Limitations

Padding of bitfields

How bitfields are stored in memory is not defined by the C standard. pycstruct c parser tries to keep consecutive defined bitfields together, however gcc (for example) will split bitfields when alignment is needed.

Following example will be a problem:

struct a_struct {
  char m1;       // = 8 bits
  int bf1 : 2;   // = 2 bits
  int bf2 : 23;  // = 23 bits
};

pycstruct c parser will layout the struct like this using 32 bit alignment:

Size in bytes	Type	Name
1	Unsigned integer	m1
1	Unsigned integer	__pad_0[0]
1	Unsigned integer	__pad_0[1]
1	Unsigned integer	__pad_0[2]
4	Bitfield	bf1 + bf2

gcc will layout the struct like this using 32 bit alignment:

Size in bytes	Type	Name
1	Unsigned integer	m1
1	Bitfield	bf1
1	Unsigned integer	__pad_0[0]
1	Unsigned integer	__pad_0[1]
3	Bitfield	bf2

One workaround is to pack the struct using compiler directives.

Another workaround is to manually add padding bytes in the struct in-between the bitfield definitions.