[Aldor-l] patch: obsolete POSIX name for CLOCKS_PER_SEC

[Stephen Watt]

This is an example of a defect that should be identified in cconfig
and a consistent interface provided in cport.

-- Stephen

On Fri, Aug 24, 2007 at 08:04:03AM -0400, Bill Page wrote:
> Steven,
> 
> On 8/24/07, you wrote:
> > Thank you Bill.
> >
> > Do you believe that we should use CLOCKS_PER_SEC pervasively instead?
> >
> 
> Unfortunately the answer is "no". You can find a lot of comments like
> this on the web:
> 
> "Here the standards are faulty: CLK_TCK is obsoleted by CLOCKS_PER_SEC,
> but CLOCKS_PER_SEC is required to be fixed at the value of 1000000L,
> so on many architectures it _differs_ from CLK_TCK !
> clock() and times() both have clock_t as a return type, but the former
> returns a value measured according to CLOCKS_PER_SEC, the latter
> according to CLK_TCK, and this is clearly absurd."
> 
> The patch that I proposed is a common solution but it is not
> universal. This is probably a very good example of the desirability of
> converting Aldor to use a standard tool like autoconf where these sort
> of inconsistencies are addressed for the greatest number of
> architectures. I would should consider this patch as just "the best we
> can do" until we have this more general solution.
> 
> In autoconf this is related to the macro AC_HEADER_TIME.
> 
> Regards,
> Bill Page.