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

Bug 336815

Summary:	Need to re-tag some projects to fix bad byte codes
Product:	[Eclipse Project] Platform	Reporter:	Dani Megert <daniel_megert>
Component:	Releng	Assignee:	Platform-Releng-Inbox <platform-releng-inbox>
Status:	RESOLVED FIXED	QA Contact:
Severity:	major
Priority:	P3	CC:	dj.houghton, eclipse.felipe, grant_gayed, john.arthorne, kim.moir, mober.at+eclipse, Olivier_Thomann, pwebster, Silenio_Quarti, tjwatson
Version:	3.6.2
Target Milestone:	3.6.2
Hardware:	All
OS:	All
Whiteboard:

Description Dani Megert

2011-02-10 08:19:35 EST

M20110209-1607.

The following projects need to be re-tagged to fix wrong byte codes:
- org.eclipse.osgi
- org.eclipse.swt.win32.win32.x86_64
- org.eclipse.swt.win32.win32.x86
- org.eclipse.swt.win32.wce_ppc.arm
- org.eclipse.jface
- org.eclipse.ui.workbench
- org.eclipse.ui.views.log
- org.eclipse.pde.core
- org.eclipse.pde.ui
- org.eclipse.equinox.core.sdk (the feature)
- org.eclipse.equinox.core.sdk (the feature)

Comment 1 Dani Megert

2011-02-10 08:20:35 EST

I'll take care of:
- org.eclipse.jface
- org.eclipse.ui.workbench
- org.eclipse.ui.views.log
- org.eclipse.pde.core
- org.eclipse.pde.ui

Comment 2 Dani Megert

2011-02-10 08:32:52 EST

Tom or DJ, can you take care of
- org.eclipse.osgi
- org.eclipse.equinox.core.sdk (the feature)
- org.eclipse.equinox.core.sdk (the feature)

Comment 3 Dani Megert

2011-02-10 08:46:46 EST

> - org.eclipse.equinox.core.sdk (the feature)
> - org.eclipse.equinox.core.sdk (the feature)

should be:
- org.eclipse.equinox.core.sdk (the feature)
- org.eclipse.equinox.sdk (the feature)

Comment 4 Dani Megert

2011-02-10 08:50:10 EST

(In reply to comment #1)
> I'll take care of:
> - org.eclipse.jface
> - org.eclipse.ui.workbench
> - org.eclipse.ui.views.log
> - org.eclipse.pde.core
> - org.eclipse.pde.ui
Done.

Comment 5 Thomas Watson

2011-02-10 08:58:07 EST

(In reply to comment #2)
> Tom or DJ, can you take care of
> - org.eclipse.osgi
> - org.eclipse.equinox.core.sdk (the feature)

I took care of these.(In reply to comment #3)
> > - org.eclipse.equinox.core.sdk (the feature)
> > - org.eclipse.equinox.core.sdk (the feature)
> 
> should be:
> - org.eclipse.equinox.core.sdk (the feature)
> - org.eclipse.equinox.sdk (the feature)

I don't have rights to org.eclipse.equinox.sdk

Comment 6 Thomas Watson

2011-02-10 09:02:15 EST

(In reply to comment #5)
> > - org.eclipse.equinox.sdk (the feature)
> 
> I don't have rights to org.eclipse.equinox.sdk

John do you have rights to this project.  In Kim's absence we are looking for someone that can tag it.  Dani and I do not have rights.

Comment 7 John Arthorne

2011-02-10 09:25:30 EST

(In reply to comment #6)
> John do you have rights to this project.  In Kim's absence we are looking for
> someone that can tag it.  Dani and I do not have rights.

Done.

Comment 8 Grant Gayed

2011-02-10 10:34:16 EST

I've re-tagged the required projects for swt.

Comment 9 Dani Megert

2011-02-10 10:38:41 EST

Comment 10 Martin Oberhuber

2011-02-12 09:14:25 EST

I'm wondering to what extent these comparator diffs tell us that the byte code for those bundles had in fact been invalid before changing the compiler?

I'm especially surprised that there's a diff in the features since these shouldn't have been affected by the compiler change?

Do the comparator diffs tell us anything about the severity of the compiler bug that was fixed, or the correctness of the fix?

Comment 11 John Arthorne

2011-02-23 14:00:36 EST

(In reply to comment #10)
> I'm wondering to what extent these comparator diffs tell us that the byte code
> for those bundles had in fact been invalid before changing the compiler?

The most likely cause is that unrelated JDT compiler settings were introduced or changed in the newer version of JDT core. I don't think we have proof that any of these bundles that we retagged were actually affected by the invalid byte code problem.

> I'm especially surprised that there's a diff in the features since these
> shouldn't have been affected by the compiler change?

The features needed to be retagged because of a last minute p2 fix, and some p2-related feature versions had not be incremented in 3.6.2.

> Do the comparator diffs tell us anything about the severity of the compiler bug
> that was fixed, or the correctness of the fix?

No. The byte codes would need to be examined by hand to do that.

Comment 12 Olivier Thomann

2011-02-23 18:32:30 EST

(In reply to comment #11)
> (In reply to comment #10)
> > I'm wondering to what extent these comparator diffs tell us that the byte code
> > for those bundles had in fact been invalid before changing the compiler?
> 
> The most likely cause is that unrelated JDT compiler settings were introduced
> or changed in the newer version of JDT core. I don't think we have proof that
> any of these bundles that we retagged were actually affected by the invalid
> byte code problem.
The problem in this case is that the compiler used before it was changed to fix the invalid bytecode is not the compiler without just that fix. The compiler contains other changes that could also affect the generated bytecodes like some specific cases in swich code generation.

> No. The byte codes would need to be examined by hand to do that.
I did a manual examination of all .class files that were reported as changed on demand. All of them ended up being valid changes.