[ome-users] Matlap trouble "mex -O"

[Valentin Dimitrov]

Hi Tom,

I had to replace the "-f" flag with "-" flag in
OME/src/perl2/OME/Install/Util.pm

but the problem that I have now looks quite bad:
#############
MATLAB Perl API configuration:
  Install MATLAB Perl API?: yes
               MATLAB User: valentin
               MATLAB Path: /usr/local/matlab7
               MATLAB Exec: /usr/local/matlab7/bin/matlab
    Config MATLAB for dev?: no
      MATLAB .m files Path: /OME/matlab

Are these values correct ? [y/n]:
Installing MATLAB Perl API
   \_ Configuring [SUCCESS].
   \_ Compiling [SUCCESS].
   \_ Testing [SUCCESS].
   \_ Installing [SUCCESS].
Installing MEX Interface to omeis-http
   \_ Configuring omeis-http with MATLAB bindings [SUCCESS].
   \_ Compiling omeis-http with MATLAB bindings [SUCCESS].
Installing OME MATLAB .m files
   \_ Compiling MEX files [FAILURE].


Errors executing task: Unable to compile MATLAB .m files, see 
PerlModuleTask.log for details. at (eval 35) line 1

################################## PerlModuleTask.log

COMPILING MODULE -- 'make'
FAILURE -- OUTPUT: "make -C OMEIS/
  make[1]: Entering directory `/opt/OME-CVS/OME/src/matlab/OMEIS'
  mex -O -inline -I../../C/omeis-http/ -L../../C/omeis-http  -lcurl 
-lhttpOMEIS newPixels.c
  building newPixels.c
  + mkoctfile -DHAVE_OCTAVE_21 -v -o newPixels.oct mex_newPixels.cc 
/usr/lib/octave/2.1.69/site/oct/i386-pc-linux-gnu/octave-
forge/mex.o 
-I/usr/lib/octave/2.1.69/site/oct/i386-pc-linux-gnu/octave-forge -O 
-inline -I../../C/omeis-http/ -L../../C/omei
s-http -lcurl -lhttpOMEIS newPixels.c
  mkoctfile: unrecognized argument -O
  make[1]: *** [newPixels.mex*] Error 1
  make[1]: Leaving directory `/opt/OME-CVS/OME/src/matlab/OMEIS'
  make: *** [omeis_http] Error 2

###########################

Are you sure about the "-O" flag ?


many thx in advance
Cheers
Valentin









+------------------------------------------+
  Valentin Dimitrov
  TEL:   +49 (0)6841 1626094
  MOBIL: +49 (0)176 27017006
  WWW: http://www.ags.uni-sb.de/~valentin
+------------------------------------------+

On Wed, 24 Aug 2005, T. J. Macura wrote:

> Hey Valentin,
>
>> I am using Debian 3.1 with 2.6.8 kernel.
> We don't "support" Debian in so far as we don't have it installed and don't 
> test on it.
> You can read what Ilya Goldberg has to say about Debian and OME support for 
> it:
> 	http://lists.openmicroscopy.org.uk/pipermail/ome-devel/2005-August/ 
> 000493.html
>
> Some people were getting OME to install on certain versions of Debian without 
> incident. I am glad to help you fix things, but I am just explaining why 
> thinks don't always work out-of-the-box.
>
>> Unfortunately Debian doesn't know the "-f" flag , but if
>> you are trying to get a full login i.e. /home/<user>
>> "su -" is uder Debian sufficient.
> This is good to know. I suspect that the -l/-f etc. flags are now 
> unnecessary. If they are needed I will have to add some OS specific logic 
> there argh !
>
>> ######################################### PerlModuleTask.log
>> 
>> TESTING MODULE -- su -  valentin -c 'make test '
>> FAILURE -- OUTPUT: "Makefile out-of-date with respect to Makefile.PL
>>  Cleaning current config before rebuilding Makefile...
> <snip>
>
>> Do you think I should chage the Makefile as well?
> Is this from CVS? I fixed this very bug in CVS. Please try from CVS and 
> modify the su there as you did here.
> If you don't want to use CVS maybe we could work around this installer bug. 
> However, the MATLAB/OME connection in CVS is completely different than the 
> connection in 2.4.0. So many bugs you encounter with the Matlab Handler 
> (that's the part of  the Analysis Engine that executes Matlab OME Analysis 
> Modules) in OME 2.4.0, will not be relevant.
>
> Tom
>
>