file format. Tested for ELF on x86 Linux, COFF and Mach-O prepared but
commented out since I don't have such systems available. Please have a look
a CMakeDetermineCompilerId.cmake and enable the test for them too.
Only add the option for using chrpath if the executable format is ELF
Alex
BUG: FindSDL now has flag it responds to so it will not find/link against SDLmain. This is required to build libraries instead of applications since they don't have main().
ENH: All modules have a predictable search order, where environmental variables are searched before system paths. This is designed to make automation easier for those that need to automatically build projects without intervention but may be using alternative install locations for isolated testing.
ENH: New modules for OpenSceneGraph, Freetype, GDAL, Lua, QuickTime, GIFLIB, Producer, OpenThreads.
STYLE: Added documentation explaining peculuar SDL_LIBRARY_TEMP variable in SDL
module when library find is incomplete.
installing without having to link the target again -> can save a lot of time
chrpath is handled very similar to install_name_tool on the mac. If the
RPATH in the build tree file is to short, it is padded using the separator
character.
This is currently disabled by default, it can be enabled using the option
CMAKE_USE_CHRPATH. There are additional checks whether it is safe to enable
it. I will rework them and use FILE(READ) instead to detect whether the
binaries are actually ELF files.
chrpath is available here
http://www.tux.org/pub/X-Windows/ftp.hungry.com/chrpath/
or kde svn (since a few days): http://websvn.kde.org/trunk/kdesupport/chrpath/
Alex
major issues:
-access() doesn't return false for an empty string (#ifdefed in cmake)
-dlopen() doesn't return 0 on failure (#ifdefed in cmake and fixed now in Syllable)
-the kwsys and Bootstrap tests fail with timeout due to the fact that I'm doing all that in qemu, which is quite slow
-RPATH is now supported, so without modifying the test adapting DLL_PATH in Syllable is required for the tests to succeed
-the Plugin test fails with an undefined reference to example_exe_function() in example_mod_1, it seems this isn't supported under Syllable
Alex
STYLE changes including: deprecation of wxWidgets_USE_LIBS in favor of using
standard FIND_PACKAGE COMPONENTS, removed some CMake 2.4.2 compatibility patches,
use of execute_process instead of exec_program, etc.
the main project. I think it doesn't make sense if a project disables
RPATH, uses CHECK_C_SOURCE_RUNS() to see if something is able to run, and
this succeeds because it has been built with RPATH, but an executable built
within the project won't be able to run since it has been built without
RPATH.
Alex
It seems rpmbuild can't handle paths with spaces, it complains that
Buildroot takes only one tag (or something like this), quoting and escaping
don't seem to help.
Alex
on NetBSD where there are separate curses and ncurses libraries, and where
the curses library is found, which doesn't work for ccmake while the
existing ncurses library would work.
With this change it should be possible to test whether the found curses lib
provides ncurses functionality.
Alex
ENH: change test for endianess from TRY_RUN() to TRY_COMPILE() by testing
the binary image of a 16bit integer array, tested on Linux x86, Intel Mac
and Sun (big endian)
Alex
cmMakefile.cxx, but now in the platform files and are now valid for the
target platform, not the host platform.
New variables CMAKE_HOST_WIN32, CMAKE_HOST_UNIX, CMAKE_HOST_APPLE and
CMAKE_HOST_CYGWIN have been added in cmMakefile.cxx (...and have now to be
used in all cmake files which are executed before
CMakeSystemSpecificInformation.cmake is loaded). For compatibility the old
set is set to the new one in CMakeDetermineSystem.cmake and reset before the
system platform files are loaded, so custom language or compiler modules
which use these should still work.
Alex
the cmake run and add macros print_enabled/disabled_features() and
set_feature_info(), so projects can get a nice overview at the end of the
cmake run what has been found and what hasn't
FIND_PACKAGE() automatically adds the packages to these global properties,
except when used with QUIET
Maybe this can also be useful for packagers to find out dependencies of
projects.
Alex
FindThreads.h
BUG: improve CheckC(XX)SourceRuns.cmake so that it works with cross
compiling, the return value has to go in the cache but shouldn't overwrite
the actual return value, and it should go only in the cache if we have a
result from try_run() otherwise we won't get here again if we have a result
later on
Alex
-add /usr/openwin/include and /usr/openwin/lib to the default search paths
-add /${CMAKE_INSTALL_PREFIX}/(lib|bin|include) to the default cmake search
paths -> this should help users who install stuff in their home
Alex
-remove the unscriptable commands also from the cpack cmake
-use CPACK_PACKAGE_CONTACT in CMakeCPack.cmake, it's used in the nsis and
the deb generator
-make set_properties() scriptable
-use a non-const char array for adding the python modules
Alex
loaded
Additionally the makefile in cmCPackGenericGenerator is now protected
instead of private, so with these two changes the cpack generators should
now be able to find their tools and how to call these tools from cmake
scripts, instead of hardcoding the search order and command line (as done
e.g. in cmCPackZIPGenerator.cxx)
Alex
FIND_PACKAGE_HANDLE_STANDARD_ARGS(), so cmake modules can specify their own
better failure messages. If the default is ok use "DEFAULT_MSG".
