cmake branch (Was: operator [])

[...]
> As the proting to cmake in the experimental branch is finished, and everything
> seems to work very well now, I would like to make this branch the trunk in
> the near future. So now is a good time to try out the experimental branch and
> give comments/feedback. I tested everything with newmat/lti/boost/ginac.

I had a (very) quick glance and I tested

* the combo boost/newmat (from debian package) on a debian testing
machine with g++ 4.1 which seems to work perfectly fine (apart from
the bug I reported to the ML friday about the warning (note to self:
not in bugzilla yet)!

* Trying to crosscompile for powerpc (using boost and the libnewmat
ppc debian package), generating the makefiles works out fine after
some tweaking


Orocos BFL version 0.4.9
Build type set to 'BFL' by user.
Found Doxygen -- documentation can be built
Searching for matrix lib newmat

wmeeusse's picture

cmake branch (Was: operator [])

> Some further smaller/bigger remarks:
> - Installation instructions? The INSTALL file seems not updated yet?

The installation instructions are updated in docs/installation_guide.tex
The same instructions are used on the webpage

> - The ChangeLog is not adapted yet

Added (all?) changes. What is actually the policy for the ChangeLog? In the
beginning of the log only BFL releases are mentioned, but towards the end
(most recent logs) it starts to look more and more like a copy of the commit
messages. What should (not) be mentioned in the ChangeLog?

> - How can I generate a static library?

This should still be fixed. Something for a bug report?

> - There is build code which seems bfl-unrelated? e.g.
> config/FindGthread.cmake

This code has been removed (and config/FindCppUnit.cmake has been added)

> - There is only one test-executable generated.
> suppose this is intentional?

This wasn't intentional. Now we have multiple tests

> - The CMAKE_BUILD_TYPE defaults to BFL. This
> seems somewhat
> counterintuitive (I expect there Release, or Debug etc)?

The BFL build type is removed. Default is Debug now.

Wim
_______________________________________________
I hereby promise not to top-post on the
BFL mailing list
BFL [..] ...
http://lists.mech.kuleuven.be/mailman/listinfo/bfl

cmake branch (Was: operator [])

[...]
> Some further smaller/bigger remarks:
> - Installation instructions? The INSTALL file seems not updated yet?
> - The ChangeLog is not adapted yet
> - How can I generate a static library?
> - There is build code which seems bfl-unrelated? e.g. config/FindGthread.cmake
> - There is only one test-executable generated. I suppose this is intentional?
> - The CMAKE_BUILD_TYPE defaults to BFL. This seems somewhat
> counterintuitive (I expect there Release, or Debug etc)?

At first sight, it also seems there is no detection if cppunit is installed?

Klaas
_______________________________________________
I hereby promise not to top-post on the
BFL mailing list
BFL [..] ...
http://lists.mech.kuleuven.be/mailman/listinfo/bfl

wmeeusse's picture

cmake branch (Was: operator [])

> At first sight, it also seems there is no detection if cppunit is
> installed?

I added a check for cppunit. If cppunit is not found, the tests are disabled.

Wim
_______________________________________________
I hereby promise not to top-post on the
BFL mailing list
BFL [..] ...
http://lists.mech.kuleuven.be/mailman/listinfo/bfl

cmake branch (Was: operator [])

On 4/30/07, Wim Meeussen <wim [dot] meeussen [..] ...> wrote:
> > > Can you svn up and try again?
> >
> > Same error (I even reran configure to be sure)
>
> en nog eens! :-)

Now it works, 2 more problems I spotted
- The built library is not installed, only the header files (at least
when installing with make install DESTDIR=/my/embedded/rootfs
- the generated pkgconfig file is not correct (but that's due to its
louzy support for cross-compilation)

Klaas
_______________________________________________
I hereby promise not to top-post on the
BFL mailing list
BFL [..] ...
http://lists.mech.kuleuven.be/mailman/listinfo/bfl

wmeeusse's picture

cmake branch (Was: operator [])

> Now it works, 2 more problems I spotted
> - The built library is not installed, only the header files (at least
> when installing with make install DESTDIR=/my/embedded/rootfs

Fixed!

