Re: Re2: [Vxl-users] Cygwin / MinGW build problems

[Amitha Perera <perera@cs...>]

I think this is not as easy with CMakeSetup. My guess: the problem is
that when CMakeSetup runs, it runs configuration checks with the
"current" compiler. As you start, it won't have the "-mno-cygwin"
flag. By the time you get to enter it into the flags, the
configuration part is already over, and the system has figured out
what the Cygwin g++ has, and not what the MingW g++ has.

When you run cmake directly, the compiler flags are already set by
the time the configuration checks run, and all is well.

Perhaps you should add a feature request to CMake to allow the
compiler to be changed before the project's configuration checks are
run. (Passing "-mno-cygwin" is effectively changing the compiler.)

Amitha.

Re: Re2: [Vxl-users] Cygwin / MinGW build problems

[Stefan v. Wachter <svwa-dev@mn...>]

I confirm that your build command with cmake (command line tool) works 
in my environment perfectly.

Seems like the CMakeSetup.exe (the version with GUI) 2.0 Patch 6 (Patch 
5 has the same behaviour) makes the problems. After configuring with 
this tool and the same config values as above I have in CMakeCache.txt 
these lines, which are wrong:
VXL_HAS_PTHREAD_H:INTERNAL=1
VXL_STDLIB_HAS_SRAND48:INTERNAL=1
VXL_STDLIB_HAS_DRAND48:INTERNAL=1

It is strange that in the build directory of the good build (with 
cmake.exe not CMakeSetup.exe) the file CMakeOutput.log  has:
" Determining if the include file pthread.h exists passed with the 
following output:
Building object file CheckIncludeFile.o...
c++.exe -o CheckIncludeFile.o             -c 
G:/src/vxl/bin_cygwin_test/CMakeTmp/CheckIncludeFile.cxx
Building executable 
G:/src/vxl/bin_cygwin_test/CMakeTmp/cmTryCompileExec.exe...
c++.exe                CheckIncludeFile.o    -o 
G:/src/vxl/bin_cygwin_test/CMakeTmp/cmTryCompileExec.exe  "

This means it found a pthread.h file!

Could you please test how CMakeSetup.exe works in your environment?

Stefan

RE: Re2: [Vxl-users] Cygwin / MinGW build problems

[Wheeler, Frederick W \(Research\) <wheeler@cr...>]

I am able to configure and build vxl using MinGW from cygwin using these =
exact commands ...

mkdir -p $TOP/vxl_bld_min
rm -f $TOP/vxl_bld_min/CMakeCache.txt
cd $TOP/vxl_bld_min
# use the windows version of cmake, but make Unix Makefiles
"/cygdrive/c/Program Files/CMake20/bin/cmake" \
  -G"Unix Makefiles" \
  -DCMAKE_INSTALL_PREFIX:PATH=3D"`cygpath -m $TOP/vxl_usr_min`" \
  -DCMAKE_C_COMPILER:FILEPATH=3D"`cygpath -m /usr/bin/gcc`" \
  -DCMAKE_CXX_COMPILER:FILEPATH=3D"`cygpath -m /usr/bin/g++`" \
  -DCMAKE_C_LINK_SHARED:FILEPATH=3D"`cygpath -m /usr/bin/gcc`" \
  -DCMAKE_CXX_LINK_SHARED:FILEPATH=3D"`cygpath -m /usr/bin/g++`" \
  -DCMAKE_C_FLAGS:STRING=3D"-Wall -g -O0 -mno-cygwin" \
  -DCMAKE_CXX_FLAGS:STRING=3D"-Wall -g -O0 -mno-cygwin" \
  -DX11_X11_INCLUDE_PATH:PATH=3DIGNORE \
  -DVXL_FORCE_V3P_ZLIB:BOOL=3DYES \
  -DVXL_FORCE_V3P_PNG:BOOL=3DYES \
  -DVXL_FORCE_V3P_JPEG:BOOL=3DYES \
  -DVXL_FORCE_V3P_MPEG2:BOOL=3DYES \
  -DVXL_FORCE_V3P_TIFF:BOOL=3DYES \
  -DVXL_FORCE_V3P_DCMTK:BOOL=3DYES \
  -DVXL_BUILD_DCMTK:BOOL=3DYES \
  -DBUILD_VGUI:BOOL=3DYES \
  -DCMAKE_VERBOSE_MAKEFILE:BOOL=3DOFF \
  "`cygpath -m $TOP/vxl_src`"
cd $TOP/vxl_bld_min
make

This is a complete new build from scratch.  TOP is an envronment =
varaiable that holds a directory name.

I just started this build.  It is up to vgui with no problems.  Where in =
the build process do you have a problem?

If the above does not work I suggest examining 'printenv' carefully. =20

What follows may be irrelevant.

