Re: Compiling lposix.c (patch)

[David Jones <drj@...>]

On 5 Oct 2006, at 21:51, Sam Roberts wrote:

> On Thu, Oct 05, 2006 at 08:32:31PM +0100, David Jones wrote:
> > > And more importantly, nobody has posted here that *it doesn't work*.
> >
> > It doesn't work!  On OS X 10.4.
>
> Maybe using sysconf() if _SC_CLK_TCK is defined is a good way to go,
> then, falling back on CLK_TCK otherwise?

I just added:

#undef CLK_TCK
#define CLK_TCK (sysconf(_SC_CLK_TCK))

and it compiles just fine.  On OS X 10.4 CLK_TCK doesn't work with or  
without -pedantic btw.  And sysconf works with or without it  
(slightly to my surprise).

I'd be loath to clutter the code with #ifdefs.  Better to use the  
sysconf interface and see how many people complain.  It says here  
(the sysconf man page on OS X) that sysconf was defined in 1988,  
_surely_ 18 years is long enough for most OSes to have got round to  
implementing it (see also C99 discussion :) ).  But I'm sure LHF has  
an opinion on that too.

Oh dear.  Umm.  I take it all back.

I've now looked at how lposix _uses_ CLK_TCK.  It's used to  
manipulate the results of the times(3) call.  Which, in OS X, is  
documented as A) obsoleted and B) returning values in CLK_TCK's of a  
second.  But sadly, as I've discovered, CLK_TCK isn't defined on OS X  
10.4 so that makes times(3) completely useless.

> Will try when I get home, but I haven't paid the apple tax yet, I'm
> still at OS X 10.3. I hope you weren't using -pedantic to knee-cap
> otherwise working code. :-)

No, the only reason I was playing around with lposix.c was because of  
the discussion here on the list.

I'm a great fan of -ansi -pedantic -Wall though.

drj