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

Bug 478151

Summary: Critical bug in Buildship and SimRel
Product: Community Reporter: Mickael Istria <mistria>
Component: Cross-ProjectAssignee: David Williams <david_williams>
Status: RESOLVED FIXED QA Contact:
Severity: critical    
Priority: P3 CC: cdtdoug, david_williams, dennis.huebner, etienne, fbricon, konstantin, manderse, marcel.bruch, mknauer, mlippert, sebastian.zarnekow, stepper, webmaster
Version: unspecified   
Target Milestone: ---   
Hardware: All   
OS: All   
Whiteboard:
Bug Depends on: 478054    
Bug Blocks:    

Description Mickael Istria CLA 2015-09-23 06:05:53 EDT
BuildShip version included in SimRel (Mars.1 RC4) apparently has a severe bug that affects the whole IDE usage: bug 478054
Depending on how likely users are going to face this bug, it may make sense to respin Mars.1 to include the fixed version of BuildShip.

As BuildShip is the very new thing of this released, we can expect that many users will give it a try and may be highly disappointed if they are affected by the bug.

Denis reported that simply installing BuildShip against DSL package "activates" the bug and prevents from selecting a workspace.
Comment 1 Martin Lippert CLA 2015-09-23 06:10:03 EDT
strong +1 for a re-spin (from my end)
Comment 2 Etienne Studer CLA 2015-09-23 06:14:37 EDT
> Denis reported that simply installing BuildShip against DSL package
> "activates" the bug and prevents from selecting a workspace.

Just to clarify, the bug only reveals itself _after_ running at least one Gradle task in the TaskView since only then is the bundle.start() method called.
Comment 3 Etienne Studer CLA 2015-09-23 06:15:09 EDT
+1 for a re-spin from my side, too.
Comment 4 Mickael Istria CLA 2015-09-23 06:22:11 EDT
+1 for me too (individual opinion, not an "official" one from the whole Red Hat crew)
Comment 5 Marcel Bruch CLA 2015-09-23 06:31:05 EDT
+1
Comment 6 Dennis Huebner CLA 2015-09-23 06:53:20 EDT
+1
Comment 7 Sebastian Zarnekow CLA 2015-09-23 07:19:12 EDT
+1 for a respin
Comment 8 Max Rydahl Andersen CLA 2015-09-23 07:30:10 EDT
+1 if at all possible.

If the issue was *just* affecting buildship I would be -1, but here we are talking about a bug that once someone tries out one of the biggest new features in Eclipse release train in the last 2-3 years he will render his whole IDE damaged.
Comment 9 Fred Bricon CLA 2015-09-23 08:00:42 EDT
+1 for a respin
Comment 10 David Williams CLA 2015-09-23 09:10:56 EDT
I've asked some members of Planning Council to take a closer look at this (such as if using -data <workspace> works around the problem, once it occurs) and if they agree its respin worthy to take it to the Planning Council's Exception Process. That is a pretty low bar, so I'd expect a decision late today. 

CC'ing webmaster since they do some things so the builds are available early, on Thursday, to "Friends" (I think) so this is a "head up" that release date *may* be changing. 

[If not obvious, if we do a respin, I'd recommend a one week slip in formal release date.]
Comment 11 David Williams CLA 2015-09-23 18:06:42 EDT
We will delay the release by one week, to do a respin, and reset the "quiet clock". 

I will close this bug by pasting what I wrote to cross-list: 

https://dev.eclipse.org/mhonarc/lists/cross-project-issues-dev/msg12402.html

= = = = = = = = 

The Mars.1 release is now officially delayed until October 2, 10 AM. 

If not obvious, this means all participants in "coordinated release train" should not make your releases visible on 9/25, but wait until 10/2 10 AM to make them visible, and announce your official releases. 

As readers of this list are aware, a bug was found, that many thought deserved to delay the release, and re-spin the repository, and rebuild the EPP packages. 

Thanks to all who made constructive comments. And, to those finding the bug(s) and especially to those providing the correct fix! 
I've also asked for a less discussed bug to be fixed, which is "simply" to remove a duplicate bundle (but, it is a low level one, and can cause problems with "invalid loader constraints at run time). The two bugs are discussed and documented in 

Bug 478054- Eclipse does not prompt for workspace any more after installing Buildship, and 
Bug 477148- Contribute JSch 0.1.53 rather than 0.1.51 to Mars.1 

I've created a branch, named Mars_maintenance.1, pointed all contributions in that branch to pull from our "final build candidate", and am doing a test build now of that "identical" build. Just to confirm the process. I've asked the projects with fixes  to submit their updates to that branch, and will build that repo later tonight (assuming everyone is in  a nearby timezone). 

I'd estimate the final repo (with changes) will be available late tonight or early tomorrow, the EPP packages shortly after, and then we'll have another "quiet week" to confirm all went well. 


Thanks again,
Comment 12 David Williams CLA 2015-09-24 09:39:27 EDT
Just to document it here, the "new repo" is complete. 

And, there are 5 fewer bundles in the new repo. The expected 

com.jcraft.jsch [0.1.51.v201410302000]

is gone. Leaving only the 0.1.53 version. 

but also 4 others: 

org.eclipse.rap.jface [3.0.0.20150520-1729]
org.eclipse.rap.jface.databinding [3.0.0.20150520-1729]
org.eclipse.rap.rwt [3.0.0.20150601-1556]
org.apache.commons.fileupload [1.2.2.v20111214-1400]

I checked and in the original repo and confirmed there were multiple versions of these bundles available, so now just one, the more recent version. 

org.apache.commons.fileupload_1.3.1.201405120000.jar
org.eclipse.rap.jface_3.0.1.20150709-1405.jar
org.eclipse.rap.jface.databinding_3.0.1.20150709-1405.jar
org.eclipse.rap.rwt_3.0.1.20150908-1130.jar

Overall, I'm assuming this is "goodness" but, if anyone has a "product" that was pulling in any of the old ones, that should be re-worked, since, no longer part of the release.