plplot-general

[Plplot-general] cmake & disable fortran

[Valery P. <pi...@is...>]

Dear Group,

I wanted to package plplot to rpm. I have used the spec from fedora-7. It uses 
cmake for building plplot. The question is how to disable some features, for 
example I don't need f95 and tcl. I'm sorry for asking such a simple 
questions in the list.

best regards
Valery 

Re: [Plplot-general] cmake & disable fortran

[Arjen M. <arj...@wl...>]

> Dear Group,
>
> I wanted to package plplot to rpm. I have used the spec from fedora-7. It
> uses
> cmake for building plplot. The question is how to disable some features,
> for
> example I don't need f95 and tcl. I'm sorry for asking such a simple
> questions in the list.
>

You can use the -DENABLE_fetaure option for language bindings:

cmake -DENABLE_f95:BOOL=OFF -DENABLE_tcl:BOOL=OFF

Be sure to start with a clean sheet (i.e. no CMakeCache.txt file etc
in the build directory).

(The above form may not be the minimal, but it works. For devices
you need the -DPLD_device option: -DPLD_ps:BOOL=OFF for instance).

Regards,

Arjen

Re: [Plplot-general] cmake & disable fortran

[Valery P. <pi...@is...>]

Dear Arjen,

Thanks for the fast reply!
On Tue, 29 May 2007 06:17:56 +0200 (CEST)
Arjen Markus wrote:

> > Dear Group,
> >
> > I wanted to package plplot to rpm. I have used the spec from
> > fedora-7. It uses
> > cmake for building plplot. The question is how to disable some
> > features, for
> > example I don't need f95 and tcl. I'm sorry for asking such a simple
> > questions in the list.
> >
> 
> You can use the -DENABLE_fetaure option for language bindings:
> 
> cmake -DENABLE_f95:BOOL=OFF -DENABLE_tcl:BOOL=OFF
Yes, I tried these just by analogy with options from configure script.
It does not help 


-- Check for using stdint.h with CXX compiler - ok
-- Check for working Fortran compiler: /usr/local/intel/bin/ifc
-- Check for working Fortran compiler: /usr/local/intel/bin/ifc --
broken CMake Error: The Fortran compiler "/usr/local/intel/bin/ifc" is
not able to compile a simple test program. It fails with the following
output: /usr/bin/gmake -f CMakeFiles/cmTryCompileExec.dir/build.make
CMakeFiles/cmTryCompileExec.dir/build gmake[1]: Entering directory
`/home/vv/RPM/BUILD/plplot-5.7.3/alt/CMakeFiles/CMakeTmp' /usr/bin/cmake
-E
cmake_progress_report /home/vv/RPM/BUILD/plplot-5.7.3/alt/CMakeFiles/CMakeTmp/CMakeFiles
1 Building Fortran object
CMakeFiles/cmTryCompileExec.dir/testFortranCompiler.o /usr/local/intel/bin/ifc
-o CMakeFiles/cmTryCompileExec.dir/testFortranCompiler.o  -pipe -Wall
-O2 -march=i686 -mtune=i686
-c /home/vv/RPM/BUILD/plplot-5.7.3/alt/CMakeFiles/CMakeTmp/testFortranCompiler.f
ifc: warning: The Intel Fortran driver is now named ifort.  You can
suppress this message with '-quiet' ifort: No such file or directory
gmake[1]: *** [CMakeFiles/cmTryCompileExec.dir/testFortranCompiler.o]
Error 100 gmake[1]: Leaving directory
`/home/vv/RPM/BUILD/plplot-5.7.3/alt/CMakeFiles/CMakeTmp' gmake: ***
[cmTryCompileExec/fast] Error 2

Re: [Plplot-general] cmake & disable fortran

[Valery P. <pi...@is...>]

Dear Arjen,

Thanks for the help. The problem disappeared when I  disable  f77
as well.
best regards,
Valery

On Tue, 29 May 2007 06:17:56 +0200 (CEST)
Arjen Markus wrote:

