[CMake] Random behavior of CTEST_USE_LAUNCHERS

[Julien Malik]

Hi again,

I'm still looking for an idea about where to look.

For the nightly build dir where CTEST_USE_LAUNCHERS is set in the ctest 
script but the dashboard behaves like it wasn't, I can see in VERBOSE mode :

[ 59%] Building CXX object 
Code/Learning/CMakeFiles/OTBLearning.dir/otbChangeProfileKernelFunctor.cxx.o
cd /home/otbval/Dashboard/nightly/OTB-Release/build/Code/Learning && 
"/usr/bin/ctest" --launch --target-name OTBLearning --build-dir 
/home/otbval/Dashboard/nightly/OTB-Release/build/Code/Learning --output 
CMakeFiles/OTBLearning.dir/otbChangeProfileKernelFunctor.cxx.o --source 
/home/otbval/Dashboard/nightly/OTB-Release/src/Code/Learning/otbChangeProfileKernelFunctor.cxx 
--language CXX -- /usr/bin/c++   -DOTBLearning_EXPORTS 
-D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -D_LARGEFILE64_SOURCE 
-D_LARGE_FILES -DLAS_DISABLE_DLL -DOTB_USE_MAPNIK 
-DOTB_USE_INTERNAL_OPENTHREADS -fPIC -Wall -Wno-deprecated 
-Wno-uninitialized -Wno-unused-variable -O3 -DNDEBUG -fPIC 
-I/usr/include/freetype2  <<< ...lots of stuff here... >>>    -o 
CMakeFiles/OTBLearning.dir/otbChangeProfileKernelFunctor.cxx.o -c 
/home/otbval/Dashboard/nightly/OTB-Release/src/Code/Learning/otbChangeProfileKernelFunctor.cxx
Linking CXX shared library ../../bin/libOTBLearning.so



This makes me think it is ok inside the build dir : build commands are 
correctly wrapped by ctest.
However, in the Build.xml file which is sent to CDash, I have things like :

<BuildCommand>/usr/bin/make -j8 -i -k install</BuildCommand>
         <Warning>
                 <BuildLogLine>287</BuildLogLine>
<Text>/usr/include/x86_64-linux-gnu/bits/stdio2.h:35:43: warning: call 
to int __builtin___sprintf_chk(char*, int, long unsigned int, const 
char*, ...) will always overflow destination buffer [enabled by default]
</Text>
<SourceFile>/usr/include/x86_64-linux-gnu/bits/stdio2.h</SourceFile>
                 <SourceLineNumber>35</SourceLineNumber>
                 <PreContext>[  4%] Building C object 
Utilities/otb6S/CMakeFiles/otb6S.dir/libf2c/due.c.o
[  4%] Building C object 
Utilities/otb6S/CMakeFiles/otb6S.dir/libf2c/h_nint.c.o
[  4%] Building C object 
Utilities/ITK/Utilities/vxl/v3p/netlib/CMakeFiles/itkv3p_netlib.dir/blas/dger.o
[  4%] Building C object 
Utilities/otb6S/CMakeFiles/otb6S.dir/libf2c/i_indx.c.o
[  4%] [  4%] Building C object 
Utilities/otb6S/CMakeFiles/otb6S.dir/libf2c/l_le.c.o
Building C object 
Utilities/ITK/Utilities/vxl/v3p/netlib/CMakeFiles/itkv3p_netlib.dir/blas/dlamch.o
In file included from /usr/include/stdio.h:934:0,
                  from 
/.../src/Utilities/ITK/Utilities/kwsys/SystemInformation.cxx:87:
In function 'int sprintf(char*, const char*, ...)',
     inlined from 'bool 
itksys::SystemInformationImplementation::RetrieveClassicalCPUIdentity()' 
at /.../src/Utilities/ITK/Utilities/kwsys/SystemInformation.cxx:2040:114:
</PreContext>
                 <PostContext>In function 'int sprintf(char*, const 
char*, ...)',
     inlined from 'bool 
itksys::SystemInformationImplementation::RetrieveClassicalCPUIdentity()' 
at /.../src/Utilities/ITK/Utilities/kwsys/SystemInformation.cxx:2041:119:
</PostContext>
                 <RepeatCount>0</RepeatCount>
</Warning>



Any idea where to look ?

Cheers,
Julien




Le 27/09/2012 18:39, Julien Malik a écrit :
>
> Sorry if it was not clear enough.
>
> When I say "does not work" I mean that I set CTEST_USE_LAUNCHERS to ON 
> and it behaves as if CTEST_USE_LAUNCHERS was OFF (apart from that 
> everything is ok in the submission).
> For example : 
> http://dash.orfeo-toolbox.org/viewBuildError.php?buildid=96726
> CMakeCache and ctest script are here : 
> http://dash.orfeo-toolbox.org/viewNotes.php?buildid=96726
> (NB : it that case I even have warnings recognized as errors by ctest, 
> but let's forget about this)
>
>
> When "it works" I have :
> http://dash.orfeo-toolbox.org/viewBuildError.php?type=1&buildid=96622
> which is the kind of output I would like to have anywhere possible.
>
>
> Thanks,
> Julien
>
>
>
>
> Le 27/09/2012 18:24, Bill Hoffman a écrit :
>> On 9/27/2012 10:50 AM, Julien Malik wrote:
>>>
>>> - a ubuntu 12.4 build [2] where it does NOT work (cmake 2.8.7)
>>> - a ubuntu 10.4 machine [3] where it does work (cmake 2.8.4)
>>> - a macosx 10.8 machine [4] where it does NOT work (cmake 2.8.9)
>> When you say it does not work, what do you mean it does not work?
>>