Do this also for FindBoost.cmake (#5349)
Alex
variable is already added to the cache inside cmTryRunCommand.cxx, so the
value used here was ignored. Additionally the INTERNAL made it internal,
which shouldn't be done when cross compiling, since here the user is
required to edit this variable manually e.g. using ccmake.
Alex
two-step priority (None or Prefered)
Current order: ASM 0, C 10, Fortran 20, CXX 30, Java 40
This is the same order as automake choses:
http://www.gnu.org/software/automake/manual/html_node/How-the-Linker-is-Chosen.html
This change should be backward compatible:
if there is a project using fortran and CXX, they had to set the
LINKER_LANGUAGE explicitely, otherwise cmake complained (but still generated
the project files). Explicitely setting the linker language still overrides
automatic detection.
If somebody has a custom language for cmake and the PREFERENCE starts with
"P", its changed to 100, which gives it preference over all other languages
(except the other custom languages which have also "Prefered"). "None" is
converted to 0.
Alex
compiling an executable (amd thus cannot build the compiler-id program)
easier by providing CMAKE_FORCE_XXX() macros which force cmake to use the
given compilers anyway
Alex
CMAKE_SYSTEM_AND_C_COMPILER_INFO_FILE, CMAKE_SYSTEM_AND_CXX_COMPILER_INFO_FILE,
CMAKE_SYSTEM_AND_C_COMPILER_AND_PROCESSOR_INFO_FILE and CMAKE_SYSTEM_AND_CXX_COMPILER_AND_PROCESSOR_INFO_FILE
Instead of presetting these variables to arbitrary filenames, users should
set up CMAKE_SYSTEM_NAME and the compilers correctly and also create a
Platform/ directory so these files will all follow the official cmake style,
which should make it easier to understand and debug project which have their
own platform/toolchain support files.
-remove support for a suffix to MS crosscompilers, since this is not (yet)
supported by cmake and might confuse users
Alex
CMakeDetermineSystem.cmake, since CMAKE_SYSTEM_NAME might already be preset
when using cmake for cross compiling
use type STRING instead of FILEPATH since otherwise a strange filename was
generated
Alex
second part copies the values from the cmake variables into internal maps.
So this can now be done after the compiler-specific information has been
loaded, which can now overwrite more settings.
Alex
Tested with various gcc, XCode, MSVC7, sdcc
For OSX when doing TRY_COMPILE() CMAKE_OSX_ARCHITECTURES is used, if there are different results an error is generated. CMAKE_OSX_ARCHITECTURES can be overwritten for the TRY_COMPILES with CMAKE_TRY_COMPILE_OSX_ARCHITECTURES.
Alex
really work by reverting X11_LIBRARIES back to the old version
-add some more X11_xxx_FOUND variables
-reformat comments at the top
-always use IF(INCLUDE_DIR and LIB) for setting FOUND to TRUE
Alex
creates two cache variables, one for the RUN_RESULT, one for the RUN_OUTPUT
(if required), which can be set or preset by the user. It has now also two
new arguments: RUN_OUTPUT_VARIABLE and COMPILE_OUTPUT_VARIABLE (the old
OUTPUT_VARIABLE merges both), so if only COMPILE_OUTPUT_VARIABLE is used the
run time output of the TRY_RUN is unused and the user doesn't have to care
about the output when crosscompiling. This is now used in FindThreads.cmake,
CheckC/CXXSourceRuns.cmake and TestBigEndian.cmake, which used the output
only for the logfile (compile output is still there). Test/TryCompile/ now
also tests the behaviour of OUTPUT_VARIABLE, RUN_OUTPUT_VARIABLE and
COMPILE_OUTPUT_VARIABLE.
Alex
conversion of hex and srec files to binary.
Without this automatic conversion, everywhere where a compiled file is parsed for strings the
a file(HEX2BIN somefile binfile) command has to be added otherwise it will
not work for these compilers. I tried this with DetermineCompiler and
CheckTypeSize and nobody will do this except the users who work with such
compilers. For them it will break if they don't add this conversion command
in all these places.
If FILE(STRINGS) is used with a text file, it
will in most cases still work as expected, since it will only convert hex
and srec files. If a user actually wants to get text out of hex files, he
knows what he's doing and will see the hint in the documentation.
Anyway, it should work without having to create a temporary file, will work
on this later.
Alex
CMAKE_SOURCE_DIR can't be used there
ENH: modify CMakeCCompilerId.c and .h so that sdcc can compile them. As they
were the preprocessor produced:
9 "test.c"
static char const info_compiler[] = "INFO:compiler["
# 40 "test.c"
""
"]";
and the mixing of the preprocessing directives and the string constants
didn't work.
Alex