> > Dear Group,
> >
> > I wanted to package plplot to rpm. I have used the spec from
> > fedora-7. It uses
> > cmake for building plplot. The question is how to disable some
> > features, for
> > example I don't need f95 and tcl. I'm sorry for asking such a simple
> > questions in the list.
> >
> 
> You can use the -DENABLE_fetaure option for language bindings:
> 
> cmake -DENABLE_f95:BOOL=OFF -DENABLE_tcl:BOOL=OFF
> 
> Be sure to start with a clean sheet (i.e. no CMakeCache.txt file etc
> in the build directory).
> 
> (The above form may not be the minimal, but it works. For devices
> you need the -DPLD_device option: -DPLD_ps:BOOL=OFF for instance).
> 
> Regards,
> 
> Arjen
> 
> -------------------------------------------------------------------------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> http://sourceforge.net/powerbar/db2/
> _______________________________________________
> Plplot-general mailing list
> Plp...@li...
> https://lists.sourceforge.net/lists/listinfo/plplot-general
> 

Re: [Plplot-general] cmake & disable fortran

[Arjen M. <arj...@wl...>]

Valery Pipin wrote:

>Dear Arjen,
>
>Thanks for the help. The problem disappeared when I  disable  f77
>as well.
>best regards,
>Valery
>  
>
Hello Valery,

what problem was that? The bindings for f77 and f95 should be independent,
but perhaps your system has neither type of compiler?

Regards,

Arjen

Re: [Plplot-general] cmake & disable fortran

[Valery P. <pi...@is...>]

Hi!
> what problem was that? The bindings for f77 and f95 should be
> independent, but perhaps your system has neither type of compiler?
> 
I have intel fortran 9 installed in  /usr/local/intel.
I heard that libtool does not support ifort. Also, there is linking
problem  like :

/usr/bin/cmake -P
CMakeFiles/cmTryCompileExec.dir/cmake_clean_target.cmake /usr/local/intel/bin/ifort
-KPIC  -lm  -O2
"CMakeFiles/cmTryCompileExec.dir/testFortranCompiler.o"   -o
cmTryCompileExec -i_dynamic IPO link: can not find "(" ifort: error:
problem during multi-file optimization compilation (code 1) gmake[1]:
*** [cmTryCompileExec] Error 1 gmake[1]: Leaving directory
`/home/vv/RPM/BUILD/plplot-5.7.3/alt/CMakeFiles/CMakeTmp' gmake: ***
[cmTryCompileExec/fast] Error 2

Finally, I don't need fortran binding for plplot. I'm using python.

regards
V

Re: [Plplot-general] cmake & disable fortran

[Arjen M. <arj...@wl...>]

Valery Pipin wrote:

>Hi!
>  
>
>>what problem was that? The bindings for f77 and f95 should be
>>independent, but perhaps your system has neither type of compiler?
>>
>>    
>>
>I have intel fortran 9 installed in  /usr/local/intel.
>I heard that libtool does not support ifort. Also, there is linking
>problem  like :
>
>/usr/bin/cmake -P
>CMakeFiles/cmTryCompileExec.dir/cmake_clean_target.cmake /usr/local/intel/bin/ifort
>-KPIC  -lm  -O2
>"CMakeFiles/cmTryCompileExec.dir/testFortranCompiler.o"   -o
>cmTryCompileExec -i_dynamic IPO link: can not find "(" ifort: error:
>problem during multi-file optimization compilation (code 1) gmake[1]:
>*** [cmTryCompileExec] Error 1 gmake[1]: Leaving directory
>`/home/vv/RPM/BUILD/plplot-5.7.3/alt/CMakeFiles/CMakeTmp' gmake: ***
>[cmTryCompileExec/fast] Error 2
>
>  
>
Hello Valery,

I don't remember having tried the Intel Fortran compiler via CMake, but 
this looks
like a problem in the CMake support for that compiler. Well, it may not 
be of any
importance to you, but there may be others who do depend on it ;).

Thanks for reporting this one - I will forward this to the CMake list.

