[kepler-dev] questions about matlab actor in snow leopard.

[Christopher Brooks]

Hi Jianwu,
The free could be superfluous.  It would be interesting to
try this code using electric fence or valgrind, but I'm not sure
how well those work with JNI.

The ptmatlab code is fairly old, if you get a fix, feel free to check
it in.

_Christopher

On 4/26/10 5:29 PM, Jianwu Wang wrote:
> Hi Christopher,
>
> Thanks for your quick update. It works for me. I replaced the
> $Kepler/common/lib/ptolemy/matlab/libptmatlab.dylib by the new generated
> libptmatlab.jnilib in $PTII/ptolemy/ptolemy/makefile directory and
> tested the $Kepler/outreach/resources/demos/Matlab/MatlabExpression.xml.
> The workflow was running well. But the whole Kepler crashed when I tried
> to stop the execution.
>
> A pop up window shows that 'Kepler quit unexpectedly while using the
> libptmatlab.jnilib plugin.' Its detailed message shows that:
>
> Thread 30 Crashed: Java: MatlabExpression-debug
> 0 libSystem.B.dylib 0x00007fff8557e01e __semwait_signal_nocancel + 10
> 1 libSystem.B.dylib 0x00007fff8557df20 nanosleep$NOCANCEL + 129
> 2 libSystem.B.dylib 0x00007fff855da902 usleep$NOCANCEL + 57
> 3 libSystem.B.dylib 0x00007fff855f9eb8 abort + 93
> 4 libSystem.B.dylib 0x00007fff85511a75 free + 128
> 5 libptmatlab.jnilib 0x000000012f6788f8
> Java_ptolemy_matlab_Engine_ptmatlabEngClose + 170
>
>
> Dan and I dug it a little bit, and found out it is because of the
> sentence 162 in ptmatlab.cc : 'free(output_buf);'. We tested the
> malloc_size(output_buf) function before free() function, and it returned
> 0. The console message shows that:
>
> [run] ptmatlabCreateDoubleMatrix(input) 25 x 25
> [run] ptmatlabDestroy(input)
> [run] matlabEngine.evalString("surf(input)")
> [run] matlabEngine.close(5638846816)
> [run] ptmatlabEngClose: calling engClose(1343879520)
> [run] java(49363,0x14e000000) malloc: *** error for object 0xde5600:
> pointer being freed was not allocated
> [run] *** set a breakpoint in malloc_error_break to debug
> [run] ptmatlabEngClose: freeing 0xde5600
>
>
> I remember I also met the console message when I using 32 bit leopard
> with 64bit Matlab and Java 1.6 64 bit. But Kepler doesn't crash at
> leopard. Do you have any idea about it? Thanks a lot!
>
> Christopher Brooks wrote:
>> Hi Jianwu,
>> I just updated $PTII/ptolemy/ptolemy/makefile so that it will build
>> either
>> * 32bit
>> * 64bit
>> * Universal (32bit and 64bit).
>>
>> If you have a chance, please update your tree and rerun "make".
>>
>> Thanks,
>>
>> _Christopher
>>
>> On 4/26/10 4:01 PM, Jianwu Wang wrote:
>>> Hi Chistopher,
>>>
>>> My machine is snow leopard (10.6.3, 64 bit), matlab 2009b 64 bit, Java
>>> 1.6 64 bit. I tried to use 'make' command in ptII/ptolemy/matlab
>>> directory to generate libptmatlab file. But it has the following error.
>>> I think it is because my matlab installation do not have maci directory,
>>> only have maci64 directory. Do you think the problem is because of my
>>> installation or make file? Thanks.
>>>
>>> echo $DYLD_LIBRARY_PATH
>>> /Applications/MATLAB64/MATLAB_R2009b/MATLAB_R2009b.app/bin/maci64:/Applications/MATLAB64/MATLAB_R2009b/MATLAB_R2009b.app/sys/os/maci64
>>>
>>>
>>>
>>> kepler:matlab jianwu$ make
>>> if [ "`uname -s`" = "Darwin" ]; then \
>>> make libptmatlabUniversal.jnilib; \
>>> mv libptmatlabUniversal.jnilib libptmatlab.jnilib; \
>>> file libptmatlab.jnilib; \
>>> else \
>>> CLASSPATH="../.."
>>> "/System/Library/Frameworks/JavaVM.framework/Versions/1.6.0/Home/bin/javah"
>>>
>>> -jni -o ptmatlab.h ptolemy.matlab.Engine; \
>>> "/usr/bin/gcc" -DPT_NO_ENGGETARRAY -DPT_NO_ENGPUTARRAY
>>> -DPT_NO_MXGETNAME \
>>> "-I/Applications/MATLAB64/MATLAB_R2009b/MATLAB_R2009b.app/extern/include"
>>> \
>>> "-I/System/Library/Frameworks/JavaVM.framework/Versions/1.6.0/Home/../Headers"
>>>
>>> \
>>> "-I/System/Library/Frameworks/JavaVM.framework/Versions/1.6.0/Home/../include/darwin"
>>>
>>> \
>>> ptmatlab.cc -fno-exceptions -dynamiclib -o libptmatlab32.jnilib
>>> -L"/Applications/MATLAB64/MATLAB_R2009b/MATLAB_R2009b.app/bin/maci"
>>> -leng -lmx -lmex; \
>>> fi
>>> CLASSPATH="../.."
>>> "/System/Library/Frameworks/JavaVM.framework/Versions/1.6.0/Home/bin/javah"
>>>
>>> -jni -o ptmatlab.h ptolemy.matlab.Engine
>>> "/usr/bin/gcc" -DPT_NO_ENGGETARRAY -DPT_NO_ENGPUTARRAY
>>> -DPT_NO_MXGETNAME \
>>> "-I/Applications/MATLAB64/MATLAB_R2009b/MATLAB_R2009b.app/extern/include"
>>> \
>>> "-I/System/Library/Frameworks/JavaVM.framework/Versions/1.6.0/Home/../Headers"
>>>
>>> \
>>> "-I/System/Library/Frameworks/JavaVM.framework/Versions/1.6.0/Home/../include/darwin"
>>>
>>> \
>>> ptmatlab.cc -fno-exceptions -dynamiclib -o libptmatlab32.jnilib
>>> -L"/Applications/MATLAB64/MATLAB_R2009b/MATLAB_R2009b.app/bin/maci"
>>> -leng -lmx -lmex
>>> ld: warning: directory
>>> '/Applications/MATLAB64/MATLAB_R2009b/MATLAB_R2009b.app/bin/maci'
>>> following -L not found
>>> ld: library not found for -leng
>>> collect2: ld returned 1 exit status
>>> make[1]: *** [libptmatlab32.jnilib] Error 1
>>> mv: rename libptmatlabUniversal.jnilib to libptmatlab.jnilib: No such
>>> file or directory
>>> libptmatlab.jnilib: cannot open `libptmatlab.jnilib' (No such file or
>>> directory)
>>>
>>
>
>

-- 
Christopher Brooks, PMP                       University of California
CHESS Executive Director                      US Mail: 337 Cory Hall
Programmer/Analyst CHESS/Ptolemy/Trust        Berkeley, CA 94720-1774
ph: 510.643.9841 fax:510.642.2718	      (Office: 545Q Cory)
home: (F-Tu) 707.665.0131 cell: 707.332.0670