> - the generated pkgconfig file is not correct (but that's due to its
> louzy support for cross-compilation)

Don't know what you mean since I'm not familiar with cross-compilation. The
current pkgconfig file is actually pretty much the same as the one we had
before the cmake port (but that one probably was not correct eather :-).

I now also added the include paths for the matrix/rng/ginac libs in the
pkgconfig file.

Wim
_______________________________________________
I hereby promise not to top-post on the
BFL mailing list
BFL [..] ...
http://lists.mech.kuleuven.be/mailman/listinfo/bfl

cmake branch (Was: operator [])

On 4/30/07, Klaas Gadeyne <klaas [dot] gadeyne [..] ...> wrote:
> On 4/30/07, Wim Meeussen <wim [dot] meeussen [..] ...> wrote:
> > > > Can you svn up and try again?
> > >
> > > Same error (I even reran configure to be sure)
> >
> > en nog eens! :-)
>
> Now it works, 2 more problems I spotted
> - The built library is not installed, only the header files (at least
> when installing with make install DESTDIR=/my/embedded/rootfs
> - the generated pkgconfig file is not correct (but that's due to its
> louzy support for cross-compilation)

And maybe yet another one?

./tests/test_runner
DiscretePdf::Samplefrom(int, void *): No such sampling method
Box-Muller not implemented yet!
Box-Muller not implemented yet!
WeightedSample::WeightSet() Weights should only be positive
...F.....E

pdf_test.cpp:211:Assertion
Test name: PdfTest::testLinearAnalyticConditionalGaussian
equality assertion failed
- Expected: 11.599999
12.800000

- Actual : 5.800000
6.400000

##Failure Location unknown## : Error
Test name: MatrixwrapperTest::testMatrixwrapperValue
uncaught exception of unknown type

Failures !!!
Run: 8 Failure total: 2 Failures: 1 Errors: 1
_______________________________________________
I hereby promise not to top-post on the
BFL mailing list
BFL [..] ...
http://lists.mech.kuleuven.be/mailman/listinfo/bfl

cmake branch (Was: operator [])

> And maybe yet another one?
>
> ./tests/test_runner
> DiscretePdf::Samplefrom(int, void *): No such sampling method
> Box-Muller not implemented yet!
> Box-Muller not implemented yet!
> WeightedSample::WeightSet() Weights should only be positive
> ...F.....E
>
> pdf_test.cpp:211:Assertion
> Test name: PdfTest::testLinearAnalyticConditionalGaussian
> equality assertion failed
> - Expected: 11.599999
> 12.800000
>
> - Actual : 5.800000
> 6.400000
>
>
> ##Failure Location unknown## : Error
> Test name: MatrixwrapperTest::testMatrixwrapperValue
> uncaught exception of unknown type
>
> Failures !!!
> Run: 8 Failure total: 2 Failures: 1 Errors: 1

This one should be solved now.
Apparently when constructing a vector in NEWMAT it is not initialised with
zero elements.

Tinne
_______________________________________________
I hereby promise not to top-post on the
BFL mailing list
BFL [..] ...
http://lists.mech.kuleuven.be/mailman/listinfo/bfl

Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm

wmeeusse's picture

cmake branch (Was: operator [])

