Stefan Dösinger wrote:
Apparently gcc pads the structure size to DWORD alignment, which leaves 3 padding bytes which trigger differences in the memset. This fixes memory "leaks" in the fragment pipeline replacement without needing a memset
You should add a note to the source about this. Or someone might change the size of this structure and the old bug will return.
However the best way to fix this problem would be to pack this structure:
#include <pshpack1.h> struct ffp_settings { ... }; #include <poppack.h>
Vitaliy.