Created attachment 61474 [details]
thread dump

Hi, Jungi.  This looks like a known bug, which I thought I had fixed on May 13, see issue 134577.  I can't tell from
your report the date of the build, so I don't know if this is this means you need a new build or the bug was not
correctly fixed.

Hi David, this build is from todays morning CET. Hudson says: Build #1944 (16.5.2008 7:33:47)[1]


[1]: http://deadlock.netbeans.org/hudson/job/trunk/1944/

same is happening for Jarda who uses its own build from sources which were fetched 3-4hours ago...

I could reproduce using today's installer build using JDK 1.6.0_02
The "Looking for a local MySQL database server" task was blocked - progress bar not moving.

I could not reproduce using JDK 1.5

Created attachment 61506 [details]
thread dump 200805150003 installer build , jbaker

this seems to be different issue than 134577

Can lookupDatabaseServer be executed in another thread instead of blocking with synchronized?

pls disregard my previous comment.  My workspace was not up-to-date.

But I think the problem is in getDatabaseServer() where there are multiple clients trying to retrieve
a DatabaseServer.  So, as the comment states in getDatabaseServer, this needs to support concurrent access

Thanks for looking at this John.  I'll be offline (in airplane), but I'll be working on this.

Here's what I think I'm seeing:

The folder recognizer is doing a lookup which causes the DB root node to initialize itself.

But as part of DB root node initialization, it ultimately calls ServerNodeProvider.checkNodeArray(), which itself does a
lookup of the DatabaseServer.

I had tried to eliminate this kind of double-lookup by moving the lookup of DatabaseServer to
ServerNodeProvider.getNodes(), but obviously to no avail.

The longer-term solution is to fix the RootNode initialization so it doesn't do lookups on the same thread in which it
is initialized, as this happens as part of a lookup. The child initialization should happen asynchronously on another
thread and notify the node tree when it is complete.

In the meantime I may eliminate doing lookups inside db.mysql to avoid circular lookups, and instead hardcode the
DatabaseServer instance to be a MySQLDatabaseServer implementation.

Hi David, did you change anything in the DB code?

I'm asking because in javaee build #1964 (18.5.2008 13:12:08 CET) I cannot reproduce UI freeze but I can't see _any_
nodes in services tab - there's no servers, dbs, webservices etc. node there (that would be just another P1...)

Also when I try to close the IDE I get an info about still running Registering local JavaDB process and when I dismiss
the dialog with this info JVM is still running and I have to KILL it...

Created attachment 61534 [details]
thread dump from "closed" IDE

Yes, I changed something, and obviously still not working.   Working on a fix.

http://hg.netbeans.org/main/rev/0a9b2449aa8d

applied David's suggestion

Integrated into 'main-golden', available in NB_Trunk_Production #206 build
Changeset: http://hg.netbeans.org/main/rev/0a9b2449aa8d
User: John Baker <jbaker@netbeans.org>
Log: #135206 IDE freeze during the first startup; temporary fix

*** Issue 135425 has been marked as a duplicate of this issue. ***

I just updated sources and still can reproduce the bug.

Created attachment 61640 [details]
thread dump

Could you try the latest nightly trunk build also.  It's possible your sources are not up-to-date

If you look at the change set John checked in, we no longer call lookupDatabaseServer

@ -67,7 +68,7 @@ public class DatabaseServerManager {
//
// TODO - fix the DB Explorer so it refreshes its children
// on a separate thread, so we don't get into this situation.
- DatabaseServer server = lookupDatabaseServer();
+ DatabaseServer server = MySQLDatabaseServer.getDefault();

but in the stack trace you posted, the call to lookupDatabaseServer is still there.

[ideTestRunner]         at
org.netbeans.modules.db.mysql.DatabaseServerManager.lookupDatabaseServer(DatabaseServerManager.java:58)
[ideTestRunner]         at org.netbeans.modules.db.mysql.DatabaseServerManager.<clinit>(DatabaseServerManager.java:51)


So you must somehow be running off of old sources...  Can you please try again?

Thanks!

David

Hi, this happens on Build 20080520224434 from hudson #1993(22:44) as well

Hi, Janie.  Can you please post a thread dump when it hangs?   Thanks.

*** Issue 135560 has been marked as a duplicate of this issue. ***

dkolar: can you reproduce from the latest build?
janie: can you please give me a thread dump? 

Marking incomplete because I can't reproduce and the stack traces so far represent the code tree before the fix.

After the fix it doesn't happen 90% of the time, but only about 25%. I'm attaching the dump, but even I've my sources up
to date, the dump is pretty similar to previous one.

Created attachment 61759 [details]
thread dump

Please list the new steps to reproduce the 25% of the time freeze

OK, I found the problem, the DEFAULT member variable is being initialized with lookupDatabaseServer().

I'll check in a fix and this should finally be resolved.

See if this fixes the freeze - from David's suggestion
I could not reproduce the freeze while testing this fix.

http://hg.netbeans.org/main/rev/c3b9a4df4b69

Well, now it seems to be ok finally.

Integrated into 'main-golden', available in NB_Trunk_Production #221 build
Changeset: http://hg.netbeans.org/main/rev/c3b9a4df4b69
User: John Baker <jbaker@netbeans.org>
Log: #135206  IDE freeze during the first startup

*** Issue 135908 has been marked as a duplicate of this issue. ***

Verified.