SDL 1 CI build fixes

[MyreMylar]

This is a grab bag of small fixes to try and get the SDL 1 build building and passing tests again on Travis CI.

There is more detail in this draft PR where I discovered most of these errors in the process of trying to write some tests for the overlay module:

#1830

[illume]

👍

[MyreMylar]

Looks like two of the new mask tests are also failing on SDL 1, is that because of this bit of code perhaps:

pygame/src_c/mask.c

Lines 2072 to 2076 in f439234

	#if IS_SDLv1
	SDL_SRCALPHA,
	#else
	PGS_SRCALPHA,
	#endif

Maybe @charlesej will know.

EDIT: The failed SDL 1 tests:

======================================================================
FAIL: test_to_surface__different_srcalphas (pygame.tests.mask_test.MaskTypeTest)
Ensures an exception is raised when surfaces have different SRCALPHA
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/virtualenv/python3.5.6/lib/python3.5/site-packages/pygame-2.0.0.dev9-py3.5-linux-x86_64.egg/pygame/tests/mask_test.py", line 4573, in test_to_surface__different_srcalphas
    mask.to_surface(surface, setsurface, unsetsurface)
AssertionError: ValueError not raised
======================================================================
FAIL: test_to_surface__different_srcalphas_with_created_surfaces (pygame.tests.mask_test.MaskTypeTest)
Ensures an exception is raised when surfaces have different SRCALPHA
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/virtualenv/python3.5.6/lib/python3.5/site-packages/pygame-2.0.0.dev9-py3.5-linux-x86_64.egg/pygame/tests/mask_test.py", line 4595, in test_to_surface__different_srcalphas_with_created_surfaces
    mask.to_surface(setsurface=setsurface, unsetsurface=unsetsurface)
AssertionError: ValueError not raised
----------------------------------------------------------------------

[charlesej]

@MyreMylar I think there needs to be a check added to the check_surface_pixel_format function for SDL 1.

pygame/src_c/mask.c

Lines 2023 to 2036 in f439234

	static int
	check_surface_pixel_format(SDL_Surface *surf, SDL_Surface *check_surf)
	{
	if ((surf->format->BytesPerPixel != check_surf->format->BytesPerPixel) ||
	(surf->format->BitsPerPixel != check_surf->format->BitsPerPixel)
	#if IS_SDLv2
	|| (surf->format->format != check_surf->format->format)
	#endif
	) {
	return 0;
	}
	return 1;
	}

Something like:

static int
check_surface_pixel_format(SDL_Surface *surf, SDL_Surface *check_surf)
{
    if ((surf->format->BytesPerPixel != check_surf->format->BytesPerPixel) ||
        (surf->format->BitsPerPixel != check_surf->format->BitsPerPixel)
#if IS_SDLv2
        || (surf->format->format != check_surf->format->format)
#else
        || ((surf->flags & SDL_SRCALPHA) != (check_surf->flags & SDL_SRCALPHA))
#endif
    ) {
        return 0;
    }

    return 1;
}

[MyreMylar]

Hooray, with that last fix from Charles it all runs. Probably should merge it in quick before we all break it again in every other PR 🍻 🤦

[illume]

It makes me a little nervous that these aren't correct on SDL1. Are the tests correct, or is SDL1 incorrect? Has the behaviour changed in these blit routines now? Do the SDL2 ones not match the SDL1 ones?

[illume]

This doesn't need a response necessarily :) I was just thinking out loud.

[MyreMylar]

No, you are right to ask I'm definitely going to poke at it and see if I can figure out why this works differently.

I mean in theory you don't need to do pre-multiplied alpha blending in any of those cases because both surfaces don't have alpha but I believe I would expect it to just work boringly without going through the fancy intrinsic code....

Anyway, it just didn't seem the moment to go off on a tangent figuring that out in this PR.

[illume]

👍 thanks, and sorry I messed this up before.