http://public.kitware.com/Bug/print_bug_page.php?bug_id=10077*
*
James
*
*
On Tue, Mar 29, 2011 at 11:56 AM, David Cole wrote:
> Hi all,
>
> Now that we have released CMake 2.8.4, *now* would be a great time to
> prioritize bug fixes for the next release of CMake.
>
> Replies requested. Read o
The recent thread on ctest and git submodules made me start considering
using submodules in a few projects. My question is how to integrate the two
build systems (both cmake based).
Right now I do a find_package() to resolve the dependent library. If I move
to a submodule, and both the superproj
Be sure that you have
INCLUDE_DIRECTORIES( [path to Internal.h and External.h] )
set.
James
On Tue, Nov 16, 2010 at 12:24 PM, David Doria wrote:
> Ryan,
>
> The issue is that these headers are only used internally to this
> collection of files, so I don't want to have to include them in my
>
On Wed, Sep 1, 2010 at 8:06 AM, Michael Jackson wrote:
> This probably isn't the best solution but until CMake fixes the bug and
> includes the updated FindHDF5.cmake file in the next or future release this
> is what you have to do:
>
> Copy the FindHDF5.cmake file from the current CMake distribu
On Wed, Sep 1, 2010 at 8:06 AM, Michael Jackson wrote:
>
> Now, edit the FindHDF5.cmake file (the one that you copied into your
> project) to remove the offending lines/fix the bug.
>
Well - that is the real trick. I am not sure where the offending lines
are...
It may be that the path of least
FYI, I just replicated this on Mac OSX 10.6.4 with HDF5 1.8.2 and CMake
2.8.1. If I do not have the "hl" libs installed things break with the same
message on the configure step.
James
On Wed, Sep 1, 2010 at 7:51 AM, James C. Sutherland <
james.sutherl...@utah.edu> wrote:
>
...@bluequartz.net
> BlueQuartz Software Dayton, Ohio
>
>
> On Sep 1, 2010, at 9:17 AM, James C. Sutherland wrote:
>
> I was aware of the CMake build for the newer HDF5 releases.
>>
>> However, I need this to work for users who are not able/willing to build
>> th
; REQUIRED NO_MODULE PATHS
> "$ENV{ProgramFiles}/HDF5")*
>
>
>
> [To fix the problem on the systems without re-installing, just remove
> references to hdf5_hl from the FindHDF5 stuff in your cmake installation
> module]
>
>
>
> JB
>
>
>
> *From:* cmake-bo
I typically have full HDF5 installations on my systems but am now going to a
system where the "hl" versions of the hdf5 libraries are not installed.
Although I don't need these, it appears that CMake is unwilling to let me
proceed without them. I get the following error:
CMake Error: The followi
Is the FindMatlab.cmake package maintained? It seems to fail miserably on
Mac.
Are there any patches out there?
Thanks for any tips...
James
___
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/o
I have a project that may need to use a fortran compiler depending on the
user configuration of the project.
The problem is that I don't want to force usage of fortran (in the project()
statement) since in some cases a user may not have a fortran compiler
installed and I want the configure to proc
On Mon, Jul 5, 2010 at 10:44 AM, John R. Cary wrote:
> On 7/5/2010 9:37 AM, James C. Sutherland wrote:
>
>
>
> On Mon, Jul 5, 2010 at 9:34 AM, John Cary wrote:
>
>> I have simply:
>>
>>FIND_PACKAGE(BLAS)
>>FIND_PACKAGE(LAPACK)
>>
>&
On Mon, Jul 5, 2010 at 9:34 AM, John Cary wrote:
> I have simply:
>
>FIND_PACKAGE(BLAS)
>FIND_PACKAGE(LAPACK)
>
> but cmake gives
>
> -- Looking for BLAS... - NOT found (Fortran not enabled)
> -- Looking for LAPACK... - NOT found (Fortran not enabled)
>
> Is this more complicated somehow?