> However, when trying to build, I get
> $ make VERBOSE=1
> [ 3%] Building CXX object
> src/CMakeFiles/orocos-bfl.dir/wrappers/matrix/matrix_wrapper.o
> /opt/amltd/bin/powerpc-linux-g++ -Dorocos_bfl_EXPORTS -O2 -DNDEBUG
> -fPIC -I/tmp/bfl-cmake -I/tmp/bfl-cmake/src/wrappers/matrix
> -I/tmp/bfl-cmake/build-ppc-linux/src -I/tmp/bfl-cmake/src/wrappers/rng
> -I/tmp/bfl-cmake/src/sample -I/tmp/bfl-cmake/src/pdf
> -I/tmp/bfl-cmake/src/model -I/tmp/bfl-cmake/src/filter
> -I/tmp/bfl-cmake/src/smoother -o
> src/CMakeFiles/orocos-bfl.dir/wrappers/matrix/matrix_wrapper.o -c
> /tmp/bfl-cmake/src/wrappers/matrix/matrix_wrapper.cpp
> In file included from
> /tmp/bfl-cmake/src/wrappers/matrix/vector_wrapper.h:217, from
> /tmp/bfl-cmake/src/wrappers/matrix/matrix_NEWMAT.h:27, from
> /tmp/bfl-cmake/src/wrappers/matrix/matrix_wrapper.h:272, from
> /tmp/bfl-cmake/src/wrappers/matrix/matrix_wrapper.cpp:1:
> /tmp/bfl-cmake/src/wrappers/matrix/vector_NEWMAT.h:28:29:
> newmat/newmatio.h: No such file or directory

Can you svn up and try again?

Wim
_______________________________________________
I hereby promise not to top-post on the
BFL mailing list
BFL [..] ...
http://lists.mech.kuleuven.be/mailman/listinfo/bfl

cmake branch (Was: operator [])

On 4/30/07, Wim Meeussen <wim [dot] meeussen [..] ...> wrote:
> > However, when trying to build, I get
> > $ make VERBOSE=1
> > [ 3%] Building CXX object
> > src/CMakeFiles/orocos-bfl.dir/wrappers/matrix/matrix_wrapper.o
> > /opt/amltd/bin/powerpc-linux-g++ -Dorocos_bfl_EXPORTS -O2 -DNDEBUG
> > -fPIC -I/tmp/bfl-cmake -I/tmp/bfl-cmake/src/wrappers/matrix
> > -I/tmp/bfl-cmake/build-ppc-linux/src -I/tmp/bfl-cmake/src/wrappers/rng
> > -I/tmp/bfl-cmake/src/sample -I/tmp/bfl-cmake/src/pdf
> > -I/tmp/bfl-cmake/src/model -I/tmp/bfl-cmake/src/filter
> > -I/tmp/bfl-cmake/src/smoother -o
> > src/CMakeFiles/orocos-bfl.dir/wrappers/matrix/matrix_wrapper.o -c
> > /tmp/bfl-cmake/src/wrappers/matrix/matrix_wrapper.cpp
> > In file included from
> > /tmp/bfl-cmake/src/wrappers/matrix/vector_wrapper.h:217, from
> > /tmp/bfl-cmake/src/wrappers/matrix/matrix_NEWMAT.h:27, from
> > /tmp/bfl-cmake/src/wrappers/matrix/matrix_wrapper.h:272, from
> > /tmp/bfl-cmake/src/wrappers/matrix/matrix_wrapper.cpp:1:
> > /tmp/bfl-cmake/src/wrappers/matrix/vector_NEWMAT.h:28:29:
> > newmat/newmatio.h: No such file or directory
>
> Can you svn up and try again?

Same error (I even reran configure to be sure)

Klaas
_______________________________________________
I hereby promise not to top-post on the
BFL mailing list
BFL [..] ...
http://lists.mech.kuleuven.be/mailman/listinfo/bfl

cmake branch (Was: operator [])

On 4/30/07, Klaas Gadeyne <klaas [dot] gadeyne [..] ...> wrote:
> [...]
> > As the proting to cmake in the experimental branch is finished, and everything
> > seems to work very well now, I would like to make this branch the trunk in
> > the near future. So now is a good time to try out the experimental branch and
> > give comments/feedback. I tested everything with newmat/lti/boost/ginac.

BTW, can you put a bug (project in bugzilla) for this?

Klaas
_______________________________________________
I hereby promise not to top-post on the
BFL mailing list
BFL [..] ...
http://lists.mech.kuleuven.be/mailman/listinfo/bfl