Bug 1350

Summary:	CGSUpdateManager warning during createContext on OS X 10.11.6
Product:	[JogAmp] Jogl	Reporter:	Peter.C.Mehlitz
Component:	macosx	Assignee:	Sven Gothel <sgothel>
Status:	RESOLVED WORKSFORME
Severity:	normal	CC:	gouessej, jg
Priority:	P4
Version:	2.4.0
Hardware:	All
OS:	macosx
Type:	DEFECT	SCM Refs:
Workaround:	---
Attachments:	CGSUpdateManager warning backtrace

Description Peter.C.Mehlitz 2017-01-06 20:37:25 CET

Created attachment 796 [details]
CGSUpdateManager warning backtrace

I get reproducible CGSUpdateManager warnings on OS X 10.11.6 (MacPro 2009 with NVidia GeForce GT 120) which do not happen on other platforms (OS X 10.10.5, Ubuntu 16.04). This is with the current jogl 2.3.2 from maven central.

The OS X 10.11.6 machine also shows bug #1299 ("uncommitted CATransaction..") during shut down.

The backtrace is attached.

The program in question is from https://github.com/nasarace/race-client-example.git. It needs SBT (http://www.scala-sbt.org/) to build, which for instance can be installed via homebrew.

  brew install sbt

The steps to build and run are

  git clone https://github.com/NASARace/race-client-example.git
  cd race-client-example
  sbt stage
  ./race-client-example config/ISS.conf

This will bring up a NASA WorldWind window that shows the current ISS position. Dragging the globe to rotate it will produce the warning.

To shut down, enter "9" at the command prompt in the terminal from where the application was started. This should also report the "uncommitted CATransaction.." bug.

-- Peter

Comment 1 Julien Gouesse 2017-01-10 09:49:18 CET

Please provide a SSCCE without external dependencies.

Comment 2 Julien Gouesse 2017-01-17 11:06:48 CET

We're currently planning a maintenance release (2.3.3). Please consider providing a SSCCE:
http://sscce.org

We cannot include a test case with third party dependencies mainly because there is a risk that a part of the root cause is within them and not in JogAmp and because we cannot add such test into our own non regression unit tests. Moreover, writing a SSCCE helps to isolate the root cause of the bug by simplifying the source code involved to reproduce it. I admit that it's almost obvious that the problem lies within JOGL itself but a SSCCE would still be helpful.

Comment 3 Peter.C.Mehlitz 2017-01-18 20:22:17 CET

will try but given the size and complexity of the JOGL client (NASA WorldWind geospatial viewer) this will take a while.
Please note the problem does not occur on other platforms, including other OS X versions.

Comment 4 Julien Gouesse 2017-01-18 20:47:59 CET

Maybe you can run the tests provided with JOGL. If one of them allows you to reproduce this bug, you'll only have to mention it here instead of spending a lot of time in writing a SSCCE.

Comment 5 Peter.C.Mehlitz 2017-01-19 00:02:19 CET

the tests for the current jogl from https://github.com/sgothel/jogl.git all pass, but the reported effect only shows warnings on the screen, it doesn't crash and hence might go un-noticed. I didn't see respective warnings while running the tests though.

Is there any specific test you want me to try?

Comment 6 Julien Gouesse 2017-01-19 00:10:01 CET

I need to analyse a bit more the warnings. I have the feeling of shooting in the dark, I don't know yet how to reproduce this bug. I'll investigate tomorrow.

Comment 7 Jesse G 2018-02-27 21:07:33 CET

Hi Peter and Julien,

I see this same warning printout (the CGSUpdateManager::log() message, with same stack order as listed in your attached log) on OS X 10.11.6 and JOGL 2.3.2 just by running the JOGL demos.  The warning happens every time, and is always printed 10 seconds after launch.  The software seems to continue running just fine despite the warning.

e.g., I see it when executing: 
java demos.vertexBufferObject.VertexBufferObject 
java demos.gears.Gears 

I tried both Java 8 and 9 and got the same results.

Comment 8 Sven Gothel 2019-03-30 05:32:36 CET

All OSX issues will be validated for version 2.4.0

Comment 9 Sven Gothel 2019-04-10 07:39:45 CEST

As of today, our latest OSX tests were all successful w/o such warnings.

We only have the latest 10.14.* installed and cannot test on older machines.

I close this bug now, may somebody re-open it if recurring,
AFAIK Peter is also not having access to this machine anymore.