Difference between revisions of "Developer Meetings/20121009"
From Slicer Wiki
m (→Conclusion) |
m (→Conclusion) |
||
Line 33: | Line 33: | ||
* Commit message | * Commit message | ||
** Should we enforced a style? Having git hook ? May be we should create a 'setupfordevelopment' script like it exists for VTK | ** Should we enforced a style? Having git hook ? May be we should create a 'setupfordevelopment' script like it exists for VTK | ||
− | |||
* Pure Git setup: This is something on the radar and planned for 4.3. | * Pure Git setup: This is something on the radar and planned for 4.3. | ||
* Gerrit: Will take more time. Need to talk more about it. | * Gerrit: Will take more time. Need to talk more about it. | ||
* And discussed assert | * And discussed assert |
Revision as of 21:13, 9 October 2012
Home < Developer Meetings < 20121009Attendees:
To discuss
- Tutorial testing status
- ChangeTracker standing issues:
- Crash if close slicer w/o closing scene: http://www.na-mic.org/Bug/view.php?id=2590
- 1 matrix and 12 vtkCommands are leaking: http://www.na-mic.org/Bug/view.php?id=2588
- ChangeTracker standing issues:
- Discuss commit r21147
- related issue 2534
- Style and Coding Conventions discussion topics
- Commit message style. See here
- Backwards compatibility and deprecation
- Q_ASSERT vs qDebug, and vtkErrorMacro
- Memory management from 'Create' methods [1]
Conclusion
- Talked about the factory and Linux VM before every body joined
- Steve looked at output of glxinfo on the linux VM running on virtual box. It seems to be executed using mesa. Jc will investigate how to use mesa also on the factory linux VM.
- Initiated discussion about Tutorial testing
- Nicole mentioned that the Brain Atlas dataset taking more than one hour to load on a linux debug build (SPL workstation)
- Andriy discussed the crash happening when closing slicer without "closing the scene" first after running ChangeTracker. See issue 2590
- Andriy also mentioned the work done related the memory leaks associated with ChangeTracker tutorial. See 2588. Julien pointed the Slicer wiki page listing debugging approach that could be used to "Track a crash while accessing already deleted object pointer". See http://www.slicer.org/slicerWiki/index.php/Documentation/Nightly/Developers/Tutorials/Troubleshooting
- Discussed warnings occurring when loading Brain atlas scene (nac-brain-atlas-1.0/brain-atlas-hierarchy+snapshots.mrml):
vtkMRMLParser (0x6b3a990): Failed to CreateNodeByClass: vtkMRMLTGParametersNode
.Command line module Model Maker is version "4.1" but parameter set from MRML file is version "3.2". Parameter set may not load properly
There is a danger here. What if AddNode or RemoveNode were called prior, the NodeIDsMTime would be in sync without having the map in sync.
- On windows, on a debug build, visual studio spends a lot of time in STL function. Julien mentioned there is a way to tweak it.
- Talked about the lazy update
- Julien:
- We should probably think twice before pushing for lazy update
- There are not much node and model (500 * 200)
- What about sharing models between views ?
- Julien:
- Commit message
- Should we enforced a style? Having git hook ? May be we should create a 'setupfordevelopment' script like it exists for VTK
- Pure Git setup: This is something on the radar and planned for 4.3.
- Gerrit: Will take more time. Need to talk more about it.
- And discussed assert