[RFC] Add __arraycount from NetBSD to sys/cdefs.h

[Konstantin Belousov]

On Mon, Sep 08, 2014 at 09:29:54AM -0700, yaneurabeya at gmail.com wrote:
> On Sep 3, 2014, at 21:17, Warner Losh <imp at bsdimp.com> wrote:
> 
> > On Sep 3, 2014, at 9:45 PM, Garrett Cooper <yaneurabeya at gmail.com> wrote:
> > 
> >> Hi all,
> >>   In order to ease porting code and reduce divergence with NetBSD
> >> when importing code (a large chunk of which for me are tests), I would
> >> like to move nitems to sys/cdefs.h and alias __arraycount to nitems.
> >>   Here's the __arraycount #define in lib/libnetbsd/sys/cdefs.h:
> >> 
> >> 44 /*
> >> 45  * Return the number of elements in a statically-allocated array,
> >> 46  * __x.
> >> 47  */
> >> 48 #define __arraycount(__x)       (sizeof(__x) / sizeof(__x[0]))
> >> 
> >>   Here's the nitems #define in sys/sys/param.h:
> >> 
> >> 277 #define nitems(x)       (sizeof((x)) / sizeof((x)[0]))
> >> 
> >>   sys/cdefs.h gets pulled in automatically with sys/param.h, so
> >> anything using nitems will continue to function like before (see below
> >> for more details). I've attached a patch which addresses all hardcoded
> >> definitions in the tree added by FreeBSD developers.
> >>   If there aren't any major concerns with my proposed change, I'll
> >> put it up for review on Phabricator.
> >> Thank you!
> >> -Garrett
> >> 
> >> $ cat cdefs_pound_define.c
> >> #include <sys/param.h>
> >> 
> >> #ifdef _SYS_CDEFS_H_
> >> #warning "sys/cdefs.h has been included"
> >> #endif
> >> $ cc -c cdefs_pound_define.c
> >> cdefs_pound_define.c:4:2: warning: "sys/cdefs.h has been included" [-W#warnings]
> >> #warning "sys/cdefs.h has been included"
> >> ^
> >> 1 warning generated.
> >> $ cc -D_KERNEL -c cdefs_pound_define.c
> >> cdefs_pound_define.c:4:2: warning: "sys/cdefs.h has been included" [-W#warnings]
> >> #warning "sys/cdefs.h has been included"
> >> ^
> >> 1 warning generated.
> >> $ gcc -c cdefs_pound_define.c
> >> cdefs_pound_define.c:4:2: warning: #warning "sys/cdefs.h has been included"
> >> $ gcc -D_KERNEL -c cdefs_pound_define.c
> >> cdefs_pound_define.c:4:2: warning: #warning "sys/cdefs.h has been included?
> > 
> > I wouldn?t bother changing the nitems #define. There?s no need, really, to do that.
> 
> Rethinking my proposal, I agree. I had lofty hopes for unifying the macros, but the functional duplication (1 line) is harmless.
> 
> > I?d also be more inclined to believe the test if you tested what the thing does rather than test for an artificial, implementation defined side effect.
> 
> Sure. I provided a lazy proof instead of a full proof :).
> 
> > But honestly the amount of duplication saved here is rather tiny?
> 
> Indeed! Thank you for the input :) ? I?ve attached a new patch which doesn?t disturb nitems or sys/param.h.

Since this is seriously discussed on arch@, I think it is worth look
at the following:

git grep -e 'sizeof.*/sizeof' -- | grep -v '^contrib' | grep '#.*define' | wc -l
     118
These are only defines, not uses, not counting direct use of
sizeof(x)/sizeof(x[0]) in the code.

Not all of them are for equiv macros, but significant part repeats ntimes()
in variations.  At least Linuxish ARRAY_SIZE/DRM_ARRAY_SIZE is much more
popular than __arraysize() which appears three (!) times.

I do not see any reason to pollute central header cdefs.h with sparcely
used macro for some compat code.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-arch/attachments/20140909/2a043aa8/attachment-0001.sig>