Mantis - ISSUES

0010401: Cannot build dll's with MinGW due to 32k limit

The variable archiveCommandLimit in cmMakefileLibraryTargetGenerator.cxx is used to generate static libraries even if there is an issue with the 32k limit under windows.

Unfortunately this variable is not applied when building shared libraries only.
I have a library here that consists of more then 700 (generated) code files and cannot build it as a dll with MinGW because of the"The parameter is incorrect" error.

I'd suggest setting the archiveCommandLimit to 30k by default wether or not there is a static library being build.

0010409: Add revision column to update in dashboard

It would be very helpful to see what revision the build information is relating to, without following the link to the update information.
For subversion this could be easily shown in single column, as it is represented by a single revision number.
It might be more difficult for other version control systems as CVS, which do have such a concept of one single revision number for everything, but a timestamp might be used instead in these cases.

0010408: Time based reload for project dashboard page

It would be very helpful if the dashboard page could automatically reload itself
for example once every five minutes.

0010208: Add support for Clang Static Analyzer (similar to 1st class valgrind support)

CTest/CDash currently have great integrated support for valgrind. It can be invoked to run tests, and the HTML dashboard page shows those results in a special section.

I'd like to see something similar for the Clang Static Analyzer.

If you're not familiar with it, see:
http://clang-analyzer.llvm.org/ [^]

Here is their summary:
"The term "static analysis" is conflated, but here we use it to mean a collection of algorithms and techniques used to analyze source code in order to automatically find bugs. The idea is similar in spirit to compiler warnings (which can be useful for finding coding errors) but to take that idea a step further and find bugs that are traditionally found using run-time debugging techniques such as testing.

Static analysis bug-finding tools have evolved over the last several decades from basic syntactic checkers to those that find deep bugs by reasoning about the semantics of code. The goal of the Clang Static Analyzer is to provide a industrial-quality static analysis framework for analyzing C and Objective-C programs that is freely available, extensible, and has a high quality of implementation."

The project is cross platform and open source. One downside is that it does not yet support C++.

I used it to find 4 bugs in freetype, a component of VTK.

I imagine two possible UIs
- a special section for analyzer warnings, like for valgrind
- an additional column to the right of compiler warnings, showing the number of warnings flagged by the analyzer

0010407: Visibility is bad for 2d plots

XP, local server, trunk.

Open disk_out_ref.exo. All variables on. Apply.
Filters/ Data Analysis/ Plot over Line. Apply.
Split 2d, XY Plot View.

<> - Notice that the highlighted viewport is now the lower right viewport, the eyeball in the pipeline browser is on, but nothing is drawing. Click on the eyeball. The XY plot now draws.

0010406: TextureMapToSphere prevent seam is wrong

As per Greenfield, our TextureMapToSphere "Prevent Seam" is wrong. What it tries to do is draw the texture on the sphere twice. What it should do is draw the texture on the sphere, then CONTINUE to draw it one more time around. That way, if there are minor holes, the second draw of the texture will catch it.

A good test texture is here: Use the world map, found at http://www.oera.net/How2/PlanetTexs/EarthMap_2500x1250.jpg [^]

0010344: Windows CMake compilation under mingw-w64

When building CMake under Windows using a binary CMake 2.8.0 release and using a mingw-w64 toolchain, compilation fails because the Windows reserved keyword BOOL is used as a property name. Also, an #ifdef structure does not check for mingw, so it does not define a critical function.

These are all trivial patches (renaming 1 variable in +/- 11 files), and adding a conditional for mingw(-w64)

0010176: Add Support for IAR compiler

Add support for IAR compiler. CMakeCCompilerId.c.in has such definition already
/* IAR Systems compiler for embedded systems.
http://www.iar.com [^]
Not supported yet by CMake
#elif defined(__IAR_SYSTEMS_ICC__)
# define COMPILER_ID "IAR" */

0010090: check_prototype_definition() - Check if the protoype we expect is correct

check_prototype_definition(FUNCTION PROTOTYPE RETURN HEADER VARIABLE)

This check if the function exists and the prototype we want to use is correct. There are still functions which have different prototypes on UNICES :(

check_prototype_definition(getpwent_r
"struct passwd *getpwent_r(struct passwd *src, char *buf, int buflen)"
"NULL"
"unistd.h;pwd.h"
SOLARIS_GETPWENT_R)

Comments are welcome :)

0009414: Add a module/function to do the version check.

Attached is a module (FindPackageVersionCheck.cmake) which allows the author of a find module to let a function do the version check. I've created a function for that which helps to have less code in the module. It sets _FOUND so that find_package_handle_standard_args() can work with it. Only if REQUIRED is set, it will stop immediately.