On May 12, 2010, at 11:43 AM, James C. Sutherland wrote:
> I would like to execute a command like:
> getrates -n < myFileName
>
> I am trying to add this so it is processed after the target "getrates" is
> built
>
> set( infile ${CM
I would like to execute a command like:
getrates -n < myFileName
I am trying to add this so it is processed after the target "getrates" is built
set( infile ${CMAKE_CURRENT_BINARY_DIR}/myFileName )
add_custom_command( TARGET getrates
POST_BUILD
COMMAND "${CMAKE_CURRENT_BINARY_DIR}/get
>>
>> 2. The instructions on http://www.cmake.org/Wiki/CMake:MatlabMex don't seem
>> to work either.
>> I tried following the approach of setting MEX as the CC and CXX compiler,
>> but even on the trivial "hello
>> world" example posted on the wiki page, CMake fails during the configure
>> phas
I have been trying to get CMake to build Matlab MEX files without any success.
I have hit several problems.
1. It seems that FindMatlab.cmake is not very functional. I am on a Mac with
Matlab 2010a and this script cannot find my installation. It seems to be tuned
only for windows, and old ver
I am trying to use CMake to build a very simple CUDA program. I am seeing the
following errors when I try to build:
> /usr/include/c++/4.2.1/new(95): error: first parameter of allocation function
> must be of type "size_t"
> /usr/include/c++/4.2.1/new(96): error: first parameter of allocation f
>
> Why not just let the user decide where they'd like to get boost, and simplify
> your build system by just using find_package(Boost)?
>
> For example, this is all you need for a simple boost-using app...
>
> find_package(Boost 1.34.0 REQUIRED COMPONENTS filesystem system)
> include_directori
>
> include_directories(
> boost/src
> boost/build/lib/Exports.cmake
> )
>
I think that you want
include( boost/build/lib/Exports.cmake )
instead. Also, you will need to specify the full path to the boost directory.
__
On Mar 18, 2010, at 7:47 AM, Marcel Loose wrote:
> I ran into a problem with FindBoost where it fails to find the correct
> version of a component library. The cause is pretty clear to me.
This isn't an answer to your question, but have you tried the CMake version of
boost? I have found it much
On Dec 8, 2009, at 4:09 AM, Olivier Pierard wrote:
> Dear all,
>
> I'm a little bit confused with the use of findmpi.cmake. Does anyone
> has an example of its usage (for compilation, not yet for running) ?
>
> Since we have two mpi implementations (mpich and openmpi-ib), I set the
> compiler
Is there a way in CMake to determine the number of shared-memory cores on a
system?
___
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
Please keep messages on-topic and check th
It is probably worth considering
find_package( MPI )
This will not only set the MPI compilers but also defines the mpi executable,
etc.
James
On Nov 22, 2009, at 1:36 AM, Taesoo Kwon wrote:
> I am sorry that this should be an recurring question.
> But I wasn't able to find any workin
On Nov 19, 2009, at 8:24 AM, Michael Jackson wrote:
> So, there are a few of us quickly port the HDF5 1.8 code to CMake. I'm
> thinking that we should put in an HDF5Config.cmake file for other projects
> use. Simple question:
>
> What goes in one of those? Is there a tutorial somewhere? Where
On Nov 11, 2009, at 9:05 AM, Brad King wrote:
> James C. Sutherland wrote:
>> Thanks. To be clear, it only works if I use
>>
>> set( ExprLib_LIBRARIES @TPL_LIBRARIES@ )
>>
>> because otherwise I have no way of propagating the dependents of
>> Exp
Brad, thanks for the feedback. See comments below.
On Nov 10, 2009, at 11:30 AM, Brad King wrote:
> James C. Sutherland wrote:
>> The library I am building pulls in several other libraries (MPI, BLAS,
>> Boost, etc). I would like any down-stream apps that use my library to
&
On Nov 9, 2009, at 12:00 PM, Brad King wrote:
James C. Sutherland wrote:
However, I just discovered that this is a deprecated feature, and
that
install(EXPORT...)
should be used instead.
