I don't know about this, until now it could show the version even if it wasn't initialized which kind of makes sense, so following that this should also just show the current version instead of raising an exception. However, this makes sense too, it just doesn't conform to the previous versions.

@Matiiss here is what _ft_get_version looked like before #3379:

static PyObject *
_ft_get_version(PyObject *self, PyObject *_null)
{
    /* Return the linked FreeType2 version */
    return Py_BuildValue("iii", FREETYPE_MAJOR, FREETYPE_MINOR,
                         FREETYPE_PATCH);
}

Although, I'm pretty sure that comment is wrong and those macros are the compiled version. That code still exists after the PR merge from above, but the linked version requires a FreeTypeInstance, and the one initialized in pygame doesn't get filled until pygame.freetype.init() is called. So, the compiled version can be returned without raising an exception, but the linked version cannot currently.

In fact, here's the declaration of those macros in the freetype.h header in the FreeType2 lib files:

  /**************************************************************************
   *
   * @enum:
   *   FREETYPE_XXX
   *
   * @description:
   *   These three macros identify the FreeType source code version.  Use
   *   @FT_Library_Version to access them at runtime.
   *
   * @values:
   *   FREETYPE_MAJOR ::
   *     The major version number.
   *   FREETYPE_MINOR ::
   *     The minor version number.
   *   FREETYPE_PATCH ::
   *     The patch level.
   *
   * @note:
   *   The version number of FreeType if built as a dynamic link library with
   *   the 'libtool' package is _not_ controlled by these three macros.
   *
   */
#define FREETYPE_MAJOR  2
#define FREETYPE_MINOR  11
#define FREETYPE_PATCH  1

But I think I have actually found a solution, so I'll add another commit

Would it not be better just to do:

if (linked) {
    if (!inst) {
            PyErr_SetString(PyExc_RuntimeError,
                            "FreeType is not be initialized");
        return NULL;
    }

Right now it seems like this would be leaking memory from initialising two versions of freetype and the one in get_version() not being cleaned up?

@MyreMylar I see two routes:

1. The way I have done it (with fixes to make sure freetype gets cleaned up properly in get_version)
2. If pygame.freetype is not initialized, either throw an error or return something like (0, 0, 0)

With option 1, I understand that it's not the most ideal solution because of the extra initialization of freetype. But, I also would like to see this method behave similarly to the other getters, who don't need module initialization within pygame to work. Which is why I opted to pursue that route (and then forgetting to clean up after myself). Also considering that this method is intended to help diagnose problems with pygame, I think it would be bad form to throw an error or return a dummy value just because a module isn't initialized.

I'd like to hear your thoughts about it. In the meantime, I'll go ahead and push a commit hopefully cleaning up properly (for some reason, FreeType2's docs are really hard to read for me)

Thoughts--

• This only broke because of the linked/compiled change to the get_version() function.
• I agree with Myre it would be simpler just to error, but I like Andrew's thoughtfulness to want to make it behave like everything else, and it doesn't seem too complex to do so.
• Why the changes to ft_wrap.c / ft_wrap.h?

Why the changes to ft_wrap.c / ft_wrap.h?

It’s more of a “returning those lines back to the way they were before the PR that broke this function in the first place

Should we add a test that freetype.get_version() can be called before pygame.freetype.init?