You can find an example at: http://www.cynapses.org/tmp/patches/kde/FindPackageVersionCheck.cmake [^]

0010404: TextureMapToSphere reverses textures

I have a user that wants to map the world to a sphere.
ParaView 3.6.2, XP, local server.

* Sources/ SuperQuadric. Theta resolution == 128. Phi Resolution = 128.
* Turn off Toroidal.
* Apply.
* On the Display tab, Apply Texture.
* Use the world map, found at http://www.oera.net/How2/PlanetTexs/EarthMap_2500x1250.jpg. [^] [^] (Enter this on the Display tab)
* Notice that the earth looks good.

* Filters/ Alphabetical/ TextureMapToSphere.
* On the Display tab, Apply Texture.
* Notice that the earth looks reversed.

We are speculating this is because the original developer of this filter wanted to be able to wrap the texture on a sphere, and that the user then moved the eye INSIDE of the sphere - thus making the sphere the 3d window background. Thus, please make a user selectable option allowing the user to reverse the texture if desired.

0010405: TextureMapToSphere is not multiple server aware

I have a user that wants to map the world to a sphere.
ParaView 3.6.2, XP, 4 remote servers.

* Sources/ SuperQuadric. Theta resolution == 128. Phi Resolution = 128.
* Turn off Toroidal.
* Apply.

* Filters/ Alphabetical/ TextureMapToSphere.
* Turn off Prevent Seam.
* Color by Texture Coords.
* Notice that the colors have zones, but should actually smoothly wrap around the sphere.

0008486: Additional platforms support (Windows CE and Symbian)

Hi, we have implemented the subj and you can find the source here: http://github.com/AlexeyS/cmake/tree/master [^]

There are two main branches in the repo:
* WindowsCE - Windows CE MSVS project generator
* Symbian - Symbian .mmp project generator

Also we have implemented one additional feature: defining a set of valid values for a CMake variable. This feature was inspired by the need to enter long and boring Windows CE platform name (like "Windows Mobile 5.0 Smartphone"). This feature introduces SET_VALID_VALUES command, so CMake now can validate all the values you assign to the variable. Most importantly this allows us to introduce comboboxes for such variables in the CMake GUIs. The comboboxes feature is also implemented for the Qt frontend and for the MFC one. You can find this code in the ValidValues branch.

The rest are merged branches:

* WindowsCEwithValidValues - [WindowsCE + ValidValues] - WindowsCE generator now fills the combobox with platform names for which installed SDKs were detected.

* all - [WindowsCEwithValidValues + Symbian]

We would be happy to contribute this code to the CMake upstream.
Thank you for such a great tool as CMake.

Looking forwards for your comments, criticism, else.

0010385: Marks: Top axis labels are incorrectly positioned.

* Plot Over Line
* New XY Plot
* Open View Settings
* Click "Show Axis" for Top Axis.
Bugs:
- Labels appear under the top axis instead of above it. (Same is true for right axes, labels should appear on the right of the axis).
- Labels from bottom axis disappear (and don't reappear even if you turn off top axis).

0010403: Allow Cube Axes to be set by user

Please allow a user to set the min and max range of the cube axes, in all 3 axes. Also, please make this setting reset to auto range (i.e., current functionality) every time ParaView is restarted.

As per a discussion with Utkarsh, lets get this into 3.8.0 if it takes under a day of work. If it becomes problematic, lets defer to 3.10.

0010386: Marks: Log scale doesn't work for bottom/top axes

Log scale doesn't seem to work for bottom or top axes, it works for left/right axes.

0010402: Documentation needed for test DEPENDS property

See http://article.gmane.org/gmane.comp.programming.tools.cmake.user/27792 [^]

It would be nice to see documentation for the DEPENDS property under the 'Properties on Tests' documentation section and 'set_tests_properties()' command documentation.

Thanks.

0010399: 2.8.0 uses low resoultion title bar icon on Windows

2.8.0 uses low resoultion title bar icon on Windows where as 2.6.4 uses the nicer high resolution version. See attached image.

0010334: Linux release binaries missing GNU_STACK which results in +x mmaps under ctest

I downloaded the 2.8 cmake linux binary package
(http://www.cmake.org/files/v2.8/cmake-2.8.0-Linux-i386.tar.gz) [^] and the
ctest binary does not have a GNU_STACK section. This causes my kernel to
set the READ_IMPLIES_EXEC personality flag for ctest and all its children,
which means that all mmaps have +x set -- which causes one of my tests to
fail, in addition to being a security faux pas if ctest doesn't really need
+x memory.

