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.
In recent builds stopped to work javadoc in code completion and go to source functionality for NetBeans modules. To reproduce: - run IDE with empty userdir - open e.g. nb_all/ant project and wait until scanning finishes - open e.g. org.apache.tools.ant.module.AntCustomizer.java - move cursor to 39:26 position (somewhere in DialogDescriptor import) - invoke code completion (Ctrl+Space) and "Javadoc not found." message is shown in popup. - try Alt+G or Alt+O at the same position but only message "Source file for org.openide.DialogDescripto not found." is printed at status bar. Build 200512132030, JDK1.5.0_06, WindowsXP.
Created attachment 27828 [details] IDE log with option -J-Dorg.netbeans.api.java.queries.JavadocForBinaryQuery=0
Works for me on linux with fresh build. Tomasi are you able to reproduce on Windows? Thanks.
true, it's reproducible for me in "5.1" 2005121319 on WinXP
Reprodicible also for external modules. In netbeans-5_0-daily-bin-200512061900-6_Dec_2005_1900.zip works, in netbeans-5_0-daily-bin-200512072015-7_Dec_2005_2015.zip does not. Some change between must have broken it.
Did you have associated sources and javadoc when trying it with external modules? Trying it with fresh build would be helpful. There was a P1 fixed today. Not sure if it can be a culprit.
Aha, it matter *which* sources i associate with platform. I downloaded ~100MB sources zip file, and it works. But if I associate the same sources, unzipped to a dir, it does not works. Martine, any change in NBPM which could break this??
Yes, my steps: 1) fresh userdir 2) associate sources in NBPM 3) create module, set dependency to e.g. Filesystems API 4) create new class, type some class and try ALT+O, ALT+G As i said, it depends on type of sources which i do associate.
When you check out whole NB cvs tree, open an e.g. mentioned NB_CVS/ant module, you shouldn't need to associate any sources AFAIK. We can try it together tomorrow Tomasi on your machine. Jesse feel free to reassign if you don't have any fast ideas.... I'm working on that leak (70032) - trying to play with GUI vs. assertGC unsuccessfully so far - but off topic here, I suppose :)
> Martine, any change in NBPM which could break this?? I'm not aware of any.
Working fine for me in a dev build. Try running with -J-Dorg.netbeans.api.java.queries.SourceForBinaryQuery=0 if you can still reproduce. Do you have other stuff checked out, like nbbuild and openide and so on?
Sorry, did not see preceding comments for some reason. If there is still a bug reproducible in current dev builds, please include complete steps to reproduce from scratch. And specify on which operating systems you can reproduce, and include a fresh diagnostic log. The preceding comments leave it ambiguous what exactly you are talking about and who can reproduce what.
1) using build 5.0 200512132030, on Windows XP, with installation 2) downloaded zip file sources to this build 3) unpacking zip to some folder. 4) clear userdir 5) starting ide, setting sources association in NBPM 6) creating new standalone module, setting it's dependency to e.g. Filesytems api 7) creating new class in that module, typing e.g. "FileObject x;" 8) invoking ALT+G and ALT+O -> does not work note that if i associate not folder with sources, but zip, it works.
> musilt2 Wed Dec 14 16:20:41 This is rather another issue (not P2 I think). Original reporter complained about modules in NB cvs tree regression. Should be filed separatelly (assigned to me), since I'm not sure if anybody checked this before at all.
