File KnobKraft-orm.changes of Package KnobKraft-orm
------------------------------------------------------------------- Sat Dec 06 08:29:27 UTC 2025 - Edgar Aichinger <edogawa@aon.at> - Update to version 2.8.2+git.14.e739382: * Fix Python on Mac for local builds without hopefully destroying the code signing in CI * Rollback MacOS codesigning changes. This is highly fragile. * Addressing #136: Generic timeout mechanism by forwarding empty MIDI messages into the callback handler when a timeout has passed. By default, the timeout is 30 seconds in order not to break any existing slow synths, but shorter timeouts can be reported by the adaptations via the messageTimings function. Adaptation Programming Guide was adapted. * Third attempt to fix MacOS codesigning again * Second try to fix the MacOS codesigning which broke earlier. * Nicer vertical alignment * Two column layout for KeyboardMacroView * Add in the Macro definition of the Regular flag that we added lately. This also required setting up a scrollbar because the list of Macros now is getting too long * More improvements for #408: - Macro detection now edge-triggered: The-Orm/KeyboardMacroView.cpp tracks active macro states and only fires when a macro chord transitions from not-matched to matched, preventing repeated triggers while holding the chord and playing/releasing other notes. - All Notes Off (CC123) is now respected: incoming CC123 clears the internal MidiKeyboardState and resets macro armed states while still forwarding the message to the synth, so held external notes no longer block shortcut detection. * Stop hiding toggled patches because it confuses navigation with the macros keys. This should now only toggle the icon on the button but not refresh the list and make the patch go away. * Address #408: Macro keyboard does not properly allow page transitions forward and backward. * Fix MacOS Python framework bundling * Adressing #499 - reloading categories when switching between two different databases with two different category sets. * Fix dev version on MacOS not having Python ready. * Add release notes for hotfix 2.8.2 * Rollback full import creation - this needs more thought in how the UI should react to a patch being in multiple imports, and in which order the imports should show * Fix sorting and filtering, allow sorting Imports * Bug fixes - failure to query patches came due to column name suddenly being ambiguous. Duplication happened because updatePatch would rewrite the import list - previously, this was a column, now it is a row, which led to the duplication. * Release notes for hotfix release * Fixing migration problem to schema version 17 caused by EditBufferImports not being unique * Adapting migration test to the problem of duplicate import.id causing the migration to schema 17 to fail. * #497: Sort synths alphabetically so we have a stable sort order * Fix nullptr crash * Starting the release notes for the 2.8.0 release coming up * Remove duplicated code * Fix wrong submodule commit, and Linux warning on fallthrough * Add fixes for Yamaha DX-7II as reported and provided by @AgtSlick * Use merged MidiKraft branch * Remove unused function to get Mac build to pass * Improving drag and drop operations now that Imports are lists themselves it is easier to integrate them. * Dropping the old sourceID column which has now been replaced by the import list * Removing the sourceID mapping now is possible without feature loss * nitpick * Removing the use of sourceID column * Added cleaned MidiKraft * Prevent crash on null pointer * one more review comment * Bring the migration test closer to the real implemented database structure * Align test with new import ID scheme * One more schema fix * Address the EditBufferImport not unique ID problem * Adding doctest properly as a submodule * Include speedup for list querying * Accelerate list loading by preventing redundant change messages getting sent out when actually the synth wasn't changed. * Added a dedicated step after the main build to build and execute patch_database_migration_test.exe, so the migration doctest suite runs as part of the Windows pipeline (.github/workflows/builds-windows.yml:57). * Extend the migration tests to cover all scenarios we had in our plan. * Add test for the migration code * Remove the debug output of the new count method because we don't have it in this branch * Use correct MidiKraft version * Adding docs created during the large refactoring, and a new python doctest that tests a few SQL statements together with a recreated v14 database * Improve logging so we get a log even if the program exits immediately. * Add debug output of new counting function * Do create the