|
|
I found that if I added this line to the add pyton test line in the Testing directory CMakeLists.txt file:
SLICER_ARGS --additional-module-paths ${CMAKE_BINARY_DIR}/lib/Slicer-4.1/qt-loadable-module
the reporting module would be found and loaded and the logic found for my test. It looks like because the testing automatically ignores the slicerrc file, the extra module path that pointed to the reporting module wasn't being set (I didn't install Reporting as an extension, just built it outside of Slicer and added the additional module path). |
|
nicole
2012-07-12 09:39
administrator
~0005127
Last edited: 2012-07-12 09:53
|
For the generic tests, the Reporting-build/AdditionalLauncherSettings.ini file does have additional paths in it, but the first ones in the library paths reference the cli specific paths that the loadable Reporting module doesn't have, and the path setting doens't include the loadable path at all:
[LibraryPaths]
1\path=/projects/birn/nicole/Slicer4/Reporting-build/lib/Slicer-4.1/cli-modules/.
2\path=/projects/birn/nicole/Slicer4/Reporting-build/lib/Slicer-4.1/qt-loadable-modules/.
size=2
[Paths]
1\path=/projects/birn/nicole/Slicer4/Reporting-build/lib/Slicer-4.1/cli-modules/.
size=1
I tried adding the loadable modules bin dir to SlicerBlockAdditionalLauncherSettings.cmake
but it didn't fix the problem with the generic module tests for the reporting module:
[LibraryPaths]
1\path=/projects/birn/nicole/Slicer4/Reporting-build/lib/Slicer-4.1/cli-modules/.
2\path=/projects/birn/nicole/Slicer4/Reporting-build/lib/Slicer-4.1/qt-loadable-modules/.
size=2
[Paths]
1\path=/projects/birn/nicole/Slicer4/Reporting-build/lib/Slicer-4.1/cli-modules/.
2\path=/projects/birn/nicole/Slicer4/Reporting-build/lib/Slicer-4.1/qt-loadable-modules/.
size=2 |
|
|
|
Should the cmake macro be modified to always include the paths that can be populated by a module?
In Slicer-build/lib/Slicer-4.1 these are
ITKFactories cli-modules qt-loadable-modules qt-scripted-modules
Should they all be added for an extension in <my-module-build>/lib/Slicer-4.1 ? |
|
|
|
Fixed in r20697
See the following commits:
http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&revision=20697
http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&revision=20696
http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&revision=20695 |
|
nicole
2012-07-26 09:43
administrator
~0005259
Last edited: 2012-07-26 11:10
|
The commits haven't fixed the Reporting module py_ReportingRoundTripTest so that it actually runs for me.
Slicer starts up and instantiates the Reporting module, but the test print outs don't show up on the console, nor does the test CTK DICOM database get created. That last is assumed by the fact that, at exit, there's only a message about the Reporting database being still active, nothing about the ReportingTesting database that the test is supposed to create/open.
I currently use a default location of /projects/igtdev/nicole/LocalDCMDB for the module database during regular run (since the slicerrc is ignored and my setting isn't loaded this default value gets used in the test). The test is supposed to create the db in
dbpath = slicer.app.slicerHome + '/Testing/Temporary/TestingDCMDB/ctkDICOM.sql'
Selected output from running the test after updating to 20700:
5: Loading module "Reporting"
5: Warning: In /projects/birn/nicole/Slicer4/Reporting/Logic/vtkSlicerReportingModuleLogic.cxx, line 138
5: vtkSlicerReportingModuleLogic (0x4dec790): InitializeDICOMDatabase: no DatabaseDirectory path found, please update the settings.
5: Using /projects/igtdev/nicole/LocalDCMDB
5:
[...]
5: Loading module "WebGLExport"
5: Loading module "Welcome"
5: Number of loaded modules: 34
5: Uninstantiating: "Welcome"
5: Uninstantiating: "WebGLExport"
[...]
5: Uninstantiating: "Reporting"
[...]
5: Uninstantiating: "CropVolume"
5: QSqlDatabasePrivate::removeDatabase: connection 'Reporting' is still in use, all queries will cease to work.
