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

Bug 350889

Summary:	Null pointer exception caused by solver attribute initialized not correctly modeled as transient
Product:	[Technology] STEM	Reporter:	James Kaufman <jhkauf>
Component:	Core	Assignee:	Stefan Edlund <sedlund>
Status:	CLOSED FIXED	QA Contact:
Severity:	critical
Priority:	P3
Version:	unspecified
Target Milestone:	1.1.2
Hardware:	PC
OS:	Mac OS X - Carbon (unsup.)
Whiteboard:

Description James Kaufman

2011-06-30 18:36:35 EDT

The ecore model for a solve has an attribute "initialized" that is not set to Transient. This causes a null pointer exception for newly created scenarios.
To reproduce, create a new scenario using the finite difference solver and try to run it. 
The first attempt to run it will cause a null pointer exeption in the FiniteDifferenceImpl class line 74
This happens because the partitioner is null
the Partitioner is null because the newly created scenario (see the xml code) has "initialized=true"

I think the fix is to change the emf model so that initialized is transient. It already has an EDefault of false but this is not enough because the creation process initializes it.
Question for Stefan: Should not ALL solver components be transient by default? Even decorators etc.... ?

After the first attempt fails to run with a null pointer, subsequent attempts will run ok.

Comment 1 James Kaufman

2011-06-30 18:37:26 EDT

Assigning to Stefan

Comment 2 Stefan Edlund

2011-07-06 19:46:56 EDT

Thanks for reminding me about this EMF bug, I remember seeing it some time ago but looks like it was never fixed. You're right, all components should be transient in the solvers.

Comment 3 James Kaufman

2011-07-18 21:47:33 EDT

Complete