- Update to version 2.7.1 * Fix inclusion mechanism for FFTs which could cause a host application to crash in certain circumstances due to conflicting versions of C-linkage symbols version 2.7: * Add ability to PluginLoader to list plugins only in (or not in) certain library files * Fix fixed-sample-rate output timestamps in printout from the simple host, and add regression test script using test plugin * Switch the convenience FFT interface for plugin usage from the very slow reference implementation previously provided, to the somewhat faster KissFFT * Add a top-secret compiler flag to switch internal FFTs to single-precision only * Fix some small but long-standing memory leaks and minor bugs version 2.6: * Add a simple C-linkage API for plugin library discovery and loading -- the missing part when trying to use Vamp plugins from C or a language with only C-linkage foreign-function interface * Fix bug in PluginBufferingAdapter reset() function, which failed to reset timestamp rewrites (symptom: timestamps appeared to continue to ascend from end point on next run after a reset) * Fix sometime 1-sample rounding error in PluginInputDomainAdapter * Update build files for more recent OS versions * Add -L option to simple host, as synonym for --list-full * Add a combined Visual C++ solution that includes the plugin SDK, host SDK, example plugins, and simple host projects
- Build with -fPIC
- Update to version 2.5 * Fix incorrect handling of FixedSampleRate outputs in the PluginBufferingAdapter. Un-timestamped features on these outputs were incorrectly being timestamped from the process timestamp, where they should have been timed relative to the previous feature on the output according to the output's sample rate. * Fix return of uninitialised memory through getCurrentProgram when plugin specifies no programs * Ensure output sample rate is initialised (to 0) for variable-rate outputs where the plugin forgets to set a rate - Use download Url provided by upstream - Remove vamp-plugin-sdk-2.0-libdir.patch; use a hack instead - Remove redundant documentation related files with fdupes - Perform spec-cleaner on spec file
- Update to version 2.2.1, split libs into libvamp-hostsdk3 and libvamp-sdk2 sub packages and moved binaries into main package. - Enabled api doc build and generate man pages with help2man. - Upstream changes : * Add ProcessTimestampMethod to PluginInputDomainAdapter, offering the ability to decide how data buffering and timestamping works for frequency-domain plugins (whose process timestamps have to be at the centre of each processing block rather than the start) * Make PluginBufferingAdapter take into account any timestamp shift introduced by a PluginInputDomainAdapter that it wraps. * Fix crash in PluginAdapter when plugin library getDescriptor fails on plugin construction. * Add plugin skeleton files as starting point for new developers.
- initial package created from fedora spec
- Update to version 2.10 * Add a method to PluginWrapper, the host-side base for adapters like PluginInputDomainAdapter that modify the processing behaviour of a plugin, that tells it to "disown" the wrapped plugin. The former behaviour, and still the default, is for the wrapper to take ownership of the wrapped plugin. The alternative behaviour makes it easier to mix these classes with some modern C++ styles that use managed pointers version 2.9 * Fix non-thread-safe behaviour in PluginAdapter. Plugins built using the adapter classes in version 2.8 or earlier cannot safely be used simultaneously across threads with other instances of themselves or of other plugins in the same library (i.e. shared object). Hosts have been required to provide synchronisation for such cases. Version 2.9 introduces synchronisation in the plugin, making this usage safe. Unfortunately this does not make host code safe when using older plugin builds, as the problem and its fix are in the plugin side of the SDK. Caution is still required, but this fix does allow updated plugin builds to avoid problems with some existing hosts * Change required C++ language standard from C++98 to C++11. This is because of the use of std::mutex in the above fix version 2.8: * Fix off-by-one rounding errors in frame-to-ns conversions. Unlike the other changes here which are invisible to plugin code, this change can lead to different results in the lowest significant figures from existing plugins if relinked against the newer code * Fix theoretical possibility of integer overflow in RealTime constructor * Fix use of undefined behaviour in PluginRateExtractor - Spec cleanuo