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

Bug 276457

Summary: Install fails without a trace using I20090514-2000.
Product: [Eclipse Project] Equinox Reporter: Thomas Hallgren <thomas>
Component: p2Assignee: P2 Inbox <equinox.p2-inbox>
Status: VERIFIED FIXED QA Contact:
Severity: critical    
Priority: P3 CC: dj.houghton, henrik.lindberg, Kevin_McGuire, thomas
Version: 3.5Flags: dj.houghton: review+
Target Milestone: 3.5 RC1   
Hardware: PC   
OS: Linux   
Whiteboard:
Attachments:
Description Flags
Fix v01
none
Regression test none

Description Thomas Hallgren CLA 2009-05-15 05:16:07 EDT 
Steps to reproduce:

1. Install a fresh Eclipse IDE of version I20090514-2000.
2. Start it on a fresh workspace.
3. Open the Install wizard and select the Galileo repository
4. Install something (I installed Buckminster, but it doesn't seem to matter)
5. Click next and the finish.

A progress flickers and then nothing happens.

6. Restart the IDE
7. Open the Install wizard again and click on "already installed"

Nothing has been installed and I don't find any errors. No .metadata/.log and no log in <eclipse>/configuration. No entry in Installation History.
Comment 1 John Arthorne CLA 2009-05-15 10:36:20 EDT 
Created attachment 136000 [details]
Fix v01
Comment 2 John Arthorne CLA 2009-05-15 10:37:23 EDT 
Henrik, I think this is the fix, but I don't understand why it was ever canceling here. Note it also seems timing related. Sometimes install will succeed and sometimes it will fail. Stepping through in the debugger it always seems to succeed.
Comment 3 John Arthorne CLA 2009-05-15 11:02:34 EDT 
Created attachment 136008 [details]
Regression test
Comment 4 John Arthorne CLA 2009-05-15 11:10:27 EDT 
Note this regression was introduced by the fix for bug 275975. The first patch in that bug changes the monitor from a NullProgressMonitor to the real progress monitor. Previously, canceling the monitor had no effect because it was always a null monitor that was never passed to anyone else (which is why I can't see why it was there in the first place). This was changed to pass in the real progress monitor so cancelation requests would be propagated correctly.

I'm asking DJ for review because we need to get the fix in ASAP to get an RC1 rebuild going. Henrik, I would still appreciate a review when you get a chance.
Comment 5 John Arthorne CLA 2009-05-15 11:16:06 EDT 
Fix released to HEAD.
Comment 6 Henrik Lindberg CLA 2009-05-15 14:57:30 EDT 
(In reply to comment #4)
> I'm asking DJ for review because we need to get the fix in ASAP to get an RC1
> rebuild going. Henrik, I would still appreciate a review when you get a chance.
> 
Thanks for catching this - that "cancel" did not make sense at all when the monitor is passed in. I totally missed that the streamed read did that on close - it made sense earlier.
Comment 7 John Arthorne CLA 2009-05-15 15:52:57 EDT 
Fix verified in I20090515-1143.
Comment 8 John Arthorne CLA 2009-05-19 12:05:26 EDT 
*** Bug 276932 has been marked as a duplicate of this bug. ***