Small Device C Compiler suite

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hynek Sladky schrieb:

> It seems that second part of bitfield starts at next byte location from 
> bit 0 :-(

Why not? AFAIK the C standard doesn't guarantee alignment in bitfields,
it doesn't even guarantee alignment in structures.

Philipp

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGoGoAbtUV+xsoLpoRAgQaAJ0VVUZb66R/JEpW2OOpdu16mh8mNgCfRH/G
7ZAsFqGRChAJlQyMEJyrLW8=
=y96K
-----END PGP SIGNATURE-----

But it is sometimes used for such purposes (defining e.g. HW register 
content - and I adopted it as well) so I don't understand why it is used 
such way if it is not guaranted... Is there correct way for it?

BTW - my structure works in all other compilers I have used.

Thanks,
Hynek Sladky


Philipp Klaus Krause napsal(a):
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Hynek Sladky schrieb:
> 
> 
>>It seems that second part of bitfield starts at next byte location from 
>>bit 0 :-(
> 
> 
> Why not? AFAIK the C standard doesn't guarantee alignment in bitfields,
> it doesn't even guarantee alignment in structures.
> 
> Philipp

I think it is a bug because:

struct {
   unsigned short CCC:12;
   unsigned short READ_BL_LEN:4;
} CSD;
----> works as expected (2 bytes long)

struct {
   unsigned short READ_BL_LEN:4;
   unsigned short CCC:12;
} CSD;
----> works wrong (3 bytes long)

I think that this should work the same way shouldn't it?
Unfortunately I need the second version... :-(

Hynek Sladky

The last note:
if bitfield in SDCC for the first time crosses byte boundary, it always 
starts at bit zero. The strange thing is that second example produces 3 
byte long structure (no 4 bytes long) so second byte-boundary crossing 
seems working correctly.

// correct
struct {
   unsigned int bit1:3;
   unsigned int bit2:5;
   // byte boundary
   unsigned int bit3:7;
   unsigned int bit4:1;
} struct1;

// wrong
struct {
   unsigned int bit1:3;
   // byte boundary, next field doesn't continue in first byte of structure
   unsigned int bit2:6;
   unsigned int bit3:6;
   unsigned int bit4:1;
} struct2;

These two structures don't generate same result: sizeof(struct1) = 2 
while sizeof(struct2) = 3...

Hynek

>From "C++: The Complete Reference", by Herbert Schildt, page 176:

  You cannot know, from machine to machine, whether the fields
  will run from right to left or from left to right; this implies
  that any code using bit-fields may have some machine
  dependencies.  Other restrictions may be imposed by various
  specific implementations, so check the user manual for your
  compiler.

A long time ago, I had two Windows C compilers that produced
bitfields in opposite directions.  I wasted a lot of time tracking
down that 'bug'.  So I no longer trust bitfields.  I like bitwise
operators better anyway.

-Ken Jackson


Hynek Sladky writes:
 > I think it is a bug because:
 > 
 > struct {
 >    unsigned short CCC:12;
 >    unsigned short READ_BL_LEN:4;
 > } CSD;
 > ----> works as expected (2 bytes long)
 > 
 > struct {
 >    unsigned short READ_BL_LEN:4;
 >    unsigned short CCC:12;
 > } CSD;
 > ----> works wrong (3 bytes long)
 > 
 > I think that this should work the same way shouldn't it?
 > Unfortunately I need the second version... :-(
 > 
 > Hynek Sladky

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hynek Sladky schrieb:
> I think it is a bug because:
> 
> struct {
>    unsigned short CCC:12;
>    unsigned short READ_BL_LEN:4;
> } CSD;
> ----> works as expected (2 bytes long)
> 
> struct {
>    unsigned short READ_BL_LEN:4;
>    unsigned short CCC:12;
> } CSD;
> ----> works wrong (3 bytes long)
> 
> I think that this should work the same way shouldn't it?

AFAIk compilers may align larger struct members to different from
smaller ones (like aligning ints to 16 bit boundaries, aligning bytes to
8 bit boundaries). I suppose it's the sme for bitfields.

Philipp

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGoMJ3btUV+xsoLpoRAntAAJ9eimXP9rQwxreLmlAlmd9tHXQ/rACg5CAm
Fs6R4XfPrg9+1oUQ0BkxCCw=
=xebT
-----END PGP SIGNATURE-----

On Friday 20 July 2007 10:11, Philipp Klaus Krause wrote:
> Hynek Sladky schrieb:
> > I think it is a bug because:
> >
> > struct {
> >    unsigned short CCC:12;
> >    unsigned short READ_BL_LEN:4;
> > } CSD;
> > ----> works as expected (2 bytes long)
> >
> > struct {
> >    unsigned short READ_BL_LEN:4;
> >    unsigned short CCC:12;
> > } CSD;
> > ----> works wrong (3 bytes long)
> >
> > I think that this should work the same way shouldn't it?
>
> AFAIk compilers may align larger struct members to different from
> smaller ones (like aligning ints to 16 bit boundaries, aligning bytes to
> 8 bit boundaries). I suppose it's the sme for bitfields.

I don't know much about sdcc, but what about something like #pragma pack(1) or 
__attribute__ ((packed)), shouldn't this force tight packing ?

Alex