Notes |
|
(0030332)
|
David Gobbi
|
2013-02-12 12:44
|
|
But there should not be any modules that depend on vtkWrappingPython.
The only one that does is Rendering/Matplotlib, but it only needs the
system python libraries, it doesn't actually need wrapping at all. It can
call find_package(PythonLibs) directly. |
|
|
(0030333)
|
Robert Maynard
|
2013-02-12 12:58
|
|
Yes the bug was found by the fact that matplotlib depends on vtkWrapPython. My concern is that we don't have an explicit checks to make sure that going forward no module depends on vtkWrapPython causing a failure in creating the __init__.py file. |
|
|
(0030364)
|
David Gobbi
|
2013-02-21 11:29
|
|
Commit 146e2f78f (on Feb 20, 2013) may have made this bug obsolete.
It eliminates the need for other modules to depend on vtkWrappingPython,
by adding a new module that provides just the python headers and libraries.
The Matplotlib module depends on this new module now. |
|
|
(0030365)
|
Robert Maynard
|
2013-02-21 11:32
|
|
In theory if we add a new vtk Module whose name starts with X or Y or Z and has nothing depend on it, wouldn't the python __init__ be missing since they would be configured after WrappingPython? |
|
|
(0030366)
|
Marcus D. Hanwell
|
2013-02-21 11:39
|
|
There are still some changes Utkarsh and I were working on to support building/wrapping from outside of VTK that should address this by performing wrapping in a wrap step at the end using the metadata we already generate in the cmake files for each module. I am tied up with other things right now, but I will try and get something pushed to Gerrit. It isn't a direct response to this bug, but should address any remaining issues with __init__.py |
|
|
(0031238)
|
Dave DeMarle
|
2013-07-22 20:02
|
|
moving all "tabled" bugs into "backlog" category since "tabled" is no longer used. |
|