From: Simon McVittie Date: Mon, 22 Aug 2022 11:25:33 +0100 Subject: surface_test: Don't overlap pixel rows SDL has the concept of pitch (sometimes referred to as stride or rowstride in other imaging libraries, for example GNOME's GDK), which is the number of bytes to advance through the data for each row/scanline. Typically this is either exactly the number of bytes required to store pixel data for a row, or a larger number chosen to align the beginning of each row to a convenient memory address (for example a multiple of 16 bytes for SSE2). Previously this test data used a fixed pitch of 16 bytes, even though this is less than the memory required to store each 16-pixel row, because each pixel uses 4 or 2 bytes of data (and therefore each 16-pixel row is 64 or 32 bytes long). This meant SDL would start reading the second row of pixels 16 bytes into the first row, and so on, with the data for each row overlapping, and some unused bytes at the end. This seems unlikely to have been intentional. SDL 2.23.x and 2.24.0 have better validation for parameters, introduced while adding overflow checks, which enforces that pixel rows do not overlap (pitch < bytes per row). They may either have padding for better alignment (pitch > bytes per row) or have no padding (pitch == bytes per row, as seen after this commit). Increase the pitch to match what was presumably intended. Forwarded: not-needed, fixed differently in commit 7c44df68 "Start migrating surface tests to modern pytest" --- sdl2/test/surface_test.py | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/sdl2/test/surface_test.py b/sdl2/test/surface_test.py index 98abdf2..47878af 100644 --- a/sdl2/test/surface_test.py +++ b/sdl2/test/surface_test.py @@ -17,12 +17,12 @@ rgbdepths = (8, 12, 15, 16) rgbadepths = (16, 24, 32) rgba_pixelations_16x16 = ( - # 32-bit 16x16 RGBA surface - ([x << 24 for x in range(16 * 16)], 32, 16, + # 32-bit 16x16 RGBA surface (64 bytes per row) + ([x << 24 for x in range(16 * 16)], 32, 64, (0xFF000000, 0x00FF0000, 0x0000FF00, 0x000000FF), pixels.SDL_PIXELFORMAT_RGBA8888), - # 16-bit 16x16 RGBA surface - ([x << 8 for x in range(16 * 16)], 16, 16, + # 16-bit 16x16 RGBA surface (32 bytes per row) + ([x << 8 for x in range(16 * 16)], 16, 32, (0xF000, 0x0F00, 0x00F0, 0x000F), pixels.SDL_PIXELFORMAT_RGB444), )