I noticed the following in my dashboard build setup script.  This is in =
a section for a cmake dashboard build using mingw that is no longer =
used.

  # need to set CC, CXX, CFLAGS, CXXFLAGS, LDFLAGS in environment
  # so gcc with -mno-cygwin gets used for the cmake tests
  ENVIRON_BEFORE=3D"\
  CC=3D'`cygpath -m /usr/bin/gcc`';
  CXX=3D'`cygpath -m /usr/bin/g++`';
  CFLAGS=3D'-mno-cygwin';
  CXXFLAGS=3D'-mno-cygwin';
  LDFLAGS=3D'-mno-cygwin';"

ENVIRON_BEFORE gets evaled later in the script, before cmake is run.  =
Since these are set, but not exported this would have no effect.  Hmmm.  =
Plus, this portion of the script is no longer used.  So, I'm not sure =
why it is there, but I want to mention it since it is a clue that at =
some point I thought these variables needed to be set like this.

When I did the config/build shown above, none of the env vars were set.  =
(Make sure they are unset, not set to ''.)

Fred



> -----Original Message-----
> From: vxl-users-admin@...
> [mailto:vxl-users-admin@...]On Behalf Of Stefan v.
> Wachter
> Sent: Friday, October 21, 2005 2:30 AM
> To: vxl-users@...
> Subject: Re2: [Vxl-users] Cygwin / mingw build problems
>=20
>=20
> Thanks for your answer! It turned out that my problems are cmake=20
> related: Yesterday I used the cmake.exe from cygwin. This version=20
> assumes that an X11 environment exists. Today I switched to=20
> cmakeSetup.exe from the cmake Windows binary distribution.=20
> This is a lot=20
> better, but still problems exist: Checking for the existance=20
> of include=20
> files is done with this command:
> Building object file CheckIncludeFile.o...
> c++.exe -o CheckIncludeFile.o             -c=20
> G:/src/vxl/bin_cygwin_neu/CMakeTmp/CheckIncludeFile.cxx
>  >> but it should be:
>=20
> c++.exe -mno-cygwin -o CheckIncludeFile.o             -c=20
> G:/src/vxl/bin_cygwin_neu/CMakeTmp/CheckIncludeFile.cxx
> Otherwise the include dir /usr/include and not=20
> /usr/include/mingw will=20
> be used.
> To avoid this problem cmake should use the CXX compiler command with=20
> CXX_FLAGS appended.
>=20
> I had to manually edit the CMakeCache.txt:
> VXL_HAS_PTHREAD_H:INTERNAL=3D1 changed to =
VXL_HAS_PTHREAD_H:INTERNAL=3D
> The same with srand48 and drand48
>=20
> After these manual modifications it compiled and linked!
> Did someone perform the same modifications on your dashboard?
>=20
> Stefan
>=20
> Wheeler, Frederick W (Research) schrieb:
>=20
> >Stefan,
> >
> >I have a Cygwin/MinGW build on the VXL dashboard.  It has=20
> been working for a couple of years, through several cygwin=20
> tool upgrades.  Take a look at the build notes (the icon next=20
> to the Build Name on the dashboard) to see the CMakeCache=20
> settings I used and the various tool versions.  It's=20
> basically an up-to-date Cygwin installation.  You wrote=20
> "CXX_FLAGS and CC_FLAGS".  Do you mean "CXX_FLAGS and=20
> C_FLAGS" -- with "C" instead of "CC"?  Maybe your config got=20
> messed up because .c files were not compiled with=20
> -mno-cygwin?  For CMake I set
> >
> >CMAKE_C_FLAGS:STRING=3D-mno-cygwin -Wall -O2 -g
> >CMAKE_CXX_FLAGS:STRING=3D-mno-cygwin -Wall -O2 -g
> >
> >I have not compiled VXL in the MinGW environment, just in=20
> Cygwin with -mno-cygwin.
> >
> >Fred
> >
> >The build notes ...
> >
> >http://www.cs.rpi.edu/research/groups/vxl/Testing/Sites/crd.g
> e.com/MinGW-3.8_gcc-3.4.4_-Wall_-O2_static/20051019-0300-Night
> ly/Notes.html
> >
> > =20
> >
> >>-----Original Message-----
> >>From: vxl-users-admin@...
> >>[mailto:vxl-users-admin@...]On Behalf Of Stefan v.
> >>Wachter
> >>Sent: Wednesday, October 19, 2005 2:25 PM
> >>To: vxl-users@...
> >>Cc: Stefan v. Wachter
> >>Subject: [Vxl-users] Cygwin / mingw build problems
> >>
> >>
> >>What is the recommended procedure to compile vxl (1.3.0) on Windows=20
> >>using the cygwin (without need for the cygwin1.dll) or mingw tools?
> >>I started with cygwin and the -mno-cygwin flag in CXX_FLAGS and=20
> >>CC_FLAGS. This brought me to some errors e.g.:
> >>/cygdrive/g/src/vxl/vxl-1.3.0/core/vnl/vnl_sample.cxx: In=20
> >>function `void=20
> >>vnl_sample_reseed()':
> >>/cygdrive/g/src/vxl/vxl-1.3.0/core/vnl/vnl_sample.cxx:28: error:=20
> >>`srand48' undeclared (first use this function)
> >>
> >>I think this is because there are different stdlib.h files in=20
> >>/usr/include and /usr/include/mingw . It looks like cmake=20
> doesn't use=20
> >>the /usr/include/mingw/stdlib.h
> >>The were other problems with pthread.h not found.
> >>
> >>Then I switched to a pure mingw environment. But there I=20
> >>couldn't build=20
> >>cmake. Using the cygwin ccmake brought an error with=20
> >>strstream.h not found.
> >>
> >>Please tell me what configurations are working for you.
> >>
> >>Thanks a lot,
> >>Stefan v. Wachter
> >>ALPHA-OMEGA digital GmbH
> >>Munich
> >>
> >>
> >>-------------------------------------------------------
> >>This SF.Net email is sponsored by:
> >>Power Architecture Resource Center: Free content, downloads,=20
> >>discussions,
> >>and more. http://solutions.newsforge.com/ibmarch.tmpl
> >>_______________________________________________
> >>Vxl-users mailing list
> >>Vxl-users@...
> >>https://lists.sourceforge.net/lists/listinfo/vxl-users
> >>
> >>   =20
> >>
> >
> >
> > =20
> >
>=20
>=20
>=20
> -------------------------------------------------------
> This SF.Net email is sponsored by:
> Power Architecture Resource Center: Free content, downloads,=20
> discussions,
> and more. http://solutions.newsforge.com/ibmarch.tmpl
> _______________________________________________
> Vxl-users mailing list
> Vxl-users@...
> https://lists.sourceforge.net/lists/listinfo/vxl-users
>=20

