0001710: Reproducible OpenAL Walkbacks

Viewing Issue Simple Details [ Jump to Notes ]

[ View Advanced ] [ Issue History ] [ Print ]

Category

Severity

Reproducibility

Date Submitted

Last Update

0001710

[Croquet] Jasmine

crash

always

08-18-05 18:59

08-18-05 21:15

Reporter

schwa

View Status

public

Assigned To

Priority

normal

Resolution

open

Status

new

Summary

0001710: Reproducible OpenAL Walkbacks

Description

The easiest way to reproduce this bug is to walk into an OpenAL space, and then collapse and expand the TeapotMorph. Another way is to save, quit, and restart an image where an OpenAL Teapot is open. Once these walkbacks occur, you can't even delete the teapot (just causes more walkbacks).

Ultimately, the problem is that an instance of the OpenAL class does not store a handle to an OpenAL context, so there is no way to tell whether OpenAL is in a valid state (contrast this with OpenGL>>destroy... it is clear that an OpenGL context is invalid when the 'handle' iVar is nil). The OpenAL code uses the deprecated function alutInit(), which does not do any error checking and does not return a handle to an OpenAL context. See the following links (search the latter for 'alut'):

http://opensource.creative.com/pipermail/openal/2005-February/002660.html [^]
http://www.openal.org/openal_webstf/specs/OpenAL1-1Spec.pdf [^]

Although alutInit() and alutExit() may be sufficient for a simple C program, they are not appropriate for a dynamic object environment that can be snapshotted and restarted. Instead, we should use functions like alcCreateContext() and alcDestroyContext() to explicitly manage the lifetime of OpenAL contexts (Jasmine's OpenGL support hides these in the plugin with #primDestroyRenderer: and #primCreateRenderer:x:y:w:h:).

Another example is the support that I have written for Nvidia's Cg. I haven't pushed this into Jasmine because DAS advised against major changes since Hedgehog is coming Real Soon Now, but I have added code to OpenGL to allow additional APIs to be supported without requiring further changes to OpenGL (only method additions, no new iVars necessary). If anyone wants to use this for OpenAL, too, let me know. Even though the lifetimes of a Cg and OpenGL are much more closely related than to that of an OpenAL context, I think that it would be quite reasonable to create/destroy/re-create OpenAL contexts at the same time as the other context types (eg: collapse/expand window, quit/restart image, etc.)

Additional Information

Attached Files

Relationships

Notes
(0002424 - 673 - 685 - 685 - 685 - 685 - 685) pmoore 08-18-05 21:15	alutInit() and alutExit() are never actaully called in the plugin. The plugin primitives primAlutInit and primAlutExit instead call alcCreateContext() and alcDestroyContext() directly. This hides the details of opening and closing devices and creating contexts much like the OpenGL methods you described above do. I agree that the names are confusing and poorly chosen. It should be fairly easy to have primAlutInit (or whatever the name should be) return the new context. I would be very interested in seeing what youve done for Cg. As far as what is already in Jasmine, where should I be looking? I'm guessing that you are refering to OpenGLName and OpenGLNameManager.

Issue History
Date Modified	Username	Field	Change
08-18-05 18:59	schwa	New Issue
08-18-05 21:15	pmoore	Note Added: 0002424
08-18-05 21:17	pmoore	Issue Monitored: pmoore

Mantis 1.0.8[^]

34 total queries executed.
27 unique queries executed.