Обсуждение: AW: AW: AW: AW: AW: AW: AW: Re: tinterval - operator pr oble ms o n AIX

> > We do not need any execution time checks for this at all. The objective is
> > to determine whether mktime works for any results that would be negative.
> > On AIX and IRIX all calls to mktime for dates before 1970 lead to a result of
> > -1, and the configure test is supposed to give a define for exactly that behavior.
> 
> Okay, so you call mktime with a pre-1970 date once when the system starts
> up or when the particular function is first used and then save the result
> in a static variable.

Can anybody else give an OK to this approach, that affects all platforms ?
I am not convinced, that this is the way to go.

Andreas

PS: next response not before Monday, I am off now :-)

> > Okay, so you call mktime with a pre-1970 date once when the system starts
> > up or when the particular function is first used and then save the result
> > in a static variable.
> Can anybody else give an OK to this approach, that affects all platforms ?
> I am not convinced, that this is the way to go.

Nope. So far we have consensus that #ifdef <something> is the way to go
(I just made some changes to the date/time stuff to isolate the #ifdef
__CYGWIN__ garbage, and would like to avoid more cruft), but we are not
unanimous on the way to get that set.
                         - Thomas