Tags: lucianodato/noise-repellent
Tags
Fixed ttls. Fixed issue in adaptive stereo version. Fixed issues for … …packagers. Updated libspecbleach (#95) * Chamge cpu again * Change flags again * testing with no c args * More changes * fix * Changing directories around * More fixes * Fix more * fix again * fix? * Fixing artifacts * Fix * doh * doh 2 * testing fix for artifacts again * Remove not used cache and mv fix * doh * doh 2 * doh3 * doh4 * Testing no compression * re enable mac ci * fix? * changing directories around * add mac intel builds * Add first test for windows * Testing fix * trying to saved bootstrapped env * Testing fix * fix? * fix? * Test fix * fix? * Testing fix * starting wine before building * Change wine package * Dependency fixes * Trying to fix wine32 dependencies * Testing not using the cache for the first build * Testing going back to wine stable packages * Restore correct package names for ubuntu distro * Fix wrong ttl and bin files names * Starting to re work old refactoring job * Finish connecting masking estimator class * Connecting transient detector * Work on correct masking initialization. Forcing clang formatter on files. * Connecting smoother and fixing include with guard for the common functions * Use longer and more descriptive names for functions * Exclude development branch from builds * Connecting plugin state save * Removing not used code and refactoring names appropriately * More cleanups * Rename class to more appropriate name * Getting rid of extra functions helper functions * Getting rid of macros and use libraries functions instead * Rename functions to align with the files name * Using object oriented design to make easier a c++ migration. Include all sources to be linked in compiling * Refactor plugin state to align with latest changes * Move bin info gathering to STFT * Initialize FFT configurations in initializer * Move initialization of FFT denoiser to plugin * Include noise profile in fft denoiser initialization. Reflect changes in plugin state * Fix bad naming convention. * More meaningful names * Fix all warnings. * Remove all comments. Use enums for windows types * Remove doxygen configuration * Passing object instead of having a private instance * More descriptive names * Added local environment configurations to debug and to test. Improve meson file * Find ttl files instead to consider different paths in other distros * Remove reset functions. Fixed clang and cppcheck linters warnings. Remove more comments * Fixes for a better markdown file * Removed wrong launch configuration. Added some convenient tasks * Use home user folder to install and uninstall * More work into tasks. Added a few handy ones. Removed tests from meson and added them as tasks * Refined tasks. Running lv2bench 10 times. Ignore valgrind dump files * Remove reset noise profile parameter. Fix many memory allocating issues and segfaults. Extract denoising parameters into a separate object. Many other fixes * Move plugin state management back to plugin.c. Fix segfault happening with plugin state. Added lv2apply to tasks. Added logging support for hosts that support it * Remove not needed macro and not needed includes * Simplify plugin parameters loading * Move bin info gathering functionality back to fft denoiser * Fix wrong variable being copied * Back to compiling with c99 standard * Fix unused warning and remove exposure of internal methods * Fixed meson install folder issue. Now libdir and prefix work properly. Simplify compiler configurations and move some to tasks. * Move around many functions and declarations for consistency * Added custom install folder for debugging * Forward declare internal functions. Explicit if conditions. Use curly braces always. Fix task problem matcher for gcc. Remove linters and enforce clang with default C/C++ vscode extension * Fix sord_validate reported issues * Embrace clang llvm and clang-tidy as linters and format * Fix all clang-tidy issues. Explicit ifs. Use const where possible. Remove exposed methods in headers that were not needed. * Getting rid of complicated encapsulations not needed. Using structs instead to be more c like. * Using lv2bench for valgrind and debug instead of lv2apply * Fix segmentation fault. Fix wrong task name * Embrace static when possible. Forward declare many methods in order to have better ordering of functions inside modules * Remove not needed vscode configurations * Expose enable port and make it host agnostic. Re add c++ extension options again * Enforce const as much as possible * Use std int instead of int * Cleaning the realtime audio path. Move modules compositions to initialize * Ensuring memory allocation on instantiate or log error. Consolidate style using pointers and initializations in plugin.c * Name change and better logging message * Reduce duplication by extracting common functionality into a utils module. Remove not needed parameter in STFT initialize. Explicictly declare warning level in tasks * Making sense of defines * Fix memory leak * Introduce a single entry point to the algorithm. First steps into creating a library out of the a dsp. Decouple some types from modules. Removed not needed states to store. * Starting to work on decoupling stft from the spectral denoising algorithm. * More renaming for consistency * Using generic handle instead of defined type for generic spectral processors being handled by the stft