Regards,

Arjen

Re: [Plplot-general] cmake & disable fortran

[Valery P. <pi...@is...>]

Hi!
> >I have intel fortran 9 installed in  /usr/local/intel.
> >I heard that libtool does not support ifort. Also, there is linking
> >problem  like :
> >
> >/usr/bin/cmake -P
> >CMakeFiles/cmTryCompileExec.dir/cmake_clean_target.cmake /usr/local/intel/bin/ifort
> >-KPIC  -lm  -O2
> >"CMakeFiles/cmTryCompileExec.dir/testFortranCompiler.o"   -o
> >cmTryCompileExec -i_dynamic IPO link: can not find "(" ifort: error:
> >problem during multi-file optimization compilation (code 1) gmake[1]:
> >*** [cmTryCompileExec] Error 1 gmake[1]: Leaving directory
> >`/home/vv/RPM/BUILD/plplot-5.7.3/alt/CMakeFiles/CMakeTmp' gmake: ***
> >[cmTryCompileExec/fast] Error 2
> >
> I don't remember having tried the Intel Fortran compiler via CMake,
> but this looks
> like a problem in the CMake support for that compiler.
This is a feature (or bug) of ifort. It appears for multi-file
optimization. There is a disscussion about somewhere in the net.I don't
remeber where I found it. The similar bug appeared when I tried to
compile "windows_to_the_stars" with ifort. I don't know how to switch it
off for cmake. 

> Well, it may not be of any
> importance to you, but there may be others who do depend on it ;).
Potentially it is important for me too.

best regards
V

Re: [Plplot-general] cmake & disable fortran

[Alan W. I. <ir...@be...>]

On 2007-05-29 16:39+0900 Valery Pipin wrote:

> Hi!
>>> I have intel fortran 9 installed in  /usr/local/intel.
>>> I heard that libtool does not support ifort. Also, there is linking
>>> problem  like :
>>>
>>> /usr/bin/cmake -P
>>> CMakeFiles/cmTryCompileExec.dir/cmake_clean_target.cmake /usr/local/intel/bin/ifort
>>> -KPIC  -lm  -O2
>>> "CMakeFiles/cmTryCompileExec.dir/testFortranCompiler.o"   -o
>>> cmTryCompileExec -i_dynamic IPO link: can not find "(" ifort: error:
>>> problem during multi-file optimization compilation (code 1) gmake[1]:
>>> *** [cmTryCompileExec] Error 1 gmake[1]: Leaving directory
>>> `/home/vv/RPM/BUILD/plplot-5.7.3/alt/CMakeFiles/CMakeTmp' gmake: ***
>>> [cmTryCompileExec/fast] Error 2
>>>
>> I don't remember having tried the Intel Fortran compiler via CMake,
>> but this looks
>> like a problem in the CMake support for that compiler.
> This is a feature (or bug) of ifort. It appears for multi-file
> optimization. There is a disscussion about somewhere in the net.I don't
> remeber where I found it. The similar bug appeared when I tried to
> compile "windows_to_the_stars" with ifort. I don't know how to switch it
> off for cmake.

CMake usually has certain default compilers it uses and flags for those
compilers depending on your system.  However, sometimes that default system
does not work or does not give you exactly what you want.  In that case,
to control all aspects of your compiler under CMake (whether C, C++, or
Fortran) set compiler environment variables _before_ running cmake in an
_empty_ build directory.

We already give some simple examples at http://www.miscdebris.net/plplot_wiki/.

export CC="gcc -O2"
export CXX="g++ -O2"
export FC="g77 -O2"

but you have complete control; you can change to any appropriate compiler
name for each of these variables, and any command-line option appropriate to
the compiler in question.

For example, Valery, suppose you want to use the ifort compiler with
optimization turned off.  Specify that as follows:

export FC="ifort -O0"

BTW, according to my quick Google search, the -O0 option turns off all
optimization including multi-file optimization so you will probably want to
use a much more specific combination of ifort compiler flags to turn off
multi-file optimization while, for example, specifying normal -O2
optimization for everything else.

Valery, I understand you don't have much need for fortran, but if you do
pursue this out of curiosity, please let us know the results of specifying
the FC environment variable for the case of your particular ifort compiler.

Good luck.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-general] cmake & disable fortran

[Valery P. <pi...@is...>]

Hello Allan!

