[kepler-dev] Matlab interface on the Mac
Christopher Brooks
cxh at eecs.berkeley.edu
Thu Jun 18 12:36:49 PDT 2009
Hi Dan,
Thanks, I tuned up ptmatlab.cc and can now run under
Mac OS X with r2009a. I folded in your change
and made other changes.
I had a look at the includes and it appears that
The Mathworks broke backward compatibility between r2008a
and r2009a - I don't see an easy way to use anything defined
in the matlab includes to determine what needs to be
defined in r2009a vs r2008a. The Mathworks have made this
mistake in the past, so nothing new here.
Under all platforms, I'm now having $PTII/configure set
-DPT_NO_ENGGETARRAY -DPT_NO_ENGPUTARRAY and -DPT_NO_MXGETNAME.
I think these _might_ be specific to Matlab r2009a, but apparently
do not cause problems when I compile with r2008a under Linux
with those defines. So, I've defined these defines for all platforms
as they will work in both r2008a and r2009a.
The problem I reported with the java stack trace is no longer appearing
when the Ptolemy/Matlab interface runs, but I do see the
stack trace when I invoke the Matlab executable.
It would be interesting if you upgraded to r2009a before applying
the Java patch, started matlab verified that no stack trace appeared
and then applied the patch and started matlab again.
It sounds like my change to UtilityFunctions.loadProperty() is
not yet causing problems.
I think our work here is done . . .
(Famous last words)
_Christopher
Daniel Crawl wrote:
>
> Hi Christopher,
>
> Thanks for letting me know about the updates.
>
> I updated my sources, recompiled, and also got an error
> when using Matlab 2009a. It appears to be caused by line
> 209 of ptmatlab.cc:
>
> int retval = mexPutVariable("global", str, ma);
>
> If I replace it with:
>
> int retval = engPutVariable(ep, str, ma);
>
> it works! According to:
>
> http://www.mathworks.com/access/helpdesk/help/techdoc/index.html?/access/helpdesk/help/techdoc/rn/f34-998197.html
>
>
> engPutArray() should be replaced with engPutVariable().
>
> However, I haven't applied the latest Apple Java patch:
>
> # java -version
> java version "1.5.0_16"
> Java(TM) 2 Runtime Environment, Standard Edition (build 1.5.0_16-b06-284)
> Java HotSpot(TM) Client VM (build 1.5.0_16-133, mixed mode, sharing)
>
>
> Do you want ptmatlab.cc to compile with older versions of
> Matlab? I tried with R2008a:
>
> # make
> CLASSPATH="../.." "/usr/bin/javah" -jni -o ptmatlab.h ptolemy.matlab.Engine
> "/usr/bin/gcc" -DPT_NO_ENGGETARRAY -DPT_NO_ENGPUTARRAY \
> "-I/Applications/MATLAB_R2008a/extern/include" \
>
> "-I/System/Library/Frameworks/JavaVM.framework/Versions/1.5.0/Home/../Headers"
> \
>
> "-I/System/Library/Frameworks/JavaVM.framework/Versions/1.5.0/Home/../include/darwin"
> \
> ptmatlab.cc -fno-exceptions -dynamiclib -o libptmatlab.dylib
> -L"/Applications/MATLAB_R2008a/bin/maci" -leng -lmx -lmex
> /Applications/MATLAB_R2008a/extern/include/matrix.h:1231: error:
> expected unqualified-id before string constant
> /Applications/MATLAB_R2008a/extern/include/matrix.h:1231: error:
> expected initializer before string constant
>
> This error is caused by mxGetName() defined at the top of
> ptmatlab.cc and then in matrix.h. It shouldn't be defined
> if the version < R2009a.
>
> Also, I think the deprecated functions were removed for
> all platforms, not just Mac. I haven't installed 2009a yet
> on my Linux and Windows boxes, but will try to verify in
> the next few days.
>
> --dan
>
>
> Christopher Brooks wrote:
>> Hi Daniel,
>> I'm working on getting the Matlab interface to run on my Mac.
>>
>> I had to make a few changes, see below.
>> The biggest change is that in UtilityFunctions.loadProperty(),
>> I'm now optionally prepending "lib" to the path on the Mac.
>>
>> Do you see this as a problem? I just checked in my change to the
>> ptII tree, when you have a chance, could you build from the current
>> head (not the stable version) and check it?