banks after import! * Merge the two code paths for import bank via dialog and import via clicking on the tree item to sync the active bank * Review feedback * More simplificationof AI code * Simplify the AI code written * Use merged MidiKraft lib * Fix search query construction for regular field * New feature: we add a fourth state - we had fav, hidden, and neither (undecided), but we want to track that a patch has been revisited and is not fav. Calling this regular. * Preparing for release 2.7.2 * Update adaptation implementation overview * Switch the Behringer Pro 800 program dump back to the normal, "non CC flood" behavior * New feature #468: Add a secondary MIDI output that mirrors all MIDI messages KnobKraft sends * Fix Pro 800 renaming to keep garbage after new name, which allows for passing the rename tests * Address delete bug throwing a SQL exception on ambiguous column name.(#474) * Update to latest MidiKraft * Allow patch names longer than 20 characters, increased to 50 (#482) * Add a new fill mode to create new bank - start from the active patch (#484) * Also fix the persisted fixed synth played setting * Fix #485 - MIDI Input Device was persisted as index, not as name. * [github] MacOS-13 is being retired. Try our Mac build on MacOS-14, which is an ARM64 image. * [Pro-800] Introduce handling the different versions of the patch format in the nameFromDump() and renamePatch() methods. * Add a cc flood version of the convertToProgramDump - actually this is more a convertToEditBuffer function, but to hook it up it needs to be convertToProgramDump right now. * Trying to implement the rename patch function for the Pro-800 * New MidiKraft library * New version 0.92 of the Ensoniq SQ-1 adaptation, thanks to @mostelin! * Add missing submodule commit * Use th new renameList function instead of the special renameImport function, which we no longer need * Use the migrated imports data for filling the grid, and imports now are lists for the database. ------------------------------------------------------------------- Wed Oct 01 14:12:23 UTC 2025 - Edgar Aichinger <edogawa@aon.at> - removed obsolete update-dtl.patch - Update to version 2.7.1+git.3.891e72e: * Fix crash when hitting a too new database schema * Reuse existing sysex decoder * Experimental implementation of the Behringer Pro-800 * Added feature entry for the new 2.7.1 release * Fix problems with MacOS not displaying native file chooser dialog. * Pretty up the EditCategory dialog to include the sort order column. * Add release notes for upcoming 2.7.1 * Revert wrong change of version.cpp.in - it is actually an include file wrongly called cpp * Fix file that was never compiled before * Fix Linux * Make changes to CMake files suggested by ChatGPT to address issue #459 * Updated * Added a new column to the table * Have the AI implement the rename function for renamePatch. This worked as a one shot query! This addresses #450. * Retrofitting the release notes for 2.6.4 I forgot * Fixing crash happening only in release build * Fix regression in version number capturing from git because CMake changed its cmake -P behavior * Adding release notes for 2.7.0 release, and back adding those for 2.6.4 which was missing. * Trying to address #463, make the highlighted patch easier visible * One more editing focus fix * Addressing #461, the focus and caret/selection lost when alt tabbing away. This is a complex problem, but GPT5 was able to come up with nice solutions. If this works, we can also enable this for more text editor fields * Rolling back implicit Python installation finding, because that broke the Mac build. Python version selection is a very very fragile thing when done cross platform * Starting with python 3.12, or maybe because of switching the version, we need to copy the embedded runtime not only into the installer but also into the build output directory, because otherwise Visual Studio might launch Python with the wrong runtime causing weird access violation errors and other things. * Downgrade pybind11 from 3.0.0 to last known 2.x version again, as the upgrade failed (maybe not due to pybind11 alone, but due to the changes in how to find python) * Fix error in CMake file that caused CMake warning * Set CMake policy required by newer CMake versions for FetchContent * Update SQLiteCpp to 3.3.3 * Need to add new JUCE module juce_javascript * Update JUCE to version 8.0.8 * Update to pybind11 v3.0.0 * Switch to newer Visual Studio 2022 as 2019 is no longer supported by Github * Update embedded Python to the last full build provided for Python 3.12, which is Python 3.12.10. Need to update to Python 3.13 soon * Update Windows Runner to 2022, we were still