transform. Refactoring different structures out of the stft and the spectral processor * Remove duplicated functionality for windows. * More refactoring of stft to decouple buffer from fft * More renaming and some simplifications in modules * Separated library and plugin. Start using a meson subproject to build the library separated from the plugin. * Decouple noise estimation from denoiser. * Use pointers to the same singleton declared in entry source file for noise profile and denoise parameters. * Using pointer where possible intead of copying for noise profile and denoise parameters * Extract STFT windows to a different module. * Fix typo and memory leak * Fix issue where noise profile was smaller than needed and causes segfault. Separated noise profile in it's own module. Some renaming for consistency. Isolated power magnitude and phase specta calculation into another module that could be explanded. Moved feature calculation to each spectral processor * Extracted soft bypass and generalized it to be a signal crossfade and apply it in the time domain instead of the more expensive frequency domain. * Extract spectral whitening into a separate module * Make spectral features struct opaque * Separate denoise builder into a separate module. Some memory optimizations. Added some TODOs to resolve * Using int values in enable ttl port. Remove denoise_builder as it not necessary to have it in a separate module. Moved spectral feature calculation to gain estimation. Fixed a bunch of issues with gain estimation. More memory optimizations. Initialize spectrum multipliers to 1 instead of 0. * Moved some functions to spectral utils as they could be reused and are general enough * Extract spl conversion operations into its own module * Re add adaptive noise learn port. Amend library. Save noise profile size again and check restored size. * Re add louizou noise spectrum estimator * Starting to use a proper handle for public module and for some internar needed ones * Separate configurations from parameters in different header files * Some file renaming for consistency * Expose library's denoise parameters publicly and use them in the plugin. Using nice c99 struct initialization. Getting rid of not needed header * Fix badly allocated memory for parameters between plugin and library. * Remove not needed includes * Consistency across modules. Basic checking on real time safe functions. * Use unsigned int values where possible * Change copyright legend in all sources. Isolate fft transform in its own module and reuse it across code. Now it's easily replaceable. * Move all configurations into configurations header * Using unsigned int in for loops * Move noise estimation block to the spectral denoiser module for flexibility when adaptive denoising * Fix segfault cause by wrong initialization order. Added new config to ensure some windows have been averaged when capturing manually the noise profile * Add back lv2apply tasks * Update sord validate task to lv2 validarte. Graciously deallocate plugin when feature is missing * Small rename * Have the processing function and the spectral processor as parameters in the data struct in the stft. Don't expose the data struct of the denoiser * Making sense of folder structure * Add back reset profile option * Starting to use the activate and deactivate stages in the plugin. Remove deprecated reportsLatance port property in favor of latency designation. * Fix binary symbol exposure reported by lv2lint * Configure clang tidy with extension and fixed a bunch of newly reported issues. Some fixes and improvements in lv2ttl. Added some tasks for lv2bm testing. Moved louizou macros to config file. Fix bad initialization of samplerate in dbspl module. Removed clang-tidy file. * more linter fixes * Move feature calculation to spectral denoiser module to avoid calculating them twise * Basic work in public header to be used as a shared library. Renamed interfaces and struct appropriately. Generalized spectral feature with an options so it's easily switched if needed. * Revert setting pointer function to into run callback, otherwise it won't work with current architecture. Separate pipeline for adaptive and manual noise reduction. Other misc changes * Some minor fixes and extracting common utilities into a separate module * Move parameter loading to the callback function. Softbypass envelope based on buffer size. * Fix noise profile saving. Needed to comply with atom_Vector specification. Move noise profile available to where it belongs. * Fix old argument not reflecting latest changes * Added more reminders * Better struct declaration style. Some other minor changes. * Moved noise_profile to shared modules * Started to use C/C++ extension provided clang-tidy. Some fixes of the reported issues. Some work on the lib interface * Misc * Decouple buffer from stft processor. Some clang-tidy fixes. Use common handler for spectral processor (new interface available). Some renamings to make more sense. * Windowing optimization. Removed not needed parameter. Do not install static library. * Fixing some narrowing types. Simplify STFT code. * Separate adaptive denoiser as a separate spectral processor * Exctract common denoisers functionality to utils * Simplify Adaptive denoise processor. Remove dependency on noise profile and parameters * Extracted adaptive version to a separate plugin and separated concerns in the library. * Added