Daniel Phillips wrote:
On Thursday 28 August 2003 16:31, Svetoslav Slavtchev wrote:
Hi ,
i'm tring to incorporate the dev-mapper 1.00.04 in a kernel that already
has int fls(int x)
which conflicts with the definitions in dm-io.[ch]
(the kernel comes from mandrake cooker and the fls(x) definition is added
by swsusp patches)
so i wanted to ask you whether the attached patch looks OK,
it removes the definitions of fls & generic_fls, changes log2_align to use
fls(int x),
and adjusts the only referance of log2_align to use int instead of
unsigned int.
That's fine, except there's no point in changing the unsigned ints to signed
int: unsigned describes more accurately what they are. The automatic
conversion doesn't generate a warning for you, does it?
the issue is that fls(x) function in 2.6 use int, same as the fls(x) in
the swsusp patches, while fls(x) in dm patches use unsigned int.
so which would the correct implementation of fls be?
regards,
L.