On Tue, 29 May 2007 07:35:19 -0700 (PDT)
Alan W. Irwin wrote:

> We already give some simple examples at
> http://www.miscdebris.net/plplot_wiki/.
> 
> export CC="gcc -O2"
> export CXX="g++ -O2"
> export FC="g77 -O2"
> 
> but you have complete control; you can change to any appropriate
> compiler name for each of these variables, and any command-line
> option appropriate to the compiler in question.
> 
> For example, Valery, suppose you want to use the ifort compiler with
> optimization turned off.  Specify that as follows:
> 
> export FC="ifort -O0"
> 
This makes everything! I have 
"......
+ rm /home/vv/tmp/plplot-buildroot/usr/share/man/man3/MAN-MANIFEST
+ ctest
Start processing tests
Test project /home/vv/RPM/BUILD/plplot-5.7.3/alt
  1/  9 Testing examples_c                       Passed
  2/  9 Testing examples_cxx                     Passed
  3/  9 Testing examples_f77                     Passed
  4/  9 Testing examples_f95                     Passed
  5/  9 Testing examples_python                  Passed
  6/  9 Testing examples_psttfc                  Passed
  7/  9 Testing examples_pstex                   Passed
  8/  9 Testing examples_svg                     Passed
  9/  9 Testing examples_plmeta                  Passed

100% tests passed, 0 tests failed out of 9
+ /usr/lib/rpm/brp-alt
....
"
The rpm was not built because of
"verify-elf: ERROR: ./usr/lib/libcsirocsa.so.0.0.1: RPATH ends with
":": /usr/local/intel/lib:/usr/local/lib/:/usr/local/intel/mkl/lib/32: "
I have rpath=relaxed in .rpmmacros, and it does not help.

kind regards
Valery

Re: [Plplot-general] cmake & disable fortran

[Alan W. I. <ir...@be...>]

On 2007-05-29 16:18+0900 Valery Pipin wrote:

> Start processing tests
> Test project /home/vv/RPM/BUILD/plplot-5.7.3/alt
>  1/  9 Testing examples_c                       Passed
>  2/  9 Testing examples_cxx                     Passed
>  3/  9 Testing examples_f77                     Passed
>  4/  9 Testing examples_f95                     Passed
>  5/  9 Testing examples_python                  Passed
>  6/  9 Testing examples_psttfc                  Passed
>  7/  9 Testing examples_pstex                   Passed
>  8/  9 Testing examples_svg                     Passed
>  9/  9 Testing examples_plmeta                  Passed
>
> 100% tests passed, 0 tests failed out of 9

Excellent result.  It means that it was possible to run every example with
-dev psc, and the C examples with the psttfc, pstex, svg, and plmeta devices
on your system.

> + /usr/lib/rpm/brp-alt
> ....
> "
> The rpm was not built because of
> "verify-elf: ERROR: ./usr/lib/libcsirocsa.so.0.0.1: RPATH ends with
> ":": /usr/local/intel/lib:/usr/local/lib/:/usr/local/intel/mkl/lib/32: "
> I have rpath=relaxed in .rpmmacros, and it does not help.

The trailing ":" should not happen.  This seems like a build-system bug to
me for the conditions on your platform, but I need more information
(e.g., complete cmake.out, complete make.out, complete make_install.out) to
find out what the problem is.

A more general rpath issue is some packaging tools (at least Debian and
probably your platform as well if you didn't use rpath=relaxed) don't like
it to be used.  (It makes no sense when you are installing in standard
locations.) For that reason, we have the USE_RPATH option which by default
is ON (a great convenience when you have non-standard install locations) but
which can be set to OFF via -DUSE_RPATH=OFF.  However, USE_RPATH currently
only works for the installed examples, and the installed libraries have
rpath set regardless.  This general rpath issue (USE_RPATH is currently
ignored) for our libraries should be fixed as well.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-general] cmake & disable fortran

[Valery P. <pi...@is...>]

Dear Alan,

Thanks for the help!
> > + /usr/lib/rpm/brp-alt
> > ....
> > "
> > The rpm was not built because of
> > "verify-elf: ERROR: ./usr/lib/libcsirocsa.so.0.0.1: RPATH ends with
> > ":": /usr/local/intel/lib:/usr/local/lib/:/usr/local/intel/mkl/lib/32:
> > " I have rpath=relaxed in .rpmmacros, and it does not help.
> 
> The trailing ":" should not happen.  
It comes from my .bash_profile
where I put
LD_LIBRIRARY_PATH= /usr/local/intel/lib:$LD_LIBRIRARY_PATH
export LD_LIBRIRARY_PATH.