Good, it is much more modern and more powerful.
However, it is not clear to me how to incorporate
On Nov 7, 2009, at 11:59 AM, James C. Sutherland wrote:
I have been using
export_library_dependencies()
to incorporate library dependencies when generating configuration
files for use with find_package()
However, I just discovered that this is a deprecated feature, and that
I have been using
export_library_dependencies()
to incorporate library dependencies when generating configuration
files for use with find_package()
However, I just discovered that this is a deprecated feature, and that
install(EXPORT...)
should be used instead.
However, it is
On Oct 29, 2009, at 2:03 PM, Will Dicharry wrote:
James C. Sutherland wrote:
It appears that the variable
HDF5_FOUND
is not being set in the FindHDF5.cmake module.
Any way of getting this fixed prior to the 2.8.0 release?
It is set for my systems when it is found. The standard system
It appears that the variable
HDF5_FOUND
is not being set in the FindHDF5.cmake module.
Any way of getting this fixed prior to the 2.8.0 release?
___
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/
On Oct 26, 2009, at 8:13 AM, Moreland, Kenneth wrote:
Consider using the UseLATEX.cmake collection of macros. It supports
out-of-source builds (in fact, sort of enforces them) and handles
copying .sty files as well as all sorts of other sticky little
issues with building LaTeX files. You
On Oct 22, 2009, at 8:50 AM, Bill Hoffman wrote:
James C. Sutherland wrote:
Thank you, Dave and Bill. Your suggestions fixed the problem. It
seems that this is likely a bug that should be fixed, since the
behavior is different on Mac and Linux.
Can you add some message statements and
Thank you, Dave and Bill. Your suggestions fixed the problem. It
seems that this is likely a bug that should be fixed, since the
behavior is different on Mac and Linux.
On Oct 22, 2009, at 6:27 AM, Bill Hoffman wrote:
James C. Sutherland wrote:
I am using configure_file to copy files
I am using configure_file to copy files during the build process. I
have a nested directory structure, that can be represented as:
/rootdir
CMakeLists.txt
/test
CMakeLists.txt
When processing rootdir/test/CMakeLists.txt, I have the following:
#---
get_filename_
If I add a CMake file via add_subdirectory is there a way to obtain
its absolute or relative path?
___
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
Please keep messages on-
How do I set an environment variable so that it is visible globally in
CMake? Specifically, I want to have it visible when I run something
like:
#---
# set the environment variable value
set( ENV{myVariable} ... )
# run a test through a CMake script that requ
I have a fairly large project where the header files from various
subdirectories are combined into a single include directory when the
project is installed.
For example, header "a.h" may be located in:
/root/subproject1/
in the source tree, it is located in
/include/MyProject/
I have a situation where I have a few TPLs that are built via auto
tools and another project with CMake.
tpl1 -> lib1
tpl2 -> lib2
myproject -> libmyproj
I would like to combine the two TPLs into my project library. From
looking at previous postings on this list, it s
On Oct 2, 2009, at 8:20 AM,
wrote:
CMakeLists.txt
cmake_minimum_required(VERSION 2.6)
SET(TEST CACHE STRING "test variable")
INSTALL(SCRIPT ${CMAKE_CURRENT_SOURCE_DIR}/install.cmake)
install.cmake
EXECUTE_PROCESS(
COMMAND "/bin/echo" "--test=${TESTVAR} --prefix=$
{CMAKE_I
On Sep 22, 2009, at 6:16 AM, motes motes wrote:
In a CMakeList.txt file I have:
SET(ELASTIX_BUILD E:/local/lib/elastix_sources_v4.1/src/build)
...