>This is rather another issue (not P2 I think). Original reporter complained >about modules in NB cvs tree regression. Should be filed separatelly (assigned >to me), since I'm not sure if anybody checked this before at all. nope. this is only another testcase, which reproduces *the same bug*. I am able to reproduce also the original testcase. I'll attach log produced after my "steps to reproduce"
Created attachment 27853 [details] Log file after my testcase
Working fine for me using last instructions from musilt2, on Linux with JDK 5.0u6. Downloaded netbeans-5_0-daily-bin-200512132030-13_Dec_2005_2030.zip netbeans-5_0-daily-src-200512132030-ide_sources-13_Dec_2005_2030.zip Unpacked them so that we have e.g. -rw-r--r-- 1 jglick jglick 422650 Dec 13 22:46 /tmp/70387/netbeans-200512132030/platform6/core/org-openide-filesystems.jar -rw-r--r-- 1 jglick jglick 615 Dec 13 16:17 /tmp/70387/netbeans-src-200512132030/openide/fs/build.xml Started IDE: cd /tmp/70387 ./netbeans-200512132030/bin/netbeans --userdir testud -J-Dorg.netbeans.api.java.queries.SourceForBinaryQuery=0 -J-Dorg.netbeans.modules.apisupport.project=0 In NB Plat Mgr, set Sources for Default Platform to /tmp/70387/netbeans-src-200512132030. Made new module /tmp/70387/module1, added FS API as a dep, created import org.openide.filesystems.FileObject; public class X { public X() { FileObject x; } } and code completion and Go To Source both work fine on FileObject. Perhaps you accidentally configured the source association with the directory containing netbeans-src, rather than the directory netbeans-src itself? Just a guess.
Created attachment 27856 [details] My log file, for purposes of comparison
I just added some diagnostics (-J-Dorg.netbeans.modules.apisupport.project=0) which would be helpful if this can be reproduced in a future dev build: committed Up-To-Date 1.28 apisupport/project/src/org/netbeans/modules/apisupport/project/universe/NbPlatform.java
As both guys Tomas and Jirka mentioned this problem appears on Win XP. OS of this issue is set as Win XP. Jesse and Martine why don't you try to reproduce this issue on reported platform instead of resolving as worksforme? It is obviously problem of just Windows, not Linux. I am reopening this issue because it is easily reproducible on Win XP. I hope my comment helps to clarify.
Due to misuse of file separators in patch to SourceForBinaryImpl for issue #69735.
Patch checked that the target JAR file actually had the same suffix including cluster as the expected one, e.g. platform6/modules/org-netbeans-modules-whatever.jar It used PropertyUtils.relativizeFile which always produced a '/'-separated path. This works on Unix but not Windows. Have not tested fix. Would have to waste several more hours getting a usable development environment on my Windows partition. Hopefully someone else would be able to test quickly. Ought to be able to change clusterPath = PropertyUtils.relativizeFile(cluster.getParentFile(), project.getModuleJarLocation()); to clusterPath = PropertyUtils.relativizeFile(cluster.getParentFile(), project.getModuleJarLocation()).replace('/', File.separatorChar);
Than it would be strange if all of our SourceForBinaryImplTest pass. Maybe that just nobody runs them on Windows since that fix. Probably testFindSourceRootForModuleJar() and testExternalModules() should fail. I'm also unable to test it right now. To be sure, I'll check it in cca 6 hours with Radek or Tomas, depending on who will be in the office :) Thanks to all for investigation.
I tested patch by Jesse in build 200512142030 (release50) and it works perfectly. I suggest to merge it into release50 branch because it is a big usability issue and fix is really simple.
Ok, thanks for your help Jirko. I'll integrate it soon (also into release50 branch)
Patch applied. Checking in queries/SourceForBinaryImpl.java; 1.4 -> 1.5;
*** Issue 70449 has been marked as a duplicate of this issue. ***
Jesse could you review your patch for release50? :) http://apisupport.netbeans.org/servlets/ReadMsg?list=cvs&msgNo=2990
Forgot to CC reviewer...
Sure, consider it reviewed. :-) Do you know anything about the unit test status? I tried to look at what SFBITest was doing but I didn't see anything that would test this specifically.
You can simulate Windows behaviour on linux with s/File.separator/\\/ in your patch, I believe (OS negation). Also we tried it here with Tomas on Windows and those two test failed before the patch was applied. After fix it works. There should be somebody who runs tests on Windows from time to time I think ;) I'll commit into release50 tomorrow.
Backported into release50 (the same diff)
verified in 5.0 && trunk