Default Sdk Not Respecting #pragma Pack(x)

[Daid]

Traced back the not loading of clonk to the failure of loading the bmp images. What traced back to this function:

BOOL CBitmap256Info::Valid()
	{
  if (Head.bfType != *((WORD*)"BM") ) return FALSE;
  if ((Info.biBitCount!=8) || (Info.biCompression!=0)) return FALSE;
	return TRUE;
	}

In this class:

#pragma pack(1)

class CBitmap256Info
	{
	public:
		CBitmap256Info();
	public:
		BITMAPFILEHEADER Head;
		BITMAPINFOHEADER Info;
		RGBQUAD Colors[256];
	public:
		void Default();
		void Set(int iWdt, int iHgt, BYTE *bypPalette);
		BOOL Valid();
		int FileBitsOffset();
	};

#pragma pack()

If I add this debug line:

printf("%i %i %i %i\n", (int)this - (int)&Head, (int)this - (int)&Info, sizeof(Head), sizeof(Info));

in the IsValid function, then I get this:
Window/Linux, x86: 0 -14 14 40
Linux, GP2x: 0 -16 16 40

And thus, it doesn't pack the Head structure correctly.

Traced back to this code:

#pragma pack(2)
typedef struct tagBITMAPFILEHEADER {
	WORD	bfType;
	DWORD	bfSize;
	WORD	bfReserved1;
	WORD	bfReserved2;
	DWORD	bfOffBits;
} BITMAPFILEHEADER,*LPBITMAPFILEHEADER,*PBITMAPFILEHEADER;
#pragma pack()

Any ideas?

 

[yaustar]

#pragma s are usually Microsoft compilers specific so they don't work with GCC except for the latest versions.

http://www.rtems.com/ml/rtems-snapshots/19...r/msg00062.html

 

[Daid]

Solved it, a slightly diffrent way.

Changed the loading code:

	if (!hGroup.Read(&BitmapInfo,sizeof(BitmapInfo))) return NULL;

To:

	if (!hGroup.Read(&BitmapInfo.Head,14)) return NULL;
	if (!hGroup.Read(&BitmapInfo.Info,40)) return NULL;
	if (!hGroup.Read(&BitmapInfo.Colors,1024)) return NULL;

Clonk Planet runs now Still segfails, but only on exit.

Just need to build some frontend now.

 

[Parkydr]

You could try __attribute__((__packed__)) instead, works with oopo's toolchain

E.g.

struct RGB {
		unsigned char red, green, blue;
} __attribute__((__packed__));