- Provide std::shared_ptr interface (#95)
- Windows compat and style fixups (#90) * add visibility macros to public functions * rename private namespace 'class_loader_private' to 'impl' to match ros2 branch
- use new headers to build library (#93)
- Contributors: Mikael Arguedas
- Stop checking for c++11 support (#87) all Melodic targeted platforms use gnu++14 so checking and forcing -std=c++11 doesn't make sense anymore
- [ABI breaking] use std::string references for exceptions (#86)
- deprecate .h headers in favor of .hpp headers (#81)
- provide a script with exhaustive rules for header replacement
- comply with package format2 xsd (#83)
- [ABI breaking] Exceptions fixups (#82) * inline exceptions * use throw statement rather than function
- [linter] add nolint for global std::string used for testing (#79)
- use auto for all for loops iterating on vectors/maps (#78)
- Add systemLibraryFormat and systemLibraryPrefix functions (#77)
- [ABI breaking] Bring melodic-devel closer to ros2 branch (#76) * comply with extra and pedantic compiler flags * use c++11 nullptr instead of NULL * make ABI breaking change for explicit constructors * make linters happy * no need to support console_bridge < 0.3.0 anymore * remove obsolete todo * add virtual destructor in test * vector size() returns size_t * simplify branching
- [fix warnings] c++11 requires at least one argument for ... (#71)
- [linter] Use std::string::empty instead comparing with an empty string (#69)
- [linter] wrap console bridge invocation lines (#68)
- OSRF and not willow in licence header (#67)
- Contributors: David Wagner, Mikael Arguedas
- Fix console_bridge marcos definition (#66)
- Style overhaul (#64) (#62)
- Add copyright notice to unique_ptr_test.cpp (#65)
- Contributors: Maarten de Vries, Mikael Arguedas
- switch to package format 2 (#56)
- remove trailing whitespaces (#55)
- use CONSOLE_BRIDGE_X logging macros (#52)
- Contributors: Mikael Arguedas, jmachowinski
- Made changes to two locking mechanisms inside class loader core's loadLibrary() function: 1) I added a lock to the 'addClassLoaderOwnerFor...' function to protect it against a race condition with the unloadLibrary() function. 2) I also raised the loader lock to cover the whole function. Previously the check to see if a library is already loaded and the actual loading of the library was not atomic. Multiple threads could create shared library objects, for example.
- Contributors: Jonathan Meyer
- Add ClassLoader::createUniqueInstance (#38) * Wrap comments on createInstance and friend. * Delegate createInstance and createUnmanagedInstance to private impl. * Add ClassLoader::createUniqueInstance. * MultiLibraryClassLoader: Factor out getClassLoaderForClass. * MultiLibraryClassLoader: Add unique_ptr API. * Add tests for unique_ptr API.
- Contributors: Maarten de Vries
- cleanup: don't use active_class_loaders_[library_path] for existence test (#35) * cleanup: don't use active_class_loaders_[library_path] for existence test - this accumulates map entries with NULL pointer - fixing it, allows some cleanup * list headers in CodeBlocks / QtCreator * explicitly list all headers
- Merge pull request #34 from rhaschke/fix-on-demand-unloading fix on demand unloading
- Merge pull request #32 from saarnold/fixed_unset_variable_evaluation fixed evaluation of undefined variable
- fixed evaluation of undefined variable
- not unloading the ClassLoaders (to avoid the SEVERE WARNING) doesn't work either
- bugfix: enable on-demand loading/unloading with MultiClassLoader - enforce loading of library in loadLibrary(), otherwise we cannot know - don't unload libraries in destructor when on-demand-unloading is enabled
- extra utest: MultiClassLoaderTest.lazyLoad succeeds two times in a row?
- added MultiLibraryClassLoader unittest
- Contributors: Mikael Arguedas, Robert Haschke, Sascha Arnold
- update maintainer
- Merge pull request #26 from goldhoorn/indigo-devel Added option to disable the catkin build
- Added option to disable the catkin build
- Contributors: Esteve Fernandez, Matthias Goldhoorn, Mikael Arguedas
- Fixed wrong handling of false statement (pkg-config was not installed)
- Make catkin optional again
- Contributors: Esteve Fernandez, Janosch Machowinski, Matthias Goldhoorn
- Depend on boost
- Use FindPoco.cmake from ros/cmake_modules
- Honor BUILD_SHARED_LIBS and do not force building shared libraries.
- Contributors: Esteve Fernandez, Gary Servin, Scott K Logan
- Use system-provided console-bridge
- Contributors: Esteve Fernandez
- Changed format of debug messages so that rosconsole_bridge can correctly parse the prefix
- Improved debug output
- fix race condition with multi threaded library loading (#16)
- fix missing class name in logWarn output
- check for CATKIN_ENABLE_TESTING (#10)
- fix find Poco to return full lib path (#8)
- add missing runtime destination for library under Windows
- add Boosst component system
- improve check for Poco foundation and headers (#7)
- use find_package for Poco/dl instead to make it work on other platforms
- update Poco cmake file to include libdl on non-windows systems
- No longer CATKIN_DEPEND on console_bridge
- change warning message for managed/unmanaged instance mixture in lazy loading mode
- fix all instances marked as unmanaged
- fix redundant destructor definition being pulled into plugin library for metaobjects instead of being contained with libclass_loader.so
- fix syntax error for logInform
- downgrade some warning messages to be info/debug
- add safety checks for mixing of managed/unmanaged mixing as well as pointer equivalency check between graveyard and newly created metaobjects
- fix compile issue on OSX in dependent packages (#3)
- add more debug information
- first public release for Groovy