using 2019 * Add the three new adaptations to the installer package and the README * Reduce log level to debug for the new logs in the GenericBankDumpCapability * adding windows dirs for build & builds to gitignore * Working microKORG S adaptation: patch extraction, correct MIDI bank select, and accurate program change recall. * Working microKORG S adaptation: patch extraction, correct MIDI bank select, and accurate program change recall * Added Korg_microKORG S adaptation, updated CMake and enhanced bank dump capability ------------------------------------------------------------------- Sun Aug 03 07:01:12 UTC 2025 - Edgar Aichinger <edogawa@aon.at> - Update to version 2.6.4+git.6.560782b: * Add adaptation for Behringer Wave * Novation Bass Station II Adaption * Update README.md * Trying to fix missing bundle identifier in MacOS build which might cause problems with sandboxed applications * Fix bug in Analog Rytm preventing it to send Sounds * Prepare release notes for the 2.6.3 * More explicit setup for gcc-14 * Fix CMakeLists.txt again * Try using gcc-14 on Ubuntu 24 * Update pybind11 to 2.13.6 adressing #452 * Update DTL to get the fix for gcc 14 mentioned in #453 * Remove the retry bank download altogether in order to get progress with the K5000 * Trying to address #455, Chatty says I misunderstood the syntax of the CMake rule here * Remove new retry code again * Another fix for Mac * Disable the tests for the TX81Z, because we are missing test data. * Prepare the 2.6.2 release * Added category imports for the Roland XV-3080, as tested by @summersetter. * Adding in the Yamaha TX81Z adaptation contributed by @summersetter. Thanks for that! * Try a falthrough to silence the Mac error * Adding a tool to produce a syx file from the MIDI log * Adjust the test data so the new test routine that simulates loading whole sysex files via the Librarian passes. For this, we need to specify how many patches we expect as an outcome of the load_sysex call. * Implement the load_sysex logic for the Librarian. This is good enough that we can turn on a test that the load_sysex function is able to drive the loading of the given sysex files for all adaptation. This will need some update for the test data, though. * Implement the LIbrarian in Python, mostly for documentation, but we can also use this to run a simulated conversation with a synth and the adaptation. Can be useful for testing as well! * Fix for Mac * Fixed Linux and Mac build * Trying to fix the retry bank download problem occuring with the K5000 because the messages are large and there are partial sysex messages incoming. * Added clear button to the KeyboardMacroView. That was more painful than expected, but now we don't need the additional thread, looks better. Also, evaluate macros only on key down and not key up to address #434. * Revert "Bank management and sending from kk to synth currently NOT working" * Bugfix and save feature * Bank management and sending from kk to synth currently NOT working * First implementation of the new bank export: For the Yamaha DX7 * Fix * Adjust release ntoes * Adding the Generic variant of this capability. Lot's of boilerplate code. * Start preparing a new way for exporting banks instead of individual dumps * Fix comma error preventing new database from working (#430) * Mac fix * Update release notes * Complete release notes for 2.6.0 * Merge error? * More release notes * Include updated MidiKraft library. * Fix Korg M1 autodetection * More release notes * Handle upgrade case better, when the default database is too new. * Marking Waldorf Pulse as working * Increase the number of banks for the Sequential Prophet-X to 12 from 8, allowing to access the 4 new addons banks. * Prepare Yamaha FS1R release * Fixes * Switch to main of MidiKraft * Start writing release notes for 2.6.0 * A little more C++y * Remove dead code * Show favorite and hidden properly also in the right hand side patch buttons. * Fix deleting patches via the menu under the complexity of synth banks and user lists. Addresses #386 * Clarified text which is not really an upload * Fix Generic Roland tag detection, thanks to @ivruebj for finding it! * Fix bug: The Patch Button at the top of the CurrentPatch display was not changing Colour anymore. * Disable long click,as it doesn't work yet. * Include corrected comment * Fix bug: The Patch Button at the top of the CurrentPatch display was not changing Colour anymore. * Allow right click to reset the other buttons of the same button group. This is useful both for filtering and for changing the category of a patch. I should make this default!? * Prepare handling right clicks and long clicks * Build the simplest possible