>>
>> One other issue is that the interface does not work right now for
>> me with Matlab2009a. When I try to run one of the demos, I get
>>
>> Exception in thread "AWT-EventQueue-0" java.lang.NullPointerException
>> at
>> apple.awt.CGraphicsEnvironment.resetDisplays(CGraphicsEnvironment.java:99)
>>
>> at
>> apple.awt.CGraphicsEnvironment.displayChanged(CGraphicsEnvironment.java:49)
>>
>> at apple.awt.CToolkit$4.run(CToolkit.java:1310)
>> at java.awt.event.InvocationEvent.dispatch(InvocationEvent.java:209)
>> at java.awt.EventQueue.dispatchEvent(EventQueue.java:461)
>> at
>> java.awt.EventDispatchThread.pumpOneEventForHierarchy(EventDispatchThread.java:269)
>>
>> at
>> java.awt.EventDispatchThread.pumpEventsForHierarchy(EventDispatchThread.java:190)
>>
>> at
>> java.awt.EventDispatchThread.pumpEvents(EventDispatchThread.java:184)
>> at
>> java.awt.EventDispatchThread.pumpEvents(EventDispatchThread.java:176)
>> at java.awt.EventDispatchThread.run(EventDispatchThread.java:110)
>>
>>
>> Interestingly, I get a similar stack trace when I start
>> Matlab outside Ptolemy. I suspect the recent patch to Java from
>> Apple. I've reported this as a bug to The Mathworks, we'll see
>> what happens.
>>
>> It might be interesting to try the Ptolemy/Matlab interface before
>> patching
>> Java and then after. After the Apple Java patch, I'm running:
>>
>> bash-3.2$ java -version
>> java version "1.5.0_19"
>> Java(TM) 2 Runtime Environment, Standard Edition (build 1.5.0_19-b02-304)
>> Java HotSpot(TM) Client VM (build 1.5.0_19-137, mixed mode, sharing)
>>
>> The complete changes are below.
>>
>> _Christopher
>>
>>> Author: cxh
>>> Date: 2009-06-17 14:19:59 -0700 (Wed, 17 Jun 2009)
>>> New Revision: 54467
>>>
>>> Modified:
>>> trunk/configure.in
>>> trunk/ptolemy/data/expr/UtilityFunctions.java
>>> trunk/ptolemy/matlab/ptmatlab.cc
>>> Log:
>>> Changes to get Matlab 2009a to run under Mac OS X: engGetArray() and
>>> engPutArray() are no longer defined. We need to include -lmex.
>>> UtilityFunctions.loadLibrary() now prepends 'lib' if necessary. Some
>>> of these changes might be generic to Matlab 2009a and not specific to
>>> Mac OS X.
>>>
>>> Modified: trunk/configure.in
>>> ===================================================================
>>> --- trunk/configure.in 2009-06-17 19:38:33 UTC (rev 54466)
>>> +++ trunk/configure.in 2009-06-17 21:19:59 UTC (rev 54467)
>>> @@ -2898,8 +2898,8 @@
>>> AC_MSG_CHECKING([which major type of OS we are running
>>> under])
>>> # Older versions of cygwin might not have env, but 1.1 does
>>> if env CLASSPATH="$PTII/config" "$JAVA" MajorOSName >
>>> /dev/null; then
>>> - # Run it again and get the results
>>> - MAJOR_OS_NAME=`CLASSPATH="$PTII/config" "$JAVA"
>>> MajorOSName`
>>> + # Run it again and get the results. We get just the
>>> first word: Windows, Linux, Mac etc.
>>> + MAJOR_OS_NAME=`CLASSPATH="$PTII/config" "$JAVA"
>>> MajorOSName | awk '{print $1}'`
>>> PT_MSG_RESULT($MAJOR_OS_NAME)
>>> else AC_MSG_WARN(['CLASSPATH="$PTII/config"
>>> $JAVA MajorOSName' failed.
>>> @@ -2945,6 +2945,7 @@
>>> ;;
>>> Darwin)
>>> MATLAB_LIBDIR=${MATLAB_DIR}/bin/maci
>>> + PTMATLAB_CC_FLAGS="-DPT_NO_ENGGETARRAY
>>> -DPT_NO_ENGPUTARRAY" ;;
>>> *)
>>> MATLAB_LIBDIR=${MATLAB_DIR}/extern/lib
>>> @@ -2974,7 +2975,7 @@
>>> fi
>>> else if test $PTJNI_ARCHITECTURE =
>>> "darwin"; then
>>> - PTMATLAB_LD_ARGS="ptmatlab.cc -fno-exceptions
>>> -dynamiclib -o
>>> ${PTJNI_SHAREDLIBRARY_PREFIX}ptmatlab.${PTJNI_SHAREDLIBRARY_SUFFIX}
>>> -L\"${MATLAB_LIBDIR}\" ${PTJNI_SHAREDLIBRARY_LDFLAG} -leng -lmx"
>>> + PTMATLAB_LD_ARGS="ptmatlab.cc -fno-exceptions
>>> -dynamiclib -o
>>> ${PTJNI_SHAREDLIBRARY_PREFIX}ptmatlab.${PTJNI_SHAREDLIBRARY_SUFFIX}
>>> -L\"${MATLAB_LIBDIR}\" ${PTJNI_SHAREDLIBRARY_LDFLAG} -leng -lmx -lmex"
>>> else
>>> PTMATLAB_LD_ARGS="-shared ptmatlab.cc
>>> -fno-exceptions -o
>>> ${PTJNI_SHAREDLIBRARY_PREFIX}ptmatlab.${PTJNI_SHAREDLIBRARY_SUFFIX}
>>> -L\"${MATLAB_LIBDIR}\" ${PTJNI_SHAREDLIBRARY_LDFLAG} -lc -leng -lmx"
>>> if test "`uname -m`" = "x86_64"; then
>>>
>>> Modified: trunk/ptolemy/data/expr/UtilityFunctions.java
>>> ===================================================================
>>> --- trunk/ptolemy/data/expr/UtilityFunctions.java 2009-06-17
>>> 19:38:33 UTC (rev 54466)
>>> +++ trunk/ptolemy/data/expr/UtilityFunctions.java 2009-06-17
>>> 21:19:59 UTC (rev 54467)
>>> @@ -859,11 +859,13 @@
>>> shortLibraryName = library;
>>> } else {
>>> if (!library.substring(index, index +
>>> 4).equals("/lib")) {
>>> - if (osName.startsWith("Mac OS X")
>>> - || osName.startsWith("Linux")) {
>>> + if (osName.startsWith("Linux")) {
>>> library = library.substring(index + 1);
>>> shortLibraryName = library;
>>> } else {
>>> + // Under SunOS and Mac OS X, we add lib to
>>> + // the path. If we don't do this on the
>>> Mac,
>>> + // then libptymatlab.dynlib will not be
>>> found.