Re: [Vxl-users] Re: how to install vxl to linux

[Nicolas Burrus <nicolas.burrus@en...>]

On Friday 21 October 2005 08:53, Peter Vanroose wrote:
> > how do I install the library and header file to my system? By
> > running a make install, root account, nothing is installed to my
> > prefix /usr/local.
>
> I never tried "make install", so I cannot give details on that.
> But essentially, no install is needed: normally, all libraries (.a
> and/or .so) are put in $VXLINSTALL/lib, and either moving or symlinking
> them to /usr/local/lib, or otherwise changing your LD_LIBRARY_PATH to
> point to that lib dir, should make your executables find them and run.

An install target has been added recently in the CVS version. In Vxl 1.3, make 
install does nothing. The CVS version is quite stable though.

An install target avoids:
1) Moving libraries by hand
2) Moving includes by hand (you basically don't need every file of Vxl source 
distribution)
3) Keeping every built files (only final libraries and generated includes will 
be copied)

Then you can do :
$ mkdir /tmp/build && cd /tmp/build
$ ccmake -DCMAKE_INSTALL_PREFIX=/usr/local $vxl_source_dir
$ make
# make install
$ rm -rf /tmp/build

And to use it in a simple program :
$ cd /tmp/example
$ cat CMakeLists.txt
FIND_PACKAGE(VXL)

IF(VXL_FOUND)
  INCLUDE(${VXL_CMAKE_DIR}/UseVXL.cmake)
ENDIF(VXL_FOUND)

LINK_LIBRARIES( vil vcl )
ADD_EXECUTABLE( vil_example example.cxx )
$ ccmake -DVXL_DIR=/usr/local/share/vxl/cmake .
$ make
$ ./vil_example

-- 
Nicolas Burrus

[Vxl-users] Re: how to install vxl to linux

[Peter Vanroose <peter_vanroose@ya...>]

> how do I install the library and header file to my system? By 
> running a make install, root account, nothing is installed to my
> prefix /usr/local.

I never tried "make install", so I cannot give details on that.
But essentially, no install is needed: normally, all libraries (.a
and/or .so) are put in $VXLINSTALL/lib, and either moving or symlinking
them to /usr/local/lib, or otherwise changing your LD_LIBRARY_PATH to
point to that lib dir, should make your executables find them and run.

--	Peter.

