Sylvain Becker

I don't think there is bug here. SDL_MUSTLOCK tells if you need to call lock/unlock before accessing surface->pixels. It works. For a given surface, sometime you may need to lock/unlock...

latest version of SDL_ttf_testapp.c [SDL_ttf_testapp.c.txt](https://github.com/libsdl-org/SDL_ttf/files/8672243/SDL_ttf_testapp.c.txt)

@slouken @smcv This is the latest version of the test app I use for SDL_ttf. interactive with the console (F1 for help). with 'r' it run a random fuzzing rendering...

Not sure, maybe not yet. but at least, this test program exists and can be useful when adding new feature. One can use any fonts (There are hard-coded)

Yes, this is licensed same as SDL_ttf ! Ideally, it could populate a list with some fonts found in the machine. (but I also used fixed font, emoji font, international...

Persian and Hebrew are working if you enable SDL_ttf with Harfbuzz ! And you also need to call TTF_SetDirection() TTF_SetScript()

Reversing a sub-string inside a string isn't really a big deal. At least, for a basic case: scan the string, detect latin char (being greedy) or stop / save start...

Maybe re-base the patch. so people can use in SDL_ttf, while still being optional at compilation. Also, I am not good to enough at RTL/LTR mixing rules, to see what...

I haven't re-tried that, I think it did (window for window arm platform). never try compiling on windows for android. I probably can double-check (later..)

I haven't done more testing recently. (I'll try the window compilation more later). I don't know why this is back-out. I've using these neon intrinsic routines for years now on...