SDL_CalculatePitch() too small

[SDLBugzilla]

This bug report was migrated from our old Bugzilla tracker.

Reported in version: 1.2.13
Reported for operating system, platform: Linux, x86

Comments on the original bug report:

On 2010-04-21 00:27:20 +0000, wrote:

SDL_CalculatePitch() switches surface.format->BitsPerPixel.

1. SDL doesn't allow BitsPerPixel smaller than 8 and cases are 1 and 4, which will never be matched

2. as surface.pitch is a multiple of 8 bits,1 byte respectively , 1 and 4 BitsPerPixel both result in 1 BytesPerPixel (see SDL_AllocFormat()) and therefore it doesn't make sense to treat them differently considering the byte aligning

3. to access single pixels via surface.pixels you use at least multiples sizeof(char), so that for a given surface.w insufficient memory is allocated at surface creation: (pitch+7)/8, (pitch+1)/2 (surface.w = 5, 1 BitsPerPixel, 1 BytesPerPixel produces pitch=1 (4 aligned) instead of pitch=5 (8 aligned))

On 2010-04-21 00:30:45 +0000, wrote:

Already mentioned by Sylvain Hellegouarch[1] in 2003, who got poor answers.

[1]http://lists.libsdl.org/pipermail/sdl-libsdl.org/2003-April/034932.html

On 2010-04-23 02:04:30 +0000, Sam Lantinga wrote:

Do you have a specific example of where this is a problem?

The places where I've seen bitmap data used, the pixels are packed in as tightly as possible and the final scanline is aligned on a byte or 4 byte boundary.

To access single pixels, you unpack the bits out of the scanline in order.

For an example, look at the "Bitmap data" section in the BMP file format:
http://en.wikipedia.org/wiki/BMP_file_format

I think what you're talking about is an 8 bpp image with a limited palette.

On 2010-04-24 04:46:56 +0000, wrote:

See example above: when using a 5 pixel-broad surface with 1 bit per pixel 4 bytes instead of 8 are allocated. If you try to manipulate the last $SURFACE_HEIGHT pixels you will run into memory violation, because no memory was allocated for them. With color depth other than 1 and 4 there is no problem (both not allowed by SDL, so that nobody really runs into problems, but the implementation is incorrect nevertheless)

On 2012-01-02 00:09:38 +0000, Sam Lantinga wrote:

If this is actually a problem for someone, please reopen this bug. For now, I'll close it, since I have lots of other things I need to do.