LINK_DIRECTORIES(${ELASTIX_BUILD}/Components/Optimizers/
StandardGradientDescent)
I would to be able to specify the variable: ELASTIX_BUILD w
From FindBoost.cmake:
# These last three variables are available also as environment
variables:
#
# BOOST_ROOT or BOOSTROOT The preferred installation prefix for
searching for
#Boost. Set this if the module has
problems finding
#
On Sep 15, 2009, at 2:47 PM, Tyler Roscoe wrote:
On Tue, Sep 15, 2009 at 02:20:35PM -0600, James C. Sutherland wrote:
set( CMAKE_INSTALL_PREFIX
${CMAKE_CURRENT_BINARY_DIR}
CACHE PATH "installation path"
)
This doesn't work - I assume that this is because
CMAKE_INSTALL_PREF
On Sep 15, 2009, at 12:59 PM, Tyler Roscoe wrote:
On Tue, Sep 15, 2009 at 12:37:27PM -0600, James C. Sutherland wrote:
A related question: If I set CMAKE_INSTALL_PREFIX in my
CMakeLists.txt file
set( CMAKE_INSTALL_PREFIX
${CMAKE_CURRENT_BINARY_DIR}
)
and then run ccmake, I find that the
On Sep 15, 2009, at 2:15 PM, Tyler Roscoe wrote:
On Tue, Sep 15, 2009 at 02:08:44PM -0600, James C. Sutherland wrote:
So assuming that I wanted my CMakeLists.txt file to modify the
default
install location, but allow the user to change this, wouldn't I do
something like:
On Sep 15, 2009, at 12:50 PM, David Cole wrote:
I believe the word used was "reproducible" not "reliable" -- and I
think he was talking about the fact that it is not reproducible
because it depends on human interaction. Other than the human
sitting at the keyboard, I have found ccmake (an
On another mailing list that I subscribe to someone mentioned that
ccmake was not very reliable. Is this true?
A related question: If I set CMAKE_INSTALL_PREFIX in my
CMakeLists.txt file
set( CMAKE_INSTALL_PREFIX
${CMAKE_CURRENT_BINARY_DIR}
)
and then run ccmake, I find that the CMAKE_
On Sep 11, 2009, at 12:16 PM, Boudewijn Rempt wrote:
On Friday 11 September 2009, James C. Sutherland wrote:
On Sep 11, 2009, at 8:24 AM, Boudewijn Rempt wrote:
I'm getting weird linking errors after I upgraded to Snow Leopard.
This is probably because the library you are linking to
On Sep 11, 2009, at 8:24 AM, Boudewijn Rempt wrote:
I'm getting weird linking errors after I upgraded to Snow Leopard.
This is probably because the library you are linking to is 32-bit and
Snow Leopard's development environment will produce 64-bit binaries.
Try recompiling the library/libr
On Sep 2, 2009, at 3:27 AM, Tim Chan wrote:
I am getting the error:
-- Boost version required: 1.35.0. Found: ..
CMake Error at /usr/share/cmake-2.6/Modules/FindBoost.cmake:553
(MESSAGE):
Couldn't find the Boost libraries and/or include directory, or the
version
found is too old. Plea
I have a question about the static library option. I did:
set( HDF5_USE_STATIC_LIBRARIES 1 )
which should trigger linkage with static libraries, right? But what
I see in the link line is:
/usr/lib/libz.dylib /usr/lib/libm.dylib /jcs/apps/hdf5_serial/lib/
libhdf5_cpp.dylib /usr/lib/
On Aug 28, 2009, at 9:14 AM, James C. Sutherland wrote:
On Aug 24, 2009, at 10:20 AM, Will Dicharry wrote:
All,
I've committed the FindHDF5 and SelectLibraryConfigurations modules
to the CMake CVS repository.
Thanks for your input and feel free to contact me with questions
rega
On Aug 24, 2009, at 10:20 AM, Will Dicharry wrote:
All,
I've committed the FindHDF5 and SelectLibraryConfigurations modules
to the CMake CVS repository.
Thanks for your input and feel free to contact me with questions
regarding the modules.
