A binary for both 1.8.x and 1.9.x is now shipped as part of the binary build. A stub file is used to dynamically require the correct binary based on your current version of Ruby.
Some updates to the Rakefile to help build multiple binaries.
An internal error formatting function now resorts to the default LANGID on error 1815 (ERROR_RESOURCE_LANG_NOT_FOUND). Thanks go to Little Snake for the spot.
Added a default Rake task and refactored the clean task.
Fixed a warning that showed up with MinGW/gcc caused by an unnecessary pointer dereference.
Some updates to the Rakefile and gemspec. Gem building is now handled via Rake tasks.
Minor updates to the test suite.
Reverted the change in 1.4.4. We need unsigned longs in a few cases. Consequently, you should upgrade windows-pr to 1.0.8 or later because some return values were set to -1 instead of 0xFFFFFFFF as they are now.
Updated one test to validate return value from failed function.
Fixed a bug where functions that returned a long return type were unsigned instead of signed.
The Rakefile has been refactored somewhat, including the removal of FFI related tasks. Those are now in a separate branch in the repository.
The test-unit library is now a development dependency instead of a runtime dependency.
Bug fix for mingw.
License now set to Artistic 2.0.
Updated the internal StringError() function to better handle the possibility of the English .mui file not being found. Thanks go to Michel Demazure for the spot.
Callback handling improvements.
Updated the gemspec description.
Now compatible with Ruby 1.9.x.
In what will go down as, “It seemed like a good idea at the time”, I have removed the feature where W (wide) character functions were used first if your $KCODE environment variable was set to “UTF8”. It caused too many headaches in practice, especially amongst Rails users. You must now always explicitly specify ‘W’ in the constructor if that’s what you want.
Fixed RF bug #23944 - bad error message for MSVCRT functions that failed to load properly.
Fixed RubyForge bug #23395, which was caused by inadvertently modifying a variable within a loop. This caused callbacks to fail in certain situations.
Added the Win32::API::LoadLibraryError and Win32::API::PrototypeError classes to provide more fine grained handling of possible error conditions in the constructor. These are both subclasses of Win32::API::Error.
Removed the Win32::API::CallbackError class.
Changed the upper limit on prototypes from 16 to 20. It turns out that there are actually Windows functions with more than 16 prototypes.
Refactored a high iteration test so that it counts as only one test instead of 1000.
Fixed bug in the error message for illegal prototypes and illegal return types where the character in question would come out as empty or garbage.
Passing a bad return type to Win32::API::Callback now raises an error.
Updated the error message for illegal return types to say, “Illegal return type” instead of “Illegal prototype” as it did previously.
The error message for a bad function name passed to Win32::API.new now matches JRuby’s FFI error message.
Improved the handling of msvcrt functions with regards to skipping ‘A’ and ‘W’ checks. Previously it was checking against the literal string ‘msvcrt’. Now it checks against any string that starts with ‘msvcr’.
Added test-unit 2.x as a prerequisite.
Added tests for the Win32::API::Callback#address method.
Added tests to all Win32::API classes that explicitly check error messages.
Fixed and updated callback handling.
Fixed wide string return handling for pointers and strings.
Added the Win32::API::Callback#address instance method.
All errors are now in English instead of your native language, because that’s what Ruby itself does.
Added support for the ‘S’ (string) prototype and return type. It can be used instead of ‘P’ (pointer) for const char*.
Some internal refactoring. The attempts to load ANSI and/or Wide character versions of functions are skipped for MSVCRT functions, since they do not exist. This eliminates some unnecessary LoadLibrary() calls.
Added a couple of gem building Rake tasks.
Added a few more tests.
Added the Windows::API::Function class. This is a subclass of Win32::API meant only for use with raw function pointers.
Some documentation updates in the source and README files.
Added the effective_function_name method. This allows you to see what the actual function name is that was defined, e.g. GetUserNameA vs GetUserNameW.
Replaced an instance of _tcscmp with strcmp. The case in question was always going to be ASCII.
Cleaned up some -W3 warnings.
Added the build_manifest task to the Rakefile, which is automatically run if you’re using a version of Ruby built with VC++ 8 or later. This builds a ruby.exe.manifest file (if it doesn’t already exist).
The API.new method now defaults to “W” (wide character functions) before “A” (ANSI functions) if the $KCODE global variable is set to ‘u’ (UTF8).
Minor improvements to the Rakefile.
Fixed a bug where methods that returned pointers (‘P’) could choke if the resulting pointer was 0 or NULL. In this case, nil is now returned instead.
Tweak to the extconf.rb file that helps the gem build it from source properly.
Fixed a subtle but dangerous copy-on-write bug in the API#call method.
Fixed a bug in an internal struct member that was causing segfaults. Thanks go to Lars Olsson for the spot.
Fixed the ‘install’ task in the Rakefile. This only affected native builds, not the prebuilt binary.
Added a few more tests.
Functions declared with a void prototype no longer require an explicit nil argument to fulfill the arity requirement. You can still call them with an explicit nil if you wish, however.
Fixed the gemspec for the native build.
Initial release
Generated with the Darkfish Rdoc Generator 1.1.6.
![[+]](./images/find.png "show/hide quicksearch"){kind=small}