a barebones stereo version of the adaptive plugin * Added stereo version of the manual plugin with support for different noise profiles for each channel * Replaced fixed fft size for a frame size in milliseconds so the same frequency resolution is preserved across sample rates. Separated implementation of interfaces for adaptive speech oriented and the general one. * Remove not needed methods in adaptive version * Extract noise profile to a separate module. Removed not used variable. * Move crossfade module to plugins. * Minor casting fix * Revert to one library header. Rename functions and variables accordingly. Fix mixing function not applying reduction weights to the complex side of the fft. Fix hanning window in stft. Calibrated stft configurations for Adaptive denoiser * Change parameter name to be more intuitive. Change block size to have more resolution in the low end to avoid ghost voices. * Fix spectrum alignment. Configure to have less latency. * Revert to original release curve. Working on Ardour again * Remove not needed extension methods for adaptive version * Change oversustraction factor to be a piecewise lineal function based on the a posteriori SNR like described in the paper. * Simplification of meson scripts install folder and enable c flags by default. Some renaming. Extracted bark mapping into its own module and make it more generic. Some other misc changes * Correctly free'd modules in masking module * Fixed critical band mapping for bark scale. Added snr calculations per bark band in adaptive denoiser. Function to return indexes per band. Amend other modules to reflect the modifications. Move some hardcoded values to the initialization functions. * Simplify processors to be just one module. Move all other modules to the shared folder. New oversustraction module to store all the different criterias of scaling the noise estimation. * Fixed memory leaks and critical bark bands once again. Use masking thresholds for adaptive denoising. Some other misc changes. * Work on consolidating names. Separated gain estimation functions into its own module * Change spl converter module to directly contain the absolute thresholds calculation and act as a processor instead * Remove masking parameters hardcoding. Re-add general spectral sustraction rule. Misc changes. * Fix mem leaks and added some functions to prepare zeropadding * Rework a bit configurations. Added some padding strategies * Fixed critical bands and added some more options. Use LUTs instead of formulas * Prepare library code to be in a separate repository. Center fft window when using padding * Prepare to re add postfilter postprocessing * Added postfilter to reduce musical noise during low snr parts of the signal. Fixed A posteriori oversustraction criteria. Remove masking parameter as is not needed just using the rescale instead. Consolidate oversubtraction parameters to be common to all criterias * Refactor alpha and betas to be general per noise scaling criteria. * Fixed wrong function signatures * Switchable gain estimation criterias * Misc changes. * Change URID for general plugin to avoid collision with old plugin. Zero padding amount per plugin. Fix noise profile learning. Move detection of noise profile availability to noise profile module. Other minor changes * Extracted denoising mixing into a separate module. Starting work to fix noise profile saving fix. * Remove vscode specifics * Fix noise profile storing with the session * Fixed Noise scaling by masking thresholds. Starting to work on time smoother * Finalizing smoothing algorithm. Transient aware smoothing added and made default. * Re expose transient protection but as a switch * Use percentage of smoothing instead of a release envelope. Use parameters for the smoothing module * Removed port properties to hide elements in gui as it was breaking reaper * Added smoothing to the adaptive plugin * Start using specbleach own repository as wrap * Amend README based on latest changes * Fix not included dependecy * GitHub workflow pawpaw (#88) Added github workflow that build binaries for all operating systems using paw paw https://github.com/DISTRHO/PawPaw * Use pawpaw version that is going to work and not recache packages again * Fix build * test newer version again * Work on README * Start to use specific denoiser headers * Update license headers * Remove old videos from README * Use badges with specific branches * Revert last change * Correcting wording * Remove cflag hardcoded in meson build file. Rely on Pawpaw configured flags * Opps! * Enable optimization flags on build pipeline * Don't optimize in older macs * Fix issues reported by lv2lint. Hide symbols in build shared libraries * No need to add flags that Pawpaw is already adding * Test fix for linux build * Revert to not passing argument * Use newer meson version as local * Remove not needed line in meson build file. Install meson from pip in ubuntu image in github actions. Reduce range of reduction in manual plugin to correspond with documentation * Fix wrong line break * Install ninja from repos * Set libspecbleach to use the released 0.1.0 version * Fix typo in ttl file. Work with libspecbleach as a shared dependency by default and fallback to subproject. Fix adaptive stereo version working with only one library instance * Fixed parameter issues in ttl * Switch to new libspecbleach version * Set version to 0.2.1 and update version from libspecbleach * Use global headers instead subproject ones * Update libspecbleach once again * Correction in readme * Use right cache version in win64 builds
