_windows_unit.log Buildfile: /space/src/nb_all/xtest/lib/harness.xml Executed successfully. Executing module openide, testtype unit at Jun 5, 2003 12:37:49 PM Output is redirected to /space/src/nb_all/xtest/instance/results/testrun_030605-123600/logs/openide_unit.log Buildfile: /space/src/nb_all/xtest/lib/harness.xml Executed successfully. Executing module ide, testtype qa-functional at Jun 5, 2003 12:38:31 PM Output is redirected to /space/src/nb_all/xtest/instance/results/testrun_030605-123600/logs/ide_qa-functional.log Buildfile: /space/src/nb_all/xtest/lib/harness.xml X Error of failed request: BadMatch (invalid parameter attributes) Major opcode of failed request: 73 (X_GetImage) Serial number of failed request: 186682 Current serial number in output stream: 186682 X connection to :69.0 broken (explicit kill or server shutdown). Executed successfully. Regenerating report's XMLs Transforming report's XMLs to HTMLs So maybe my use of a nested X server is causing the problem? Hard to know. Seems to have worked before, except under some conditions that I no longer remember - maybe a JDK issue. I guess you are right that java.awt.Robot can't connect to XServer. From logs it seems that testDebugging test failed. In that case our infrastructure creates a screen shot using java.awt.Robot. Something wrong in your environment? Looks like there really was some problem with the nested X server when runnning debugger tests in the QA's IDE validation suite. It's hard to say which component fault it is, but I doubt it was XTest's fault. From ant build script log of IDE validation suite, I can get this: <snip just after the IDE started> [ideexec] X connection to :69.0 broken (explicit kill or server shutdown). <now the script continues as IDE is normally closed> So IMHO it is rather bug of the nested X server (it definitely was somehow shut/crashed down), you might try to search for the cause in its log (Xerrors?). Also please try to run the commit validation suite again some other X server (on Linux you can start another X server (:1), so the tests does not interfiere with your desktop). Anyway, for now I'm closing it as 'invalid', because this really does not seem to be XTest's bug. (BTW did you get 'build failed' and some messages after printing out the summary, because of the unfinished suite?) Currently I am trying a non-nested X server (X :69 rather than Xnest :69 - both as of XFree86 4.2.0). While Xnest sometimes gives this BadOpcode thing (again I can't quite remember under which conditions), X seems to sometimes dump core (SIGSEGV). :-( Running with the main X server is pretty much out of the question, unless I want to spend an hour every day making coffee - especially if you run more than just the commit validation suite, the ide-mode tests running in SDI mode prevent normal work (and if you so much as touch the mouse while they are running, they will randomly fail). I think I was some time ago using just X (i.e. switchable new display) but ran into the SIGSEGV, so switched to Xnest, and somehow got it to work more or less reliably. But I am not sure what the critical factor was. Maybe choice of JDK. I am considering reopening issue #20829 if I cannot find any decent workaround. JDK 1.4.0 seems to work better with Xnest, at least so far - no BadOpcode's yet. Possible 1.4.2 bug, who knows. (1.4.1 is undesirable for building with generally because the javadoc tool is pretty much broken.) I've experimented with VNC server on Solaris and Linux. It seemed to be more stable than Xnest. Below is a description how to run vncserver on Solaris. On linux it is the same but you have to start your favorite window manager by xstartup script. ------------------------------------------------- It is really enough to run vncserver by command "vncserver -nolisten local". There should be automaticaly created directory ".vnc" in your home directory with file "xstartup" after running vncserver command. By default, this file contains: -------------------------------------------->8--------------------------------- #!/bin/sh xrdb $HOME/.Xresources xsetroot -solid grey xterm -geometry 80x24+10+10 -ls -title "$VNCDESKTOPDesktop" & twm & -------------------------------------------->8--------------------------------- If you wold like to run CDE, your xtsratup file has to look like below: -------------------------------------------->8--------------------------------- #!/bin/sh exec /usr/dt/bin/Xsession ----------------- I will play with VNC. Thanks for the tip. I find that I get the same BadOpcode errors sometimes with JDK 1.4.0_04. VNC seems to work better; we'll see. I added direct support to nbbuild/misc/full-build-linux.sh. |