I thought it is legal. This time I changed to construction proposed in
the manual : 
if [ -z "$LD_LIBRARY_PATH" ]
then 
 LD_LIBRARY_PATH="/usr/local/intel/lib";
else
 LD_LIBRARY_PATH="/usr/local/intel/lib:$LD_LIBRARY_PATH:$MKLPATH";
fi
export LD_LIBRARY_PATH;

Having this and with option -DUSE_RPATH=OFF the building was smoothly
ended with no errors. There were still some warning like

verify-elf: WARNING: ./usr/lib/libplplotf77d.so.9.1.1: undefined
symbol: for_iargc verify-elf:
WARNING: ./usr/lib/libplplotf77d.so.9.1.1: undefined symbol:
for_write_seq_fmt verify-elf:
WARNING: ./usr/lib/libplplotf77d.so.9.1.1: undefined symbol: for_getarg
verify-elf: WARNING: ./usr/lib/libplplotf77d.so.9.1.1: undefined
symbol: for_cpystr verify-elf:
WARNING: ./usr/lib/libplplotf95d.so.9.1.1: TEXTREL entry found: 0x0

kind regards
Valery

Re: [Plplot-general] cmake & disable fortran

[Alan W. I. <ir...@be...>]

On 2007-05-30 17:21+0900 Valery Pipin wrote:

> Having this and with option -DUSE_RPATH=OFF the building was smoothly
> ended with no errors.

To test that all is well with the trailing colon business for rpath,
-DUSE_RPATH=ON (or no option since this is the default) should be used. (I
think that will probably be OK since you now have a correct LD_LIBRARY_PATH,
but it would be nice to know for sure.) Once that is tested, then for your
ultimate build, you will probably want -DUSE_RPATH=OFF so I am glad to hear
that works for you.

> There were still some warning like
>
> verify-elf: WARNING: ./usr/lib/libplplotf77d.so.9.1.1: undefined
> symbol: for_iargc verify-elf:
> WARNING: ./usr/lib/libplplotf77d.so.9.1.1: undefined symbol:
> for_write_seq_fmt verify-elf:
> WARNING: ./usr/lib/libplplotf77d.so.9.1.1: undefined symbol: for_getarg
> verify-elf: WARNING: ./usr/lib/libplplotf77d.so.9.1.1: undefined
> symbol: for_cpystr verify-elf:
> WARNING: ./usr/lib/libplplotf95d.so.9.1.1: TEXTREL entry found: 0x0

IARGC and GETARG are important functions that should be defined (e.g., to
obtain Fortran command-line parsing).  I just checked the Intel Fortran
Libraries Reference
(http://www.intel.com/software/products/compilers/techtopics/for_lib.htm),
and it appears at least those symbols will be defined if you use the -Vaxlib
option (as well as -O0).

BTW, Valery, I know you are keen on python/Numeric, and I am as well, but I
have looked recently at Fortran 95, and it is an excellent high-level
language for number crunching with many (or all?) of the Fortran 77
irritations taken care of.  Also, there is a large number of different
Fortran 95 compilers available including the Intel one and also gfortran
which is part of the Gnu Compiler Collection and free in both senses.
Anyhow, I see a pretty bright future for Fortran 95 so I am glad we have an
f95 interface to PLplot (thanks to Arjen's efforts last year).  Our current
documentation for that interface is just a placeholder (where we took the
fortran 77 documentation and substituted 95 for 77 (!)) so those with an
interest in our f95 interface are urged to look at our f95 examples and not
pay too much attention to our formal f95 "documentation" until we update it.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-general] cmake & disable fortran

[Valery P. <pi...@is...>]

Alan W. Irwin wrote:
> On 2007-05-30 17:21+0900 Valery Pipin wrote:
> > Having this and with option -DUSE_RPATH=OFF the building was smoothly
> > ended with no errors.
>
> To test that all is well with the trailing colon business for rpath,
> -DUSE_RPATH=ON (or no option since this is the default) should be used. (I
> think that will probably be OK since you now have a correct
> LD_LIBRARY_PATH, but it would be nice to know for sure.)
Yes this works now with  RPATH=ON as well.
> > There were still some warning like
> >
> > verify-elf: WARNING: ./usr/lib/libplplotf77d.so.9.1.1: undefined
> > symbol: for_iargc verify-elf:
> > WARNING: ./usr/lib/libplplotf77d.so.9.1.1: undefined symbol:
> > for_write_seq_fmt verify-elf:
> > WARNING: ./usr/lib/libplplotf77d.so.9.1.1: undefined symbol: for_getarg
> > verify-elf: WARNING: ./usr/lib/libplplotf77d.so.9.1.1: undefined
> > symbol: for_cpystr verify-elf:
> > WARNING: ./usr/lib/libplplotf95d.so.9.1.1: TEXTREL entry found: 0x0
>
> IARGC and GETARG are important functions that should be defined (e.g., to
> obtain Fortran command-line parsing).  I just checked the Intel Fortran
> Libraries Reference
> (http://www.intel.com/software/products/compilers/techtopics/for_lib.htm),
> and it appears at least those symbols will be defined if you use the
> -Vaxlib option (as well as -O0).
I have added -Vaxlib to FCFLAGS and it does not help. It may be because the 
intel libraries are not in standard path. As rpm complaine is

verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: not found: libimf.so
verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: not found: 
libifport.so.5
verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: not found: 
libifcore.so.5
verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: undefined symbol: 
for_iargc

Both libimf and libifort are in /usr/local/intel/lib that is in 
LD_LIBRARY_PATH.
> BTW, Valery, I know you are keen on python/Numeric, and I am as well, but I
> have looked recently at Fortran 95, and it is an excellent high-level
> language for number crunching with many (or all?) of the Fortran 77
> irritations taken care of.  Also, there is a large number of different
> Fortran 95 compilers available including the Intel one and also gfortran
> which is part of the Gnu Compiler Collection and free in both senses.
> Anyhow, I see a pretty bright future for Fortran 95 so I am glad we have an
> f95 interface to PLplot (thanks to Arjen's efforts last year).  Our current
> documentation for that interface is just a placeholder (where we took the
> fortran 77 documentation and substituted 95 for 77 (!)) so those with an
> interest in our f95 interface are urged to look at our f95 examples and not
> pay too much attention to our formal f95 "documentation" until we update
> it.
Thanks for pointing this! Though, I prefer to draw results after computations.
Also, there is a principal possibility to tight plplot and maxima via 
cl-plplot which was  developed by Hazen  Babcock. Though, the maxima folks 
are keen of gnuplot.


kind regards,
Valery

Re: [Plplot-general] cmake & disable fortran

[Alan W. I. <ir...@be...>]

On 2007-05-31 13:32+0900 Valery Pipin wrote:

> I have added -Vaxlib to FCFLAGS and it does not help.

Hi Valery:

I am sure you knew this, but just to keep the record straight it should be
FC, i.e.,

export FC='ifort -O0 -Vaxlib'

not FCFLAGS.

> It may be because the
> intel libraries are not in standard path. As rpm complaine is
>
> verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: not found: libimf.so
> verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: not found:
> libifport.so.5
> verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: not found:
> libifcore.so.5
> verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: undefined symbol:
> for_iargc
>
> Both libimf and libifort are in /usr/local/intel/lib that is in
> LD_LIBRARY_PATH.

I think you are correct that rpm won't work for your setup since rpm expects
everything is installed in standard locations.

At this point, you may just want to turn off the f77 and f95 interfaces
since you are not primarily interested in those.  However, in case you have
some remaining interest or someone else runs into ifort problems there is
still the alternative of doing a simple PLplot build and install without
worrying about the special requirements of rpm.  That is, use the normal
cmake; make; and make install.  I think that procedure would work correctly
on any system where ifort is installed in a non-standard location. After
all, ifort knows how to link everything (i.e., it knows the locations of its
libraries), and at run-time, your LD_LIBRARY_PATH tells the run-time loader
where to find those same fortran libraries.

A comprehensive test of the fortran interface to PLplot built and installed
that way is provided if you try "make install", and then run the standard
installed examples test (e.g., run "make" and "./plplot-test.sh" in
$PREFIX/share/plplot5.7.3/examples, where $PREFIX is the PLplot installation
prefix you specified to the cmake command with the -DCMAKE_INSTALL_PREFIX
option.) You could also try "ldd -r $PREFIX/lib/libplplotf95d.so" to make
sure all libraries are found and those libraries define all required
symbols.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-general] cmake & disable fortran

[Valery P. <pi...@is...>]

