Cross-compiling update.

[Amar Takhar]

On 2016-01-14 19:04 -0800, Hal Murray wrote:
> 
> verm at darkbeer.org said:
> > This is more work than I was expecting I was nearly done but (re)discovered
> > that  keyword-gen uses libntp! 
> 
> Will it simplify things if I fix that?

Yes, by a large amount it is very worth it if we can make it depend only on 
itself.


> > I also discovered one quirk by accident when compiling for ARM:
> >   37:#error libntp supplied mktime()/timegm() do not support 64-bit time_t
> > I wonder how often users will hit that?
> 
> How did you generate that error?

I'm not sure, it's not detecting timegm in the ARM libraries I'm using for some 
reason it's probably an odd environment quirk.  I'm setting up a proper one 
tomorrow morning that I can copy over to Buildbot.


> Do we really need our copy of mktime and timegm?  Is that one more piece of 
> cruft we can get rid of?

No idea, are there any systems that don't have a mktime that we need?


> grep says that timegm isn't used.
> 
> My Linux man page says:
>    Feature Test Macro Requirements for glibc (see feature_test_macros(7)):
>        asctime_r(), ctime_r(), gmtime_r(), localtime_r():
>               _POSIX_C_SOURCE >= 1 || _XOPEN_SOURCE || _BSD_SOURCE ||
>               _SVID_SOURCE || _POSIX_SOURCE
> 
> I think that means if we are assuming POSIX, gmtime will be there.

I think this is the case, too.


Amar.