--Will
Will,
Thank you for your initiativ
01:20:39 cmake-requ...@cmake.org wrote:
Message: 3
Date: Tue, 18 Aug 2009 15:39:14 -0600
From: "James C. Sutherland"
Subject: [CMake] F90 and legacy F77
To: cmake Mailing List
Message-ID: <1896503e-7a81-4517-8bcb-d11ae6db7...@utah.edu>
Content-Type: text/plain; charset=US-ASCII; format=
Is there a way to detect and use both F77 and F90 compilers?
I have some legacy F77 code that uses constructs that have been
deprecated/removed in F90. I would like to compile those F77 files
with a F77 compiler, compile the F90 files with an F90 compiler, and
then link them with the F90 c
The new boost libraries (as of version 1.40) will not have all of the
extra information like compiler info and boost version tagged onto the
library name.
Is there going to be a patch released for the FindBoost.cmake to
conform to the new library naming convention?
Will the change be a br
Out of curiosity, was a module ever submitted for this? I would love
to have a copy ;-)
Specifically, I am interested in detecting statically linked libraries
and the c++ library and header(s). I also would like to see a list of
libraries like the "z" library for compression.
Utility ex
Aah, now I see what you want...
#define TEST_VAR_VALUE ${TEST_VAR_VALUE}
is what you want.
But if I do
set( TEST_VAR_VALUE 0 )
then this results in TEST_VAR_VALUE being undefined rather than
having the value of 0 as I want. If the value is anything other
than zero it works. This is wha
In my configure.h.in file I have:
#cmakedefine TEST_VAR_VALUE
In my CMakeLists.txt file I have:
configure_file( ${PROJECT_SOURCE_DIR}/src/config.h.in $
{PROJECT_SOURCE_DIR}/src/configure.h )
set( TEST_VAR_VALUE "0" )
This results in a configure.h file containing:
/* #undef TEST_VAR_VALUE *
On Aug 9, 2009, at 10:33 AM, James C. Sutherland wrote:
On Aug 9, 2009, at 10:20 AM, Michael Wild wrote:
On 9. Aug, 2009, at 18:06, James C. Sutherland wrote:
In my configure.h.in file I have:
#cmakedefine TEST_VAR_VALUE
In my CMakeLists.txt file I have:
configure_file
On Aug 9, 2009, at 10:20 AM, Michael Wild wrote:
On 9. Aug, 2009, at 18:06, James C. Sutherland wrote:
In my configure.h.in file I have:
#cmakedefine TEST_VAR_VALUE
In my CMakeLists.txt file I have:
configure_file( ${PROJECT_SOURCE_DIR}/src/config.h.in $
{PROJECT_SOURCE_DIR}/src
In my configure.h.in file I have:
#cmakedefine TEST_VAR_VALUE
In my CMakeLists.txt file I have:
configure_file( ${PROJECT_SOURCE_DIR}/src/config.h.in $
{PROJECT_SOURCE_DIR}/src/configure.h )
set( TEST_VAR_VALUE "0" )
This results in a configure.h file containing:
/* #undef TEST_VAR
On Aug 5, 2009, at 1:33 AM, motes motes wrote:
I get the point of INCLUDE_DIRECTORIES (the various files used must
be located). But what is the point of LINK_DIRECTORIES and
TARGET_LINK_LIBRARIES?
LINK_DIRECTORIES specifies the path to search for libraries whereas
TARGET_LINK_LIBRARIES
On Jul 14, 2009, at 6:48 AM, Henrique Almeida wrote:
Hello,
When running "make test", some of my unit tests require that the
result is given by a correct output, instead of (just) the correct
return value (because, I'm specifically testing the ability to do
output). Is there a way to configure
On Jul 6, 2009, at 11:54 AM, James C. Sutherland wrote:
execute_command(
COMMAND ${test_cmd}
COMMAND ${CMAKE_COMMAND} -E compare_files ${output_blessed} $
{output_test}
RESULT_VARIABLE test_not_successful
OUTPUT_QUIET
ERROR_QUIET
)
if( test_not_successful )
message( SEND_ERROR
execute_command(
COMMAND ${test_cmd}
COMMAND ${CMAKE_COMMAND} -E compare_files ${output_blessed} $
{output_test}
RESULT_VARIABLE test_not_successful
OUTPUT_QUIET
ERROR_QUIET
)
if( test_not_successful )
message( SEND_ERROR "${output_test} does not match $
{output_blessed}!" )
endif( te
Thank you - this was a very useful explanation!