Alan W. Irwin wrote:
> On 2007-05-31 13:32+0900 Valery Pipin wrote:
> > I have added -Vaxlib to FCFLAGS and it does not help.
>
> Hi Valery:
>
> I am sure you knew this, but just to keep the record straight it should be
> FC, i.e.,
>
> export FC='ifort -O0 -Vaxlib'
>
> not FCFLAGS.
I'm sorry the previous message was misleading. I have used  
export FC='ifort -O0 -Vaxlib' exactly. During the build I observed that cmake 
understood it.
>
> > It may be because the
> > intel libraries are not in standard path. As rpm complaine is
> >
> > verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: not found:
> > libimf.so verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: not
> > found: libifport.so.5
> > verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: not found:
> > libifcore.so.5
> > verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: undefined symbol:
> > for_iargc
> >
> > Both libimf and libifort are in /usr/local/intel/lib that is in
> > LD_LIBRARY_PATH.
>
> I think you are correct that rpm won't work for your setup since rpm
> expects everything is installed in standard locations.
>
> At this point, you may just want to turn off the f77 and f95 interfaces
> since you are not primarily interested in those.
I thought  this interface should work on my mashine unless  the warnings.  
> However, in case you have 
> some remaining interest or someone else runs into ifort problems there is
> still the alternative of doing a simple PLplot build and install without
> worrying about the special requirements of rpm.  That is, use the normal
> cmake; make; and make install.  I think that procedure would work correctly
> on any system where ifort is installed in a non-standard location. After
> all, ifort knows how to link everything (i.e., it knows the locations of
> its libraries), and at run-time, your LD_LIBRARY_PATH tells the run-time
> loader where to find those same fortran libraries.
>
> A comprehensive test of the fortran interface to PLplot built and installed
> that way is provided if you try "make install", and then run the standard
> installed examples test (e.g., run "make" and "./plplot-test.sh" in
> $PREFIX/share/plplot5.7.3/examples, where $PREFIX is the PLplot
> installation prefix you specified to the cmake command with the
> -DCMAKE_INSTALL_PREFIX option.) You could also try "ldd -r
> $PREFIX/lib/libplplotf95d.so" to make sure all libraries are found and
> those libraries define all required symbols.
This is worth to try. 

kind regards
Valery

Re: [Plplot-general] cmake & disable fortran

[Valery P. <pi...@is...>]

BTW, I have others warnings like
Verifying ELF objects in /home/vv/tmp/plplot-buildroot 
(arch=normal,fhs=normal,rpath=normal,stack=normal,textrel=relaxed,unresolved=relaxed)
verify-elf: WARNING: ./usr/lib/libcsirocsa.so.0.0.1: undefined symbol: sqrt
verify-elf: WARNING: ./usr/lib/libcsirocsa.so.0.0.1: undefined symbol: floor
verify-elf: WARNING: ./usr/lib/libcsirocsa.so.0.0.1: undefined symbol: ceil
verify-elf: WARNING: ./usr/lib/libcsirocsa.so.0.0.1: undefined symbol: hypot

Does it mean that libcsirocsa is compiled without -lm ?

> > I am sure you knew this, but just to keep the record straight it should
> > be FC, i.e.,
> >
> > export FC='ifort -O0 -Vaxlib'
> >
> > not FCFLAGS.
>
> I'm sorry the previous message was misleading. I have used
> export FC='ifort -O0 -Vaxlib' exactly. During the build I observed that
> cmake understood it.
>
> > > It may be because the
> > > intel libraries are not in standard path. As rpm complaine is
> > >
> > > verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: not found:
> > > libimf.so verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: not
> > > found: libifport.so.5
> > > verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: not found:
> > > libifcore.so.5
> > > verify-elf: WARNING: ./usr/lib/libplplotf95d.so.9.1.1: undefined
> > > symbol: for_iargc
> > >
> > > Both libimf and libifort are in /usr/local/intel/lib that is in
> > > LD_LIBRARY_PATH.
> >
> > I think you are correct that rpm won't work for your setup since rpm
> > expects everything is installed in standard locations.
> >
> > At this point, you may just want to turn off the f77 and f95 interfaces
> > since you are not primarily interested in those.
>
> I thought  this interface should work on my mashine unless  the warnings.
>
> > However, in case you have
> > some remaining interest or someone else runs into ifort problems there is
> > still the alternative of doing a simple PLplot build and install without
> > worrying about the special requirements of rpm.  That is, use the normal
> > cmake; make; and make install.  I think that procedure would work
> > correctly on any system where ifort is installed in a non-standard
> > location. After all, ifort knows how to link everything (i.e., it knows
> > the locations of its libraries), and at run-time, your LD_LIBRARY_PATH
> > tells the run-time loader where to find those same fortran libraries.
> >
> > A comprehensive test of the fortran interface to PLplot built and
> > installed that way is provided if you try "make install", and then run
> > the standard installed examples test (e.g., run "make" and
> > "./plplot-test.sh" in $PREFIX/share/plplot5.7.3/examples, where $PREFIX
> > is the PLplot
> > installation prefix you specified to the cmake command with the
> > -DCMAKE_INSTALL_PREFIX option.) You could also try "ldd -r
> > $PREFIX/lib/libplplotf95d.so" to make sure all libraries are found and
> > those libraries define all required symbols.
>
> This is worth to try.
>
> kind regards
> Valery
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by DB2 Express
> Download DB2 Express C - the FREE version of DB2 express and take
> control of your XML. No limits. Just data. Click to get it now.
> http://sourceforge.net/powerbar/db2/
> _______________________________________________
> Plplot-general mailing list
> Plp...@li...
> https://lists.sourceforge.net/lists/listinfo/plplot-general

