A Place for patch demos and machine development news.
KRTPluginJLatest Auto jump to where the “edit limit” is plus addition forward … Auto insert on save?
So the make works, but te modules are nowhere near ready. I’m making the “standard” plugin.cpp/plugin.hpp for all my future plugins. A generic project starter. I’ve abstracted the edits when making modules to be in modules.hpp (for the list of module sources) and menus.txt (for the context menus and threaded signalling).
Sayc, sarc .ASM a bound?
I’ve added make macjig to force menus.txt change recompilation. A stop gap as .hpp er, should standardly work, ‘cos like that’s how .cpp lines of #include go like man. Owe quey, reda di Manuelle Jaw de sure like man?
Next!
I vont to vrock’it into spass. Like a moo-sycle style like so man, hey dudumdida, ho, yes, or “well it’s like the loader is like flawed, but excellent, but needs a stop ‘till end then go, as an easy man.” 2023-03-02 yes either same TB-303*2.
Yes, the build system should now be better. It follows a few botches and generic code hacks. So now I’m thinking about maybe other languages to add. I though LOGO for fun, but TO NAME :VAR is all wrong. It should’ve been TO "NAME "VAR for evaluation consistency. As an adult it makes lisp macro look useful but primarily avoided to remove one letter and add much in the source code. Implicit " is kind of cool in setq, but suffers from inconsistency, and as a learner lingo distracted by pizza turtles and living in the sewer, erm, is it really the daddy or the diddy, butt wee wont go win to dat?
Hi ho. Next up (no postscript although cool, has much to blit and little to command line decompose). I So netpbm is graphically cool in a textual number data format. I wonder about that fractal algorithm, as it reminds me of the .fif file extension from the really early ’90s. I managed to recompress the uncompressed residual in a loop at 50% quality and after 3 iterations it was better than 100%b in less time, even though an archive .zip of the multiple outputs was required.
COBOL, ASM (machine specific), PL/1, FORTRAN (good supercomputer heritage though), spreadsheet outputs (LibreOffice), CSV, …
So I’ve classed the menus as OnMenu so multiple instances can occur. That was relatively easy, but involved a bit of forwarding of classes to callback actions. So it now is complete enough. Next is hot keys. I’ve got to read up on the docs.
error: assignment to 'int (*)(const __mpq_struct *, const __mpq_struct *)' from incompatible pointer type 'FARPROC' {aka 'long long int (*)()'} [-Werror=incompatible-pointer-types] is the latest, perhaps a cast? So …
So mac builds, lin complained at sudo so removed for assumed root, and win seemed to fail on the j link, perhaps an unneeded resource file from MS. I’ll set it to one more wizz tonight and fingers crossed only win will be left to fix in the morning. :D
So I might have solved the mingw resource compiler issue, but the shared linux container baulks about not finding the include for -luuid in premake5 building, while both the macos version fail on *** "invalid configuration release". when building efsw with the built premake5, so perhaps that’s a default config issue …
But yes closer than yesterday.
cd efsw/make/macosx && make config=release
Makefile:42: *** "invalid configuration release". Stop.
make: *** [libefsw.a] Error 2
Error: Process completed with exit code 2.
And
12 | #include <uuid/uuid.h>
| ^~~~~~~~~~~~~
compilation terminated.
make[2]: *** [Bootstrap.mak:110: linux] Error 1
make[2]: Leaving directory '/__w/jqt-chromebook-arm/jqt-chromebook-arm/premake-core'
make[1]: *** [Makefile:141: premake-core/bin/release/premake5] Error 2
make[1]: Leaving directory '/__w/jqt-chromebook-arm/jqt-chromebook-arm'
make: *** [Makefile:186: plugin-build-linux-x64] Error 2
Error: Process completed with exit code 2.
But still some way to go. I won’t know if the windows build has similar premake5 issues until it passes the j build point, which is not helped by the combined container process kill. So options are install uuid but given the container complained when trying to install premake4 which is a linux package from bionic Ubuntu onwards, but uuid-dev might work where premake4 didn’t.
WinRes MinGW might work though to compile resources.
Some nice macro expansions to make the names and variables for a module from just a #define NAME Blank at the top. Makes for an easy generic copy of the Blank template to add new modules. Nice Macros.
Ok, so UUID and uuid-dev package. I’m not sure even this will completely fix the premake5 dependency chain, as that is just the bootstrap premake5 before the actual build full which includes all sorts of SSL and curl stuff, which isn’t needed for many things. I’m on Debian Bullseye, and I’ve installed many *-dev packages doing things.
OK :D (Nice Rust ID on the SIM repo … robot giggle)
>
Friday 2023-03-10 16:00:43 [dev-bot] #47 compiles all, but the win has no j, no efsw callback working and no forking due to posix pipe support and fork not being available. There’s a nice isWindows() function though for further development, and a fileExeTension() for when the base name doesn’t work. I kept it generic justin in case I make windows fork work. So a weird hack of premake4 (lin by the repo makefiles predone, mac by a botch brew and win by ignoring such notifications of change). It’s kind of impossible without a set of makefiles for win. Any offers? Oh --os= a .gitignore and makefiles for all platforms :D.
I almost messed up on the -1 control specifications for “no control” at panel location. Likely would have seg faulted. It’s shaping up in my mind. Maybe a little brainfuck audio so that even windows has a “processor lingo” not needing fork. I like the idea, and it keeps the general GPL3 growth with this plugin instead of a dead zilch.
Feel free to use jackokring/efsw if you don’t want to go through the Makefile generation of targets you don’t have access to. It is the best CI solution for me. The evil Lua is put to for such a nice little embedable lingo. Maybe …
So some hot key functions to limit usage in the ANSI keyboard MIDI input context. QMK input devices rock. I’ve got a board someday to MX blue. The parameters need a Module or ParamWidget and so then only allow Hover and Select of the AFK-'48 set. This reserves space and tab and the function set not on Chromebook and ANSI-60. Alt \ and # are removed too as they’re internationally renowned. I do find AltGr+U a Unicode funny in the ANSI marking.
Ah, so some simple filename construction helper abstractions and tunneling some std::string down to NanoSVG to save some headache on assignments of labels, due to the Module name being quite a difficult thing to obtain. Yes, very funny ModuleWidget has no Model seg. fault. But fixed now. As you may have guessed I’m not a fan of type upcasting char* to std::string unless there is a code efficiency reason to do it. There maybe further upcasts later, but for now …
So some functions to handle transparent file initialization from the presets folder for a module, and to transparently make any necessary directories. Also making it an interface so as to not place any files where they shouldn’t go. It’s not fully secure in that /../ is not replaced and it still easy to access other files. It’s more of a “restriction interface”.
Just uninstalled the C++ VSCode extension. Oh how fast she goes now! It’s almost like gedit. Hurrah for ridding the stupid idea of code completion in a lingo where type prefixes and wild card identifier introduction is illogical and macros morph the code beyond application of regex.
So I’ve started on some simple uiTheme SVG files for the backgrounds. Simple light and dark versions of the base colour. I’ll have to decide on the controls next. Keeping it very black/white seems like the easy way.