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.

Bug 40420

Summary:	Startup time regression caused by 500x slower o.o.util.actions.CallbackSystemAction.findGlobalContextAction()
Product:	platform	Reporter:	Antonin Nebuzelsky <anebuzelsky>
Component:	Actions	Assignee:	Jaroslav Tulach <jtulach>
Status:	VERIFIED INVALID
Severity:	blocker	CC:	issues
Priority:	P2	Keywords:	PERFORMANCE, REGRESSION
Version:	3.x
Hardware:	PC
OS:	Windows ME/2000
Issue Type:	DEFECT	Exception Reporter:
Bug Depends on:
Bug Blocks:	36556
Attachments:	Snapshot of jprofiler window with the method in NB351 Snapshot of jprofiler window with the method in trunk

Description Antonin Nebuzelsky 2004-02-24 14:35:37 UTC

There is a big execution time regression of
findGlobalContextAction() method. Profiler shows
me 500x worse times for the current implementation
in trunk in comparison to NB351. This participates
on the startup time regression we have.

The method was slowed down perhaps with the
version 1.30.

Comment 1 Antonin Nebuzelsky 2004-02-24 14:41:02 UTC

Created attachment 13605 [details]
Snapshot of jprofiler window with the method in NB351

Comment 2 Antonin Nebuzelsky 2004-02-24 14:41:51 UTC

Created attachment 13606 [details]
Snapshot of jprofiler window with the method in trunk

Comment 3 Antonin Nebuzelsky 2004-02-24 14:49:35 UTC

The cause is in the initialization of
CallbackSystemAction$GlobalManager and down to
o.o.windows.TopComponent.getRegistry... and down to WindowManager
initialization... :-)

Comment 4 Jaroslav Tulach 2004-02-24 15:05:30 UTC

There is a change, that is true, but the TopComponent registry would
have to be initialized anyway, even in 3.5. The time just moved from
one place to another, imho. I cannot verify this, as the screenshot
does not show what actually is happening under
GlobalManager.getDefault(), but I belive it is not more then what
would happen in another place in 35 anyway.

Comment 5 Antonin Nebuzelsky 2004-02-24 15:32:51 UTC

Correct. Anyway tracking it down I realized that WindowManager
initialization now takes four times longer than in NB351. But it's a
completely different story... ;-)