Bugzilla – Full Text Bug Listing
| Summary: | [rfc 160] Implement the new coodinator service | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | [Eclipse Project] Equinox | Reporter: | Thomas Watson <tjwatson> | ||||||||||
| Component: | Compendium | Assignee: | John Ross <jwross> | ||||||||||
| Status: | RESOLVED FIXED | QA Contact: | |||||||||||
| Severity: | normal | ||||||||||||
| Priority: | P3 | CC: | hargrave | ||||||||||
| Version: | 3.7 | ||||||||||||
| Target Milestone: | 3.7 M5 | ||||||||||||
| Hardware: | PC | ||||||||||||
| OS: | Mac OS X - Carbon (unsup.) | ||||||||||||
| Whiteboard: | |||||||||||||
| Bug Depends on: | |||||||||||||
| Bug Blocks: | 289124, 329688 | ||||||||||||
| Attachments: |
|
||||||||||||
|
Description
Thomas Watson
John is working on an implementation of this. I would like to get it released for M4. Created attachment 182672 [details] Proposed implementation as project ZIP archive Attached is iteration 1 of the proposed implementation for review. I think it's in pretty good shape. It's passing the CT which, believe it or not, seems to actually be quite thorough, although I'm sure we will come across some things to add as progress is made. For example, security permission tests still need to be added. There will be ongoing work such as continuing to add to the JUnit tests (Equinox as well as possibly the CT), clearing up https://www.osgi.org/members/bugzilla/show_bug.cgi?id=1783, optional JMX support, optional TSH support, as well as from any input received here. RFC 165 should be a big help in driving further testing and hardening the implementation since that work requires a Coordinator. See Bug 329688. Created attachment 182678 [details]
Proposed implementation 2 as project ZIP archive
Fixes the following issues.
Activator.stop() - added null checks in case of ds
Coordinator.getNextId() - fix so that Long.MAX_VALUE is used as an ID
Coordination.join() - fix so that start time is taken into account if wait() must be called more than once
CoordinationTimerTask.run() - fix to prevent timer thread from stopping
Coordination.addParticipant() - holding lock while calling Coordination.join() can cause deadlock; refactored method
There's still a safety issue with Coordination.addParticipant(). The coordination could terminate before the participant is added resulting in the participant never receiving a callback. Created attachment 182742 [details]
Proposed implementation 3 as project ZIP archive
Fixed safety issue in Coordination.addParticipant().
Fixed safety issue in Coordination.getParticipants().
We may release the code to CVS before M4, but I am moving this bug to M5 for integrating the code into the build. Created attachment 185229 [details]
Proposed implementation 4 as project ZIP archive
Contains spec updates from Walldorf F2F as well as a few that have not been finalized but appear to be heading in that direction. Among the latter would be NPE when the reason for the failure is null and IAE when a timeout related value is negative.
Also incorporates changes from Tom's review.
This will be the last submission before release barring any further feedback.
I released the latest project provided by John. Thanks for the contribution John! I opened bug 332693 to get the new coordinator bundle integrated into the equinox build. |