Refinements before release (#91) * fix * no need to rename zip file * testing not using shell scripts added macos * fix typo * typo again * remove scripts and previous ci yml files * Testing macos fixes * remove explicit arch for macos * Test with cpu type * Chamge cpu again * Change flags again * testing with no c args * More changes * fix * Changing directories around * More fixes * Fix more * fix again * fix? * Fixing artifacts * Fix * doh * doh 2 * testing fix for artifacts again * Remove not used cache and mv fix * doh * doh 2 * doh3 * doh4 * Testing no compression * re enable mac ci * fix? * changing directories around * add mac intel builds * Add first test for windows * Testing fix * trying to saved bootstrapped env * Testing fix * fix? * fix? * Test fix * fix? * Testing fix * starting wine before building * Change wine package * Dependency fixes * Trying to fix wine32 dependencies * Testing not using the cache for the first build * Testing going back to wine stable packages * Restore correct package names for ubuntu distro * Fix wrong ttl and bin files names * Starting to re work old refactoring job * Finish connecting masking estimator class * Connecting transient detector * Work on correct masking initialization. Forcing clang formatter on files. * Connecting smoother and fixing include with guard for the common functions * Use longer and more descriptive names for functions * Exclude development branch from builds * Connecting plugin state save * Removing not used code and refactoring names appropriately * More cleanups * Rename class to more appropriate name * Getting rid of extra functions helper functions * Getting rid of macros and use libraries functions instead * Rename functions to align with the files name * Using object oriented design to make easier a c++ migration. Include all sources to be linked in compiling * Refactor plugin state to align with latest changes * Move bin info gathering to STFT * Initialize FFT configurations in initializer * Move initialization of FFT denoiser to plugin * Include noise profile in fft denoiser initialization. Reflect changes in plugin state * Fix bad naming convention. * More meaningful names * Fix all warnings. * Remove all comments. Use enums for windows types * Remove doxygen configuration * Passing object instead of having a private instance * More descriptive names * Added local environment configurations to debug and to test. Improve meson file * Find ttl files instead to consider different paths in other distros * Remove reset functions. Fixed clang and cppcheck linters warnings. Remove more comments * Fixes for a better markdown file * Removed wrong launch configuration. Added some convenient tasks * Use home user folder to install and uninstall * More work into tasks. Added a few handy ones. Removed tests from meson and added them as tasks * Refined tasks. Running lv2bench 10 times. Ignore valgrind dump files * Remove reset noise profile parameter. Fix many memory allocating issues and segfaults. Extract denoising parameters into a separate object. Many other fixes * Move plugin state management back to plugin.c. Fix segfault happening with plugin state. Added lv2apply to tasks. Added logging support for hosts that support it * Remove not needed macro and not needed includes * Simplify plugin parameters loading * Move bin info gathering functionality back to fft denoiser * Fix wrong variable being copied * Back to compiling with c99 standard * Fix unused warning and remove exposure of internal methods * Fixed meson install folder issue. Now libdir and prefix work properly. Simplify compiler configurations and move some to tasks. * Move around many functions and declarations for consistency * Added custom install folder for debugging * Forward declare internal functions. Explicit if conditions. Use curly braces always. Fix task problem matcher for gcc. Remove linters and enforce clang with default C/C++ vscode extension * Fix sord_validate reported issues * Embrace clang llvm and clang-tidy as linters and format * Fix all clang-tidy issues. Explicit ifs. Use const where possible. Remove exposed methods in headers that were not needed. * Getting rid of complicated encapsulations not needed. Using structs instead to be more c like. * Using lv2bench for valgrind and debug instead of lv2apply * Fix segmentation fault. Fix wrong task name * Embrace static when possible. Forward declare many methods in order to have better ordering of functions inside modules * Remove not needed vscode configurations * Expose enable port and make it host agnostic. Re add c++ extension options again * Enforce const as much as possible * Use std int instead of int * Cleaning the realtime audio path. Move modules compositions to initialize * Ensuring memory allocation on instantiate or log error. Consolidate style using pointers and initializations in plugin.c * Name change and better logging message * Reduce duplication by extracting common functionality into a utils module. Remove not needed parameter in STFT initialize. Explicictly declare warning level in tasks * Making sense of defines * Fix memory leak * Introduce a single entry point to the algorithm. First steps into creating a library out of the a dsp. Decouple