[ome-devel] OME and matlab - trouble connecting

[debartolomeo at ece.ucsb.edu]

Hi Arpun,

Thanks for your suggestions.  I think that the setPixels problem is solved
with the switch to GCC 4.0.  However, I am still getting the error message
(both from running the Matlab Test Chain in the UI and from the command
line):
"Cannot open a connection to Matlab! at /usr/share/perl/5.8/CGI/Carp.pm
line 314."

I've altered the MatlabHandler.pm file per your suggestion and reinstalled
OME (and added some print statements to check that the modified .pm was
being used) with no luck.

Any suggestions?

Melissa

> I ran into the setPixels problem as well.  There was also a lazy
> symbol binding problem that I solved by switching to GCC 4.0.  Here
> is what worked for me:
>
> Modify the Makefile in your src/matlab/OMEIS directory.  Change the
> MEX_FLAGS line to this:
>
> MEX_FLAGS = -O -inline -I../../C/omeis-http/ -L../../C/omeis-http  -
> lcurl -lhttpOMEISMATLAB CC=gcc-4.0 LD=gcc-4.0
>
> You should remove all the *.mexglx files.  Then reinstall OME - when
> it installs the matlab connection, it should compile with these new
> flags.
>
> The Web UI problems are a whole different story.  We have been
> working to fix them, and the only real solution we found was to
> modify the _openEngine function in the src/perl2/OME/Analysis/
> Handlers/MatlabHandler.pm
>
> There is a line in _openEngine that looks something like this:
> $_engine = OME::Matlab::Engine->open("$matlab_exec $matlab_flags");
>
> Modify it so that it looks like this:
> $_engine = OME::Matlab::Engine->open("env PATH=/usr/bin:/bin
> $matlab_exec $matlab_flags");
>
> You may have to reinstall OME at this point, to make sure it's using
> this copy of the MatlabHandler.  For reasons I haven't been able to
> discern, it sometimes doesn't use your modified version of the
> Handler.  It might be helpful to put some print statements around
> these calls to make sure everything is doing what it's supposed to do.
>
> I'm not certain this will work on Linux.  One of the reasons we
> haven't committed this stuff is because we didn't get a chance to
> test it on non-Mac systems.  Looks like you'll be our guinea pig.
>
> Let us know how things turn out.  Good luck!
>
> -Arpun
>
> On Jul 28, 2006, at 6:30 PM, debartolomeo at ece.ucsb.edu wrote:
>
>> Hi,
>>
>> I'm trying to get Matlab working with ome and am running into
>> problems.
>>
>> When I try to use the UI to run the analysis chain, I get the
>> following
>> error in the UI:
>> "Cannot open a connection to Matlab! at /usr/share/perl/5.8/CGI/
>> Carp.pm
>> line 314"
>>
>> But, when I try this from the command line:
>> ome execute -d "MyDataset" -a "Matlab Test Chain", I get the following
>> error in the UI:
>> "ERROR saving pixels to omeis w/ command:
>> 	ome_output_9_PixelsArray_pix =
>> setPixels(openConnectionOMEIS('http://claw/cgi-bin/omeis'), 686,
>> ome_output_9_PixelsArray);
>>>> ??? Invalid MEX-file
>> '/home/omedev/OME/src/matlab/OMEIS/setPixels.mexglx':
>> /home/matlab/bin/glnx86/../../sys/os/glnx86/libgcc_s.so.1: version
>> `GCC_3.3' not found (required by /usr/lib/libstdc++.so.6)."
>>
>> Again, running "Histogram Equalization Preprocessing Chain" from
>> the UI, I
>> get:
>> "Cannot open a connection to Matlab! at /usr/share/perl/5.8/CGI/
>> Carp.pm
>> line 314."
>>
>> But, when I try:
>> ome execute -d "MyDataset" -a "Histogram Equalization Preprocessing
>> Chain", I get:
>> "getROI/getPixels failed for pixels 482. The returned MATLAB array has
>> dimensions (0, 0, 1, 1, 1) at
>> /usr/local/share/perl/5.8.4/OME/Analysis/Handlers/MatlabHandler.pm
>> line
>> 562,  line 1."
>>
>> I'm new at this system, so there may be something very fundamental
>> that
>> I'm not yet understanding, but any suggestions on how to proceed
>> would be
>> very appreciated.
>>
>>
>> _______________________________________________
>> ome-devel mailing list
>> ome-devel at lists.openmicroscopy.org.uk
>> http://lists.openmicroscopy.org.uk/mailman/listinfo/ome-devel
>