Re: [Plplot-general] cmake & disable fortran

[Alan W. I. <ir...@be...>]

On 2007-05-31 17:32+0900 Valery Pipin wrote:

>
> BTW, I have others warnings like
> Verifying ELF objects in /home/vv/tmp/plplot-buildroot
> (arch=normal,fhs=normal,rpath=normal,stack=normal,textrel=relaxed,unresolved=relaxed)
> verify-elf: WARNING: ./usr/lib/libcsirocsa.so.0.0.1: undefined symbol: sqrt
> verify-elf: WARNING: ./usr/lib/libcsirocsa.so.0.0.1: undefined symbol: floor
> verify-elf: WARNING: ./usr/lib/libcsirocsa.so.0.0.1: undefined symbol: ceil
> verify-elf: WARNING: ./usr/lib/libcsirocsa.so.0.0.1: undefined symbol: hypot
>
> Does it mean that libcsirocsa is compiled without -lm ?

Thanks for pointing out this build bug which I have just verified for myself
using 'ldd -r libcsirocsa.so'.

Now fixed in svn (revision 7698).

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

Re: [Plplot-general] cmake & disable fortran

[Arjen M. <arj...@wl...>]

Alan W. Irwin wrote:

>On 2007-05-31 17:32+0900 Valery Pipin wrote:
>
>  
>
>>BTW, I have others warnings like
>>Verifying ELF objects in /home/vv/tmp/plplot-buildroot
>>(arch=normal,fhs=normal,rpath=normal,stack=normal,textrel=relaxed,unresolved=relaxed)
>>verify-elf: WARNING: ./usr/lib/libcsirocsa.so.0.0.1: undefined symbol: sqrt
>>verify-elf: WARNING: ./usr/lib/libcsirocsa.so.0.0.1: undefined symbol: floor
>>verify-elf: WARNING: ./usr/lib/libcsirocsa.so.0.0.1: undefined symbol: ceil
>>verify-elf: WARNING: ./usr/lib/libcsirocsa.so.0.0.1: undefined symbol: hypot
>>
>>Does it mean that libcsirocsa is compiled without -lm ?
>>    
>>
>
>Thanks for pointing out this build bug which I have just verified for myself
>using 'ldd -r libcsirocsa.so'.
>
>Now fixed in svn (revision 7698).
>
>  
>
To answer your original question: C libraries that use the math 
functions require the math library (-lm)
to be explicitly specified when linking. At least that is the 
traditional way on UNIX and Linux (no idea
why this should be so, but it is).

Regards,

Arjen

Re: [Plplot-general] cmake & disable fortran

[Alan W. I. <ir...@be...>]

On 2007-05-31 07:40-0700 Alan W. Irwin wrote:

> Thanks for pointing out this build bug which I have just verified for myself
> using 'ldd -r libcsirocsa.so'.
>
> Now fixed in svn (revision 7698).

That should read 7699.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state implementation
for stellar interiors (freeeos.sf.net); PLplot scientific plotting software
package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of
Linux Links project (loll.sf.net); and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________