Bug 173893 - AbstractResponse.openStreamWithCanel pegs CPU
Summary: AbstractResponse.openStreamWithCanel pegs CPU
Status: RESOLVED FIXED
Alias: None
Product: Platform
Classification: Eclipse Project
Component: Update (deprecated - use Eclipse>Equinox>p2) (show other bugs)
Version: 3.3   Edit
Hardware: PC Windows XP
: P3 normal (vote)
Target Milestone: 3.3 M6   Edit
Assignee: Branko Tripkovic CLA
QA Contact:
URL:
Whiteboard:
Keywords: contributed
Depends on:
Blocks:
 
Reported: 2007-02-12 12:45 EST by Chris Grindstaff CLA
Modified: 2007-04-10 17:40 EDT (History)
7 users (show)

See Also:

Attachments
Patch to fix problem (858 bytes, patch)
2007-02-12 12:47 EST, Chris Grindstaff CLA 		no flags Details | Diff
Fixes another problem (1.08 KB, patch)
2007-02-15 16:58 EST, Chris Grindstaff CLA 		no flags Details | Diff
Modified patch to fix additional problems (3.90 KB, patch)
2007-02-19 14:07 EST, James D. Miles CLA 		no flags Details | Diff
Show Obsolete (1) View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Chris Grindstaff CLA 2007-02-12 12:45:10 EST 
Build ID: I20070210-0800

Steps To Reproduce:
1. Add a new remote site
2. Add a breakpoint in ConnectionThreadManager line 96
3. Click finish


More information:
AbstractResponse.openStreamWithCancel will peg the CPU when the runnable getInputStream is slow to return.
Comment 1 Chris Grindstaff CLA 2007-02-12 12:47:45 EST 
Created attachment 58787 [details]
Patch to fix problem

The fix is to move the join inside the for(::) loop.
I've attached a patch.
Comment 2 Chris Grindstaff CLA 2007-02-15 16:58:22 EST 
Created attachment 59102 [details]
Fixes another problem

There are 2 problems
1) A background thread would busy spin while it waited for another thread to connect. If the connector thread takes a long time, the CPU is pegged.

2) The second problem is a race-condition associated with the 2 threads. If the connection is established (or fails) quickly the polling thread may loop forever, again pegging the CPU.

This is fixed now.
Comment 3 Chris Aniszczyk CLA 2007-02-16 10:48:40 EST 
Dejan, this one is biting us pretty bad, what are your thoughts?
Comment 4 Dejan Glozic CLA 2007-02-16 10:58:38 EST 
Branko, please review the patch and release into HEAD.
Comment 5 Chris Aniszczyk CLA 2007-02-16 11:03:40 EST 
Thanks for the quick attention Dejan. This will have to be added to the beer tab.
Comment 6 James D. Miles CLA 2007-02-19 14:07:51 EST 
Created attachment 59298 [details]
Modified patch to fix additional problems

These changes preserve exception messages.
They also fix disconnect behavior
Comment 7 Dejan Glozic CLA 2007-03-21 17:25:21 EDT 
This is a localized patch - safe to backport to the maintenance stream if you want.

I released it into M6 - keep an eye on connection issues before backporting.