>>> shortLibraryName = "/lib"
>>> + library.substring(index + 1);
>>> library = library.substring(0, index)
>>>
>>> Modified: trunk/ptolemy/matlab/ptmatlab.cc
>>> ===================================================================
>>> --- trunk/ptolemy/matlab/ptmatlab.cc 2009-06-17 19:38:33 UTC (rev
>>> 54466)
>>> +++ trunk/ptolemy/matlab/ptmatlab.cc 2009-06-17 21:19:59 UTC (rev
>>> 54467)
>>> @@ -52,10 +52,36 @@
>>> // version of the ptmatlab shared library to load...
>>> #define V5_COMPAT
>>>
>>> +#define PT_NO_MXGETNAME
>>> +#ifdef PT_NO_MXGETNAME
>>> +#define mxGetName(a) "a"
>>> +#endif
>>> +
>>> #include <jni.h>
>>> #include "ptmatlab.h"
>>> #include "engine.h"
>>>
>>> +// These defines were needed for Matlab r2009a under Mac OS X
>>> +//#define PT_NO_ENGGETARRAY +//#define PT_NO_ENGPUTARRAY
>>> +
>>> +
>>> +#ifdef PT_NO_ENGGETARRAY
>>> +#ifndef PT_INCLUDE_MEX_H
>>> +#define PT_INCLUDE_MEX_H
>>> +#endif
>>> +#endif
>>> +
>>> +#ifdef PT_NO_ENGPUTARRAY
>>> +#ifndef PT_INCLUDE_MEX_H
>>> +#define PT_INCLUDE_MEX_H
>>> +#endif
>>> +#endif
>>> +
>>> +#ifdef PT_INCLUDE_MEX_H
>>> +#include "mex.h"
>>> +#endif
>>> +
>>> // The following test is a kludge for a missing #define in the matlab
>>> // include files that would allow engine interface C code to test which
>>> // matlab version it is dealing with... This is a test for an earlier
>>> @@ -155,7 +181,11 @@
>>> Engine *ep = (Engine*)(ptrlong) e;
>>> char debug = ptmatlabGetDebug(jni,obj);
>>> const char *str = jni->GetStringUTFChars(name, 0);
>>> +#ifdef PT_NO_ENGGETARRAY + mxArray *ma = engGetVariable(ep, str);
>>> +#else
>>> mxArray *ma = engGetArray(ep, str);
>>> +#endif
>>> if (debug > 1 && ma != NULL) {
>>> const int *dimArray = mxGetDimensions(ma);
>>> printf("ptmatlabEngGetArray(%s) %d x %d\n", str,
>>> dimArray[0], dimArray[1]);
>>> @@ -175,10 +205,14 @@
>>> Engine *ep = (Engine*)(ptrlong) e;
>>> const char *str = jni->GetStringUTFChars(name, 0);
>>> mxArray *ma = (mxArray*)(ptrlong)pma;
>>> +#ifdef PT_NO_ENGPUTARRAY
>>> + int retval = mexPutVariable("global", str, ma);
>>> +#else
>>> mxSetName(ma, str);
>>> const int *dimArray = mxGetDimensions(ma);
>>> if (debug > 1) printf("ptmatlabEngPutArray(%s) %d x %d\n", str,
>>> dimArray[0], dimArray[1]);
>>> int retval = engPutArray(ep, ma);
>>> +#endif
>>> jni->ReleaseStringUTFChars(name, str);
>>> return retval;
>>> }
>>>
>>> _______________________________________________
>>> Ptexternal-cvs mailing list
>>> Ptexternal-cvs at chess.eecs.berkeley.edu
>>> http://chess.eecs.berkeley.edu/ptexternal/listinfo/ptexternal-cvs
>>
>
>
--
Christopher Brooks (cxh at eecs berkeley edu) 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 (W-F) 510.655.5480
More information about the Kepler-dev
mailing list