display switch to put the author as subtitle. This could be extended into a more general info layout to put various properties into the 4 corners and the center. Also for #383. * Add Author and Info fields in the database and property editor. Also add it to the text search. Works, addresses part of #383 * Remove the save button and the dirty list from the user bank feature. This is alien to an application that is database driven, and users complain you can lose data, which I agree. Addresses #389 * Add a tool to convert the MIDI log into a syx file, thanks chatty! * Make a version which has no global state. This was reported to work * Add release notes for 2.5.3 * Include two bugfixes in the LIbrarian * Address the Waldorf Pulse problem of bank sizes (#419) * Get away without using mido for testing * More fixes and checks for the FS1R * Fix for Mac * Fix for Mac * Add release notes for 2.5.2 release * Need the fix to have larger banks * Need the fix to have larger banks * Extend the Yamaha FS1R to deal with voices as part of the patch * [Yamaha FS1R] Implement convertToEditBuffer as well * Next round of implementation - trying to do it the same way we did it for the MKS-70 and the MKS-80, by including the voices with the performances. * Oberheim OB-X8 fix: name is actually 20 characters * Oberheim OB-X8 fix: name is actually 20 characters * Fix the name length bug caused by misunderstanding of ljust in more places * Revert "Add another validation loop to prevent we are working with wrong data. The Roland format is really horribly complex." * Remove unused test * Add another validation loop to prevent we are working with wrong data. The Roland format is really horribly complex. * Fix the D50 renaming - it did not shorten the name, just pad it if it was too short. Add one more testbank, and two additional guards against programming errors. * Fix the extract all patches from bank test, that never worked. Add an option to specify that the bank is edit buffers, not program buffers. * Fix - the Oberheim OB-X8 module has again its own sysex ID screwing up auto detection. * Mark the Trigon-6 with "works" after feedback that it does. * Fix typo * Roland D-50: One more test * Yamaha FS1R: Slower autodetection, preparation for parameter request message, and test for autodetection * Fixing Yamaha FS1R after testing with the MID files downloaded * Remove the Yamaha FS1R try using the Generic Roland module - that was the wrong idea. * Yamaha FS1R - implement rename and convert to edit buffer * Implemented the missing D50 bank import function. Curious to see if all bank files have 136 messages? * Test fix? * New approach for the Yamaha FS1R - forget the Roland module, the address space works completely differently, more like message type triples. This version does treat performance and voice as two separate data types, and allows to import these from separate banks. * Fix missing import by copying locally * Some initial experiments with the Yamaha FS1R - it might be simpler than we thought. * Roland D-50: Overhaul the implementation, it needed quite a few fixes to the GenericRoland module. This was copied into the D50 file for now so the user can test it without getting a new build. * Update status for Waldorf M, and add bank select. * Just typos * Waldorf M: Add a test for friendly bank name * A few more fixes for the Waldorf M * Don't throw Exceptions as strings * Waldorf M: Do try to use the correct device ID when requesting programs. * Analog Rytm: Do not try to extract the name from non-patterns or sounds * Adding issue number to README, #247 * Adding the Juno DS adaptation contributed by @mslinn, as copied from https://github.com/mslinn/knobkraft_adaptation_juno-ds * New version of the Oberheim OB-X8, now based on the GenericSequential module which had to be extended a bit to support this synth. * Preparing the next release with a lot of new adaptations. * #388 - switching the database would not clear the synth bank panel or the recent patches list. * Adding the Oberheim Teo-5, which is under the hood a Sequential software-wise. * Fix Trigon-6, not sure why this was commited as is * Fix Analog Rytm, need to use correct name of interface function * Adding the Sequential Trigon-6 * Waldorf Pulse, a new attempt. The German manual states the difference between the bulk and the non-bulk thing is just that the bulk is stored (program dump) while the other is not (edit buffer). * Adding Elektron Digitone and Analog Rytm to the README file * And now adding the Elektron Digitone as well. This is so similar to the Analog Rytm that we should do a generic Elektron module. * Adding the Elektron AnalogRytm * Moog Voyager adaptation * Forgot to define the 16 banks * [Waldorf M] Very quick shot at Waldorf M * [Korg M1] First try of a Korg M1 adaptation * [Akai AX80] New device which has only a Tauntek firmware. Can't do much, but you'd be able to send from the KnobKraft into the synth I think * Trying to fix CMake file for Ubuntu. Why did this work on Windows? It is not specified. * Try to update cmake on Ubuntu as it might be outdated? * Adding release notes for 2.5.0 * Adding alpha version of Roland TD-07 adaptation * Add missing test data file for E-mu Morpheus * Adding Line 6 POD Series adaptation * Add new synths to README and installer * Cherry pick Python 3.12 unused op suppression * [Morpheus] fingerprinting * [Morpheus] Renaming for the Morpheus * Cherry pick Python 3.12 unused op suppression * Quick shot at an implementation for the E-mu Morpheus. Looks pretty straightforward. This is Presets only, the synth has a ton of different data types. * Add IF(APPLE) * Try to get rid of the include and bin files in the DMG put there by this? * remove unused variable * Fixes * Upload the DMG even when notarization fails * More subtle changes * Fix auto lock timeout of keychain unlock operation - the default is set to 5 minutes. This is hilariously under documented. * Revert "Test using only the TeamID" * Test using only the TeamID * Build didn't run * Run build * Fix export * Some more links * Get closer to the real world * Worked, then just use the same mechanism for the long running build * Test keychain with password * Force push * Why does it hang again? * Integrating the successful test into the main pipeline * Next test * What is the keychain? * YAML shit again * A bit of testing * We need to unlock the keychain, else we cannot access it * Try something else * Revert "Spooky quoting" * Spooky quoting * Fix * Missing quoting * Fix * Trying to repair the developer ID variable * Use the number of hardware CPUs for the parallelization * Try to accelerate the MacOS build on the Github runner * What is different on the build machine? * Setup version detection also from the Makefile * Prepare running the code signing and notarization on Github * Parameterize the codesign team ID * Remove one extra line of search paths from the trials earlier * Remove not required find_package for Python - we use pybind11 solely for this. * Keep link target which is required * Fix the distributed files so the notary action works again. * Add copying the Python library files via rsync as the embedding target does not copy the lib folder * Slight improvement for the makefile and another debug command * Enable Python from within the bundle directory, instead of the system's install. And sign all those dynlibs again because else it will crash. * Add comment * Notarization and stapling works, verification is successful. Now to the question how to do all of this in github? * Adding more steps for notarization and verifying the notarization steps. * Add a Makefile target to sign the DMG as well, not just the files inside the DMG * Add a Makefile shortening the typing of all the mount, unmount, and verify commands. * One step further - the code signing in the install code section is wrong, because it is run on the wrong files. So the DMG included files will never be signed. Extract the signing and bundle code into its own post-build step, which makes more sense and is more verbose. * Setup code signing with an additional execute process step * Fixup bundle works without runtime * Fix copyright * Remove runtime * Add Python as a Framework * One problem less - do not copy the pycache files around, they might also interfere with binary compatibilty as they are compiled Python files. * The global compiler switch seems to have done its job, then we can remove all the unnecessary cruft again. * Work around the hdiutil command failing with resource busy by just retrying it 10 times... * Set a compiler switch to remove the visibility warnings? * Fix typo in variable name * Address #375, a one off error as the Novation Summit has 1-based bank numbers in sysex * /&ยง%$(% YAML * Keep DMG for testing * Sigh * Remove the icu4c library as it is conflicting with the one from MacPorts. * Switch order of linker libraries, to hope that ld will not find the wrong library. Else I need to remove it from the brew directory? * Add a comment if UNIVERSAL_BINARY is off, where to setup the ICU path. * Remove file not needed as we have a working github version now. * Test asking macPort for the universal binary * Fix compiler error on MacOS * Remove long outdated splash screen define * yaml /&$2434 * Use MacPorts universal binary of ICU instead of homebrew, which does not support universal