some types from modules. Removed not needed states to store. * Starting to work on decoupling stft from the spectral denoising algorithm. * More renaming for consistency * Using generic handle instead of defined type for generic spectral processors being handled by the stft transform. Refactoring different structures out of the stft and the spectral processor * Remove duplicated functionality for windows. * More refactoring of stft to decouple buffer from fft * More renaming and some simplifications in modules * Separated library and plugin. Start using a meson subproject to build the library separated from the plugin. * Decouple noise estimation from denoiser. * Use pointers to the same singleton declared in entry source file for noise profile and denoise parameters. * Using pointer where possible intead of copying for noise profile and denoise parameters * Extract STFT windows to a different module. * Fix typo and memory leak * Fix issue where noise profile was smaller than needed and causes segfault. Separated noise profile in it's own module. Some renaming for consistency. Isolated power magnitude and phase specta calculation into another module that could be explanded. Moved feature calculation to each spectral processor * Extracted soft bypass and generalized it to be a signal crossfade and apply it in the time domain instead of the more expensive frequency domain. * Extract spectral whitening into a separate module * Make spectral features struct opaque * Separate denoise builder into a separate module. Some memory optimizations. Added some TODOs to resolve * Using int values in enable ttl port. Remove denoise_builder as it not necessary to have it in a separate module. Moved spectral feature calculation to gain estimation. Fixed a bunch of issues with gain estimation. More memory optimizations. Initialize spectrum multipliers to 1 instead of 0. * Moved some functions to spectral utils as they could be reused and are general enough * Extract spl conversion operations into its own module * Re add adaptive noise learn port. Amend library. Save noise profile size again and check restored size. * Re add louizou noise spectrum estimator * Starting to use a proper handle for public module and for some internar needed ones * Separate configurations from parameters in different header files * Some file renaming for consistency * Expose library's denoise parameters publicly and use them in the plugin. Using nice c99 struct initialization. Getting rid of not needed header * Fix badly allocated memory for parameters between plugin and library. * Remove not needed includes * Consistency across modules. Basic checking on real time safe functions. * Use unsigned int values where possible * Change copyright legend in all sources. Isolate fft transform in its own module and reuse it across code. Now it's easily replaceable. * Move all configurations into configurations header * Using unsigned int in for loops * Move noise estimation block to the spectral denoiser module for flexibility when adaptive denoising * Fix segfault cause by wrong initialization order. Added new config to ensure some windows have been averaged when capturing manually the noise profile * Add back lv2apply tasks * Update sord validate task to lv2 validarte. Graciously deallocate plugin when feature is missing * Small rename * Have the processing function and the spectral processor as parameters in the data struct in the stft. Don't expose the data struct of the denoiser * Making sense of folder structure * Add back reset profile option * Starting to use the activate and deactivate stages in the plugin. Remove deprecated reportsLatance port property in favor of latency designation. * Fix binary symbol exposure reported by lv2lint * Configure clang tidy with extension and fixed a bunch of newly reported issues. Some fixes and improvements in lv2ttl. Added some tasks for lv2bm testing. Moved louizou macros to config file. Fix bad initialization of samplerate in dbspl module. Removed clang-tidy file. * more linter fixes * Move feature calculation to spectral denoiser module to avoid calculating them twise * Basic work in public header to be used as a shared library. Renamed interfaces and struct appropriately. Generalized spectral feature with an options so it's easily switched if needed. * Revert setting pointer function to into run callback, otherwise it won't work with current architecture. Separate pipeline for adaptive and manual noise reduction. Other misc changes * Some minor fixes and extracting common utilities into a separate module * Move parameter loading to the callback function. Softbypass envelope based on buffer size. * Fix noise profile saving. Needed to comply with atom_Vector specification. Move noise profile available to where it belongs. * Fix old argument not reflecting latest changes * Added more reminders * Better struct declaration style. Some other minor changes. * Moved noise_profile to shared modules * Started to use C/C++ extension provided clang-tidy. Some fixes of the reported issues. Some work on the lib interface * Misc * Decouple buffer from stft processor. Some clang-tidy fixes. Use common handler for spectral processor (new interface available). Some renamings to make more sense. * Windowing optimization. Removed not needed parameter. Do not install static library. * Fixing some narrowing types. Simplify STFT code. * Separate