Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 326990

Summary:

Create markers takes 12 times more in some cases than in others

Product:

[Tools] AJDT

Reporter:

Marc Mazas <marc.mazas>

Component:

Core

Assignee:

Andrew Eisenberg <andrew.eisenberg>

Status:

RESOLVED WORKSFORME

QA Contact:

Severity:

normal

Priority:

Version:

2.1.0

Target Milestone:

2.1.1

Hardware:

OS:

Windows XP

Whiteboard:

Attachments:

Description	Flags
trace showing different execution times	none

Description Marc Mazas

2010-10-05 06:04:04 EDT

Build Identifier: Build id: 20090621-0832

The create markers phase takes much more time (12 times more) in some cases (may be the cache mechanism is bypassed in these cases)
AJDT 2.1.0.e35x-release-20100630-1500


Reproducible: Sometimes

Steps to Reproduce:
Will provide AJDT Event Trace

Comment 1 Marc Mazas

2010-10-05 06:15:04 EDT

Created attachment 180233 [details]
trace showing different execution times

File concatenating 2 event traces from 2 sessions
Comments have been added, marked by --->
Shows different execution times of delete and create markers (19 sec / 240 sec)
Using Maven
One project (Commons) hosts the aspects
Other projects reference the aspects library in Commons
One project (ProgrammesCiblesJava) is much bigger than the others : 324 files, each file declaring in average 20 inner classes (in 3 or more levels).

Comment 2 Andrew Eisenberg

2010-10-05 23:44:22 EDT

Thanks for the execution trace, but I'm having a little trouble decyphering it.  The overall times do not looks entirely bad to me.

I see a few lines where it seems to take many seconds to create a few markers for a project and other lines where it takes substantially less time.

It's hard to know if the root cause is AJDT, AspectJ, Eclipse, or some third party plugin.  In Eclipse, as I'm sure you know, there are many background tasks that take up memory and cpu.  It could be that a background task (such as updating maven indexes) is causing a wayward GC that blocks everything for a while.

Things that would help me address this:

1. can you reliably recreate this using a fixed set of steps?
2. can you send me a project that reliably creates this?

Comment 3 Marc Mazas

2010-11-04 10:49:00 EDT

(In reply to comment #2)
> Thanks for the execution trace, but I'm having a little trouble decyphering it.
>  The overall times do not looks entirely bad to me.
> 
> I see a few lines where it seems to take many seconds to create a few markers
> for a project and other lines where it takes substantially less time.
> 
> It's hard to know if the root cause is AJDT, AspectJ, Eclipse, or some third
> party plugin.  In Eclipse, as I'm sure you know, there are many background
> tasks that take up memory and cpu.  It could be that a background task (such as
> updating maven indexes) is causing a wayward GC that blocks everything for a
> while.
> 
> Things that would help me address this:
> 
> 1. can you reliably recreate this using a fixed set of steps?
> 2. can you send me a project that reliably creates this?

1. I was / am able to recreate this (happens on projet / IDE startup) 
2. I'm not able to send you a project for this, my environment is currently too big for this and it would take a few days to strip it down

You can close this bug report

Comment 4 Andrew Eisenberg

2010-11-04 11:35:48 EDT

Thanks for getting back to me.

There are some things that you can do to speed up compilation:

1. disable weaveinfo messages
2. shut down the ajdt event trace view

Also, if there are particular actions that seem to trigger full builds, when you don't think they should, please raise a bug for that and we will explore it.