binaries * Move all builds into one .builds directory to cleanup the .gitignore * New version 0.92 of the Ensoniq SQ-1 adaptation, thanks to @mostelin! * Enforce only family 0x0000 * Return -1 if not SYSEX message. * Clarifications for other POD devices. * Line 6 POD Series Adaptation * One more ignore, need to cleanup * Update to C++ 20 for gin, and fix formatting issues with the new fmt library. * Update fmt to 11.0.2 * Update Gin * Update nlohmann json to 3.11.3 * Use json schema validator 2.3.0 * Set JUCE release to 8.0.3 * Update pybind to 2.13.1 * Update sentry-native to 0.7.12 * Update spdlog to 1.15.0 * Update to SQLiteCPP to release 3.2.0 * Fix Windows build * Fix Kijimi test * Fix export on unknown MIDI channel as well for the program dump export. * Make sure to store unknown bank as NULL in the database, and not as 0. And make sure to not call numberFromDump() for non single program dumps. * Need this Midikraft change as well * Try to address #343 - when programNumber returned 0 for an edit buffer, all patches were listed as 0 instead of the system resorting to counting the edit buffers. * Fix - in case the synth is not detected, don't try to use channel -1 but rather default to 0 and issue a warning. This was triggered by bank export. * [JUCE] Replace deprecated juce::Font::getStringWidth calls with recommended juce::TextLayout::getStringWidth calls in order to fix mac builds * [Linux] Avoid name collision of Ubuntu 22 and Ubuntu 24 versions. * [JUCE] Update juce-widgets to match JUCE 8 * [Mac] Bump version minimum to 10.11, as JUCE does not seem to support older versions anymore. * Trying to fix Ubuntu 22, and upload artifacts of a successful build for testing * [Linux] Change the names of the two Ubuntu build machines * [Linux] Update to webgtk 4.1 for Ubuntu 24.04 * [JUCE] Jump to Juce 8 from Juce 7. Add an Ubuntu 24.04 build machine to see how far we can get. * [macOS] seems this is not the official way to install MacPorts * [macOS] Trying a different way to bootstrap MacPorts on the runner * [.ck yml] * [mac M1] The joy and fun - to build Universal binaries our binary input needs to be available in universal2 format as well. Switch to MacPorts for getting ICU, and use a newer Python with 3.11 that is supposedly available as universal2 * [macOS] Turn on universal binaries by default. This requires a new enough Python version that ships with unversal2 libraries. Let's see what github has. ------------------------------------------------------------------- Sat Nov 09 02:45:58 UTC 2024 - kv@kott.no-ip.biz - Update to version 2.4.4+git.0.75b4b41: * Release notes * Add new Yamaha YC adaptation file to the list of files to be shipped. * One more fix for the create bank with fill - don't crash when no patch matches the current filter. * Release notes for 2.4.3 * More .gitignore - I need to setup a pattern for all my build directories. * Include fix for #353 * initial version for #344, allowing to send sysex stuff directly to all outputs removing the need to use a tool like Midi-OX for such a simple thing. * Further improvement to nameFromDump. * Another improvemnt to nameFromDump. * Improve patch naming algorithm. return True from needsChannelSpecificDetection for better detection. * Simplify checksum computation. * Need to rebuild message checksum as data was modified. * Additional changes. * Minor improvement * Bug fixes. * Bug fix. * Bug fixes. * Minor changes. * Remove Yamaha-YC-Series.py - separate PR for that. * YamahaRefaceCP adaptation rewrite. * Yamaha YC Series (YC61/73/88) keyboard adaptation. ------------------------------------------------------------------- Fri Apr 8 05:59:07 UTC 2022 - Konstantin Voinov <kv@kott.no-ip.biz> - package python sequential ------------------------------------------------------------------- Fri Apr 08 03:29:28 UTC 2022 - kv@kott.no-ip.biz - Update to version 1.14.1+git.14.0f5c20e: * [winsparkle] The codesigning public key needs to go into the repo, yes? This is the key that is compiled into the executable which is used to check the validity of an update. * [cleanup] Fixed two typos pointed out by @kant. Thanks for this! * [python] Updating the embedded python to version 3.10 * [fix] When creating a new database, the user lists need to be refreshed! * [hotfix] Bump the version number to 1.14.1 * [fix] Don't crash on me when there is not a single synth configured (first startup) * [fix] Found this fix in my stash, looks like we need to reload the autocategories in certain situations!? * [Minilogue] First version of Minilogue Adaptation * [github] Ok, we need a specific version of the Windows SDK to build the WinSparkle. Hope that will not download for a long time?