For any others who may look at this, note that
execute_command
should be
execute_process
On Jul 4, 2009, at 1:11 AM, Michael Wild wrote:
On 4. Jul, 2009, at 0:56, James C. Sutherland wrote:
I have a test that produces output
I have a test that produces output files that I would like to compare
against a "blessed" copy. Is there a way to do this in CMake/CTest?
This is probably a very simple thing, but I have not figured out how
to do it.
I have an
add_test( ... )
that creates the test, but I am not su
So, in general, when using globbing, YOU are responsible for rerunning
CMake whenever you've added a source file. Otherwise you run the
risk of
the new file not being compiled. Furthermore, you might accidentally
compile sources that were just lying around in your directory as test
code. Delet
I am trying to add most files from a directory to a variable. I first
glob the source files, and then want to remove the unwanted files from
the resulting string. Here is what I have:
set( empty "" )
file( GLOB f90_src src/*.f90 )
string( REPLACE
${PROJECT_SOURCE_DIR}/src/ark_m.f90
If you distribute on OS X then there are some CMake facilities to
help you create a stand alone .app bundle. Basically it will run
"install_name_tool" on the dylibs to make sure they reference
dependent libraries that are stored inside the .app bundle. Usually
this is a real pain to do by hand
If you are distributing your application on a Mac, then there are
myriads of methods to distribute the shared libraries with the
application.
James,
Can you point me in the right direction?
I haven't yet tried this on Linux, so I don't know what land mines
await me there...
Thanks,
Jame
On Jun 11, 2009, at 4:49 PM, James C. Sutherland wrote:
I have an executable that links with HDF5 (by the way, is there any
plan to create a "FindHDF5.cmake" file?).
The problem is that I have two flavors of libraries:
libhdf5_cpp.a libhdf5.a
libhdf5
I have an executable that links with HDF5 (by the way, is there any
plan to create a "FindHDF5.cmake" file?).
The problem is that I have two flavors of libraries:
libhdf5_cpp.a libhdf5.a
libhdf5_cpp.dylib libhdf5.dylib
When I do
set( HDF5_LIBRARIES hdf5_cpp hdf5
Alin,
Thank you for your patience. I now have this set via
set( CMAKE_EXE_LINKER_FLAGS "-mbig-endian" )
However, this does not seem to affect the reading of the big-endian
binary file (which is in read in fortran). Perhaps I am still not
setting it properly?
James
On Mar 31, 20
Let me try this again.
I have a f90 subroutine that performs file IO. I would like to link
that with a C++ driver.
If I use a fortran driver, I am able to set the binary format via
set( CMAKE_Fortran_FLAGS "-fconvert=big-endian" )
This results in file IO in big-endian format.
When u
g++ knows of -mbig-endian not -f
g++ --help -v provides no information on any endian operations.___
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
Please keep messages on-to
Should not the flag be passed to the linker? So in this case the
linker flags are read from the c++ compiler flags?
I tried adding
add_definitions( ${CMAKE_Fortran_FLAGS} )
But the g++ linker complained about an unrecognized option (the -
fconvert=big-endian option).
When linking
I am trying to compile a mixed f90/C++ program linked with g++.
The fortran compiler flags don't seem to be automatically passed
through to the g++ linker.
For example, I want to do:
set( CMAKE_Fortran_FLAGS "-fconvert=big-endian" )
When I do this for a fortran program (linked with g
80 matches
Mail list logo