Re2: [Vxl-users] Cygwin / mingw build problems

[Stefan v. Wachter <svwa-dev@mn...>]

Thanks for your answer! It turned out that my problems are cmake 
related: Yesterday I used the cmake.exe from cygwin. This version 
assumes that an X11 environment exists. Today I switched to 
cmakeSetup.exe from the cmake Windows binary distribution. This is a lot 
better, but still problems exist: Checking for the existance of include 
files is done with this command:
Building object file CheckIncludeFile.o...
c++.exe -o CheckIncludeFile.o             -c 
G:/src/vxl/bin_cygwin_neu/CMakeTmp/CheckIncludeFile.cxx
 >> but it should be:

c++.exe -mno-cygwin -o CheckIncludeFile.o             -c 
G:/src/vxl/bin_cygwin_neu/CMakeTmp/CheckIncludeFile.cxx
Otherwise the include dir /usr/include and not /usr/include/mingw will 
be used.
To avoid this problem cmake should use the CXX compiler command with 
CXX_FLAGS appended.

I had to manually edit the CMakeCache.txt:
VXL_HAS_PTHREAD_H:INTERNAL=1 changed to VXL_HAS_PTHREAD_H:INTERNAL=
The same with srand48 and drand48

After these manual modifications it compiled and linked!
Did someone perform the same modifications on your dashboard?

Stefan

Wheeler, Frederick W (Research) schrieb:

>Stefan,
>
>I have a Cygwin/MinGW build on the VXL dashboard.  It has been working for a couple of years, through several cygwin tool upgrades.  Take a look at the build notes (the icon next to the Build Name on the dashboard) to see the CMakeCache settings I used and the various tool versions.  It's basically an up-to-date Cygwin installation.  You wrote "CXX_FLAGS and CC_FLAGS".  Do you mean "CXX_FLAGS and C_FLAGS" -- with "C" instead of "CC"?  Maybe your config got messed up because .c files were not compiled with -mno-cygwin?  For CMake I set
>
>CMAKE_C_FLAGS:STRING=-mno-cygwin -Wall -O2 -g
>CMAKE_CXX_FLAGS:STRING=-mno-cygwin -Wall -O2 -g
>
>I have not compiled VXL in the MinGW environment, just in Cygwin with -mno-cygwin.
>
>Fred
>
>The build notes ...
>
>http://www.cs.rpi.edu/research/groups/vxl/Testing/Sites/crd.ge.com/MinGW-3.8_gcc-3.4.4_-Wall_-O2_static/20051019-0300-Nightly/Notes.html
>
>  
>
>>-----Original Message-----
>>From: vxl-users-admin@...
>>[mailto:vxl-users-admin@...]On Behalf Of Stefan v.
>>Wachter
>>Sent: Wednesday, October 19, 2005 2:25 PM
>>To: vxl-users@...
>>Cc: Stefan v. Wachter
>>Subject: [Vxl-users] Cygwin / mingw build problems
>>
>>
>>What is the recommended procedure to compile vxl (1.3.0) on Windows 
>>using the cygwin (without need for the cygwin1.dll) or mingw tools?
>>I started with cygwin and the -mno-cygwin flag in CXX_FLAGS and 
>>CC_FLAGS. This brought me to some errors e.g.:
>>/cygdrive/g/src/vxl/vxl-1.3.0/core/vnl/vnl_sample.cxx: In 
>>function `void 
>>vnl_sample_reseed()':
>>/cygdrive/g/src/vxl/vxl-1.3.0/core/vnl/vnl_sample.cxx:28: error: 
>>`srand48' undeclared (first use this function)
>>
>>I think this is because there are different stdlib.h files in 
>>/usr/include and /usr/include/mingw . It looks like cmake doesn't use 
>>the /usr/include/mingw/stdlib.h
>>The were other problems with pthread.h not found.
>>
>>Then I switched to a pure mingw environment. But there I 
>>couldn't build 
>>cmake. Using the cygwin ccmake brought an error with 
>>strstream.h not found.
>>
>>Please tell me what configurations are working for you.
>>
>>Thanks a lot,
>>Stefan v. Wachter
>>ALPHA-OMEGA digital GmbH
>>Munich
>>
>>
>>-------------------------------------------------------
>>This SF.Net email is sponsored by:
>>Power Architecture Resource Center: Free content, downloads, 
>>discussions,
>>and more. http://solutions.newsforge.com/ibmarch.tmpl
>>_______________________________________________
>>Vxl-users mailing list
>>Vxl-users@...
>>https://lists.sourceforge.net/lists/listinfo/vxl-users
>>
>>    
>>
>
>
>  
>

[Vxl-users] how to install vxl to linux

[Zhaoxuan WU <mounterriver@ho...>]

I compiled the vxl 1.3.0 on Fedora Core 4 using gcc 4.0.1
My question is how do I install the library and header file to my system? By 
running a make install, root account, nothing is installed to my prefix 
/usr/local. Thanks

_________________________________________________________________
Don’t just search. Find. Check out the new MSN Search! 
http://search.msn.click-url.com/go/onm00200636ave/direct/01/