Are you deliberately building these without a GNU_STACK section? Normally
gcc should add one by default, unless you have assembler sources.
You can tell the linker to make any existing GNU_STACK section -x, which
maybe will add a section if you don't have one? (I know /usr/bin/execstack
refuses to add one), with:

-Wl,-z,noexecstack

There's some basic info here:
http://www.gentoo.org/proj/en/hardened/gnu-stack.xml [^]

I'm not 100% sure there's no downside to adding the section if you want to
run on really old systems but certainly on any relatively modern system binaries should have a GNU_STACK section.

Looks like the 2.6.4 binaries on the cmake website have the same issue.
The version distributed by Redhat for Fedora does have GNU_STACK so I never
noticed it before.

Note that as a workaround I simply built 2.8 locally from sources using my
distro's toolchain and the resulting binary is fine.

0010400: Problem reading DICOM pixel data

We have been given some DICOM data from a TOSHIBA Asteion CT scanner. We previously had some data of the same type that loaded correctly.
After using GDCMSeriesFileNames::SetInputDirectory() and then calling fileNamesGenerator->GetSeriesHelper()->GetFirstSingleSerieUIDFileSet() nothing is found.
Debugging a bit I found that the calls to SerieHelper::AddFileName() give an error, because is trying to read the pixel data from the tag [7fe0 0010] which cannot be found. However, other softwares (DicomWorks) are able to load them.
I enclose the corresponding DCM file

0009457: XY Plot does not support large numbers (double length versus single length)

XY Plots do not support double precision numbers? I have CSV tables of numbers in the range e+48 e+49. The paraview Spreadsheet view works fine, but the XY plot is empty.

Dividing by e+20 with a text editor before reading the data makes it work, but that is not good quantitative data analysis practice. The scientist wants to see the true data in real units.

0010398: At least one of ErodeObjectMorphologyImageFilter and DilateObjectMorphologyImageFilter not behaving as expected

A closed image should contain the original image. However the result of applying DilateObjectMorphologyImageFilter followed by ErodeObjectMorphologyImageFilter results in an image which is shrunk compared to the original image.

This error is not observed when one uses BinaryDilateImageFilter and BinaryErodeImageFilter

I've uploaded a sample file and attached the code for a program that will read in an input file, perform the close, and save the result back. You can comment out and uncomment the appropriate lines to change the filter being used.

You may call it with
close_filter mandible.mha closed_mandible.mha 3
The error will be most clearly visible near axial slices 24 and 25 of the closed image.

0010395: cannot use add_dependencies() for imported libraries (together with externalproject_add())

Hi,

I want to use externalproject_add(), and in order to make sure that the external project (log4cplus) is built before my targets which link against it, I tried to create the dependency by creating an imported library, which depends on the externalproject-target, and then link against this imported library:

externalproject_add(EP_log4cplus
SOURCE_DIR ${CMAKE_SOURCE_DIR}/Packages/log4cplus/log4cplus-1.0.3
DOWNLOAD_COMMAND ""
INSTALL_DIR ${CMAKE_BINARY_DIR}/Packages/Install
CONFIGURE_COMMAND /configure --prefix=
BUILD_COMMAND make -j4
)

add_library(log4cplus UNKNOWN IMPORTED)
set_property(TARGET log4cplus PROPERTY
IMPORTED_LOCATION ${CMAKE_BINARY_DIR}/Packages/Install/lib/liblog4cplus.so )
add_dependencies(log4cplus EP_log4cplus)

...
add_executable(foo main.cpp)
target_link_libraries(foo log4cplus)

When I try to do this, cmake complains:
trunk/build-281/shared/tests/SystemInformation> make
CMake Error at CMakeLists.txt:31 (add_dependencies):
add_dependencies Adding dependency to non-existent target: log4cplus

It would be nice if this could be made to work.
Otherwise I didn't find an easier way than additionally to linking against the library, also make the target dependent on the external project target:

target_link_libraries(foo -llog4cplus)
add_dependencies(foo EP_log4cplus)

It would be much more elegant if I could avoid the add_dependencies() call here and instead just use the imported library.

Alex

0010397: Options for the visual studio generators

I added support for the /Zc:wchar_t- and /GF compiler options of
visual c++

0010396: Support for solution name change

I added a CMake variable (CMAKE_SOLUTION_NAME) to choose another name for the
generated visual studio solution file. This is very neat if you have multiple
differently configured of the same solution open at the same time and
want to quickly know which configuration working on.