> > not clear that there hasn't in fact been a change in the
> > semantics here.
> 
> No, there has been no change in semantics.  ...

To the extent that only signatures changed, I'll now partly agree.

Partly: because when I looked at the alpha and ia64 code, I see
code which seems to expect 32-bit values at the end of those
pointers, even if "unsigned long" would be a 64-bit value.  (Yet
MIP64 and SPARC64 seems to use 64-bit values ... sigh.)


> This was one of these "everything knows" things which many people
> didn't know.

Surely in part because it wasn't documented in the places such things
should be documented ... and it wasn't clear whether the code or the
doc was wrong.  The documentation easily accessible said those
pointers were "void *".  It's now partly fixed, for x86 and ppc.

I still think the doc is problematic.  It seems like "everyone knows" that
bit zero is the LSB in a word.  However, I just checked the x86 (386)
and PPC (601) manuals on my shelf ... PPC says zero is MSB, while
x86 says zero is LSB.  While microprocessors lately tend to use the
little-endian bit numbering, it's not universal.  Worth documenting that
Linux uses a "zero is LSB" convention internally that may be different
from the standard one on the host processor.

- Dave








_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to