[ntp:questions] 4.2.8 can't find timepps.h or can it?

William Unruh unruh at invalid.ca
Mon Dec 22 06:08:42 UTC 2014


On 2014-12-21, Harlan Stenn <stenn at ntp.org> wrote:
> William Unruh writes:
>> On 2014-12-21, Harlan Stenn <stenn at ntp.org> wrote:
>> > It doesn't contradict itself.  It finds sys/timepps.h.  This is
>> > expected.
>> >
>> > It does not find timepps.h.  This is expected.
>> >
>> > The former lives in /usr/include/sys/timepps.h, as you note above.
>> >
>> > The latter does not live in /usr/include/timepps.h, as you note above.
>> >
>> > We both note that ntpd compiles and the ATOM driver works.
>> 
>> Why would it be searching for timepps.h when it had already found
>> sys/timepps.h?
>
> Using a general example, because it's much easier and safer to write and
> use:
>
>  AC_CHECK_HEADERS([timepps.h sys/timepps.h])

Except that is not the way configure actually works. It looks for
sys/timepps.h and long afterwards looks for timepps.h

Line 27521: for ac_header in sys/timepps.h

line 28812: for ac_header in termios.h timepps.h timex.h unistd.h

>
> than it is to unequiviocally decide that one of a choice is always
> better than the other and then write code so that we *only* look for the
> other one if the former doesn't exist.

Better? No just first, or most common.

And why would you want to look for both when one will do?

>
> H



More information about the questions mailing list