1/1 Test 0000005: py_ReportingRoundTripTest ........ Passed 16.38 sec
If I take out the passing of additional module paths to the python test (changing the CMakeLists.txt file slicer_add_python_unittest line), it fails to find the logic for the Reporting module:
5: ****
5: Executing string with python code: import slicer.testing; slicer.testing.runUnitTest(['/projects/birn/nicole/Slicer4/Reporting-build/Testing/Python', '/projects/birn/nicole/Slicer4/Reporting/Testing/Python'], 'ReportingRoundTripTest')
5: **
5: runUnitTest, path =
5: ['/projects/birn/nicole/Slicer4/Reporting-build/Testing/Python', '/projects/birn/nicole/Slicer4/Reporting/Testing/Python']
5: testname = ReportingRoundTripTest
5: -------------------------------------------
5: path: ['/projects/birn/nicole/Slicer4/Reporting-build/Testing/Python', '/projects/birn/nicole/Slicer4/Reporting/Testing/Python']
5: testname: ReportingRoundTripTest
5: -------------------------------------------
5: test_RoundTrip (ReportingRoundTripTest.ReportingTesting) ... ERROR
5:
5: ======================================================================
5: ERROR: test_RoundTrip (ReportingRoundTripTest.ReportingTesting)
5: ----------------------------------------------------------------------
5: Traceback (most recent call last):
5: File "/projects/birn/nicole/Slicer4/Reporting/Testing/Python/ReportingRoundTripTest.py", line 27, in test_RoundTrip
5: l = slicer.modules.reporting.logic()
5: AttributeError: 'module' object has no attribute 'reporting'
5:
5: ----------------------------------------------------------------------
5: Ran 1 test in 0.012s
5:
5: FAILED (errors=1)
5: Traceback (most recent call last):
5: File "<string>", line 1, in <module>
5: File "/projects/birn/nicole/Slicer4/S4-SuperBuild/Slicer-build/bin/Python/slicer/testing.py", line 26, in runUnitTest
5: exitFailure()
5: File "/projects/birn/nicole/Slicer4/S4-SuperBuild/Slicer-build/bin/Python/slicer/testing.py", line 10, in exitFailure
5: raise Exception(message)
5: Exception
I checked in the change in the Reporting/Testing/Python/CMakeLists.txt that will make the test fail more obviously. |
|
|
|
See the dashboard build with the test failing to find slicer.modules.reporting.logic() here:
http://slicer.cdash.org/testDetails.php?test=2534885&build=30461 |
|
|
|
Fixed in topic "22-fix-ReportingRoundTripTest"
See https://github.com/fedorov/Reporting/issues/22
and https://github.com/jcfr/Reporting/compare/22-fix-ReportingRoundTripTest |
|
|
|
I also created the corresponding issue on the ExtensionsIndex. See https://github.com/Slicer/ExtensionsIndex/issues/42 |
|
|
|
Commit r20710 will resolve the related issue described here:
https://github.com/fedorov/Reporting/issues/22#issuecomment-7331259
See http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&revision=20710 |
|
|
|
It is not clear how to pass additional module paths to generic unti tests, is this possible?
My python module needs to load Editor module, if I specify the module path explicitly in a test macro, it works, but I can only do this for non-generic tests.
Also, the generic tests cannot load Editor module, give errors and should be failing, but instead they pass when I run ctest. |
|
|
|
Here's example output (Reporting extension from this commit https://github.com/fedorov/Reporting/commit/58aa91405883ab3820f9fcfc6f1d02a1373eff8d):
1: Test command: /Users/fedorov/github/Slicer-Superbuild-Debug-Qt-Debug/Slicer-build/Slicer "--launcher-no-splash" "--launcher-additional-settings" "/Users/fedorov/github/Reporting-build-Release-Qt-Debug/AdditionalLauncherSettings.ini" "--launch" "/Users/fedorov/github/Reporting-build-Release-Qt-Debug/Testing/Cxx/qSlicerReportingModuleCxxTests" "qSlicerReportingModuleGenericTest"
1: Test timeout computed to be: 1500
1: [Reporting 08/05/2012 22:24:10]: WARNING: DICOM Database is not accessible.
1: PythonQt: QObject::connect() signal 'mrmlSceneChanged(vtkMRMLScene)' does not exist on qMRMLNodeComboBox
1: PythonQt: QObject::connect() signal 'mrmlSceneChanged(vtkMRMLScene)' does not exist on qMRMLNodeComboBox
1: PythonQt: QObject::connect() signal 'destroyed(QObject)' does not exist on QFrame
1: ERROR: In /Users/fedorov/github/Reporting/Logic/vtkSlicerReportingModuleLogic.cxx, line 1681
1: vtkSlicerReportingModuleLogic (0x115ab0320): GetActiveReportHierarchyID: no parameter reportID on node with id vtkMRMLScriptedModuleNode1
1:
1:
1: PythonQt: QObject::connect() signal 'destroyed(QObject)' does not exist on QFrame
1: Traceback (most recent call last):
1: File "/Users/fedorov/github/Reporting-build-Release-Qt-Debug/lib/Slicer-4.1/qt-loadable-modules/Python/qSlicerReportingModuleWidget.py", line 189, in setup
1: self.__editorWidget.setup()
1: File "/Users/fedorov/github/Slicer-Superbuild-Debug-Qt-Debug/Slicer-build/lib/Slicer-4.1/qt-scripted-modules/Editor.py", line 213, in setup
1: self.helper.mergeValidCommand = self.updateLabelFrame
1: AttributeError: 'NoneType' object has no attribute 'mergeValidCommand'
1/2 Test 0000001: qSlicerReportingModuleGenericTest ......... Passed 5.76 sec
This should NOT Pass! |
|
|
|
Fixed in r21046
See http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&revision=21046
Consider also topic associated with Reporting module. See https://github.com/jcfr/Reporting/compare/12-fix-testing-for-out-of-slicer-source-build |
|
|
|
JC, I updated the trunk and merged your fixes to Reporting, but when I run RoundTripTest, there is no evidence the test actually runs -- no print() statements show up, and the test passes:
https://github.com/fedorov/Reporting/blob/master/Testing/Python/ReportingRoundTripTest.py
So I do not know if the issue with uninitialized logic() is fixed, because I do not know if the test is running. |
|
|
|
Fixed in r21066
See http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&revision=21066
Consider also the following commit to fix the Reporting module. See https://github.com/jcfr/Reporting/commit/ae2cfdf52138a606df4045d4154931599c68e6ea |
|
|
|
Thank you |
|