adaptive denoiser as a separate spectral processor * Exctract common denoisers functionality to utils * Simplify Adaptive denoise processor. Remove dependency on noise profile and parameters * Extracted adaptive version to a separate plugin and separated concerns in the library. * Added a barebones stereo version of the adaptive plugin * Added stereo version of the manual plugin with support for different noise profiles for each channel * Replaced fixed fft size for a frame size in milliseconds so the same frequency resolution is preserved across sample rates. Separated implementation of interfaces for adaptive speech oriented and the general one. * Remove not needed methods in adaptive version * Extract noise profile to a separate module. Removed not used variable. * Move crossfade module to plugins. * Minor casting fix * Revert to one library header. Rename functions and variables accordingly. Fix mixing function not applying reduction weights to the complex side of the fft. Fix hanning window in stft. Calibrated stft configurations for Adaptive denoiser * Change parameter name to be more intuitive. Change block size to have more resolution in the low end to avoid ghost voices. * Fix spectrum alignment. Configure to have less latency. * Revert to original release curve. Working on Ardour again * Remove not needed extension methods for adaptive version * Change oversustraction factor to be a piecewise lineal function based on the a posteriori SNR like described in the paper. * Simplification of meson scripts install folder and enable c flags by default. Some renaming. Extracted bark mapping into its own module and make it more generic. Some other misc changes * Correctly free'd modules in masking module * Fixed critical band mapping for bark scale. Added snr calculations per bark band in adaptive denoiser. Function to return indexes per band. Amend other modules to reflect the modifications. Move some hardcoded values to the initialization functions. * Simplify processors to be just one module. Move all other modules to the shared folder. New oversustraction module to store all the different criterias of scaling the noise estimation. * Fixed memory leaks and critical bark bands once again. Use masking thresholds for adaptive denoising. Some other misc changes. * Work on consolidating names. Separated gain estimation functions into its own module * Change spl converter module to directly contain the absolute thresholds calculation and act as a processor instead * Remove masking parameters hardcoding. Re-add general spectral sustraction rule. Misc changes. * Fix mem leaks and added some functions to prepare zeropadding * Rework a bit configurations. Added some padding strategies * Fixed critical bands and added some more options. Use LUTs instead of formulas * Prepare library code to be in a separate repository. Center fft window when using padding * Prepare to re add postfilter postprocessing * Added postfilter to reduce musical noise during low snr parts of the signal. Fixed A posteriori oversustraction criteria. Remove masking parameter as is not needed just using the rescale instead. Consolidate oversubtraction parameters to be common to all criterias * Refactor alpha and betas to be general per noise scaling criteria. * Fixed wrong function signatures * Switchable gain estimation criterias * Misc changes. * Change URID for general plugin to avoid collision with old plugin. Zero padding amount per plugin. Fix noise profile learning. Move detection of noise profile availability to noise profile module. Other minor changes * Extracted denoising mixing into a separate module. Starting work to fix noise profile saving fix. * Remove vscode specifics * Fix noise profile storing with the session * Fixed Noise scaling by masking thresholds. Starting to work on time smoother * Finalizing smoothing algorithm. Transient aware smoothing added and made default. * Re expose transient protection but as a switch * Use percentage of smoothing instead of a release envelope. Use parameters for the smoothing module * Removed port properties to hide elements in gui as it was breaking reaper * Added smoothing to the adaptive plugin * Start using specbleach own repository as wrap * Amend README based on latest changes * Fix not included dependecy * GitHub workflow pawpaw (#88) Added github workflow that build binaries for all operating systems using paw paw https://github.com/DISTRHO/PawPaw * Use pawpaw version that is going to work and not recache packages again * Fix build * test newer version again * Work on README * Start to use specific denoiser headers * Update license headers * Remove old videos from README * Use badges with specific branches * Revert last change * Correcting wording * Remove cflag hardcoded in meson build file. Rely on Pawpaw configured flags * Opps! * Enable optimization flags on build pipeline * Don't optimize in older macs * Fix issues reported by lv2lint. Hide symbols in build shared libraries * No need to add flags that Pawpaw is already adding * Test fix for linux build * Revert to not passing argument * Use newer meson version as local * Remove not needed line in meson build file. Install meson from pip in ubuntu image in github actions. Reduce range of reduction in manual plugin to correspond with documentation * Fix wrong line break * Install ninja from repos * Set libspecbleach to use the released 0.1.0 version