[Kea-users] Selection of awk binary in configure script

[Patrik Lundin]

Hello,

I was notified by someone running bulk builds of OpenBSD ports that the
kea build had failed because it detected "gawk" during its configure
stage, which had then been removed in the background while the build was
completing.

There are two ways to deal with this: either I add a build dependecy on
gawk to make sure it is present, or I make the configure script not
look for gawk at all.

Since the build is perfectly happy with the awk supplied by the base
system, adding another dependency is a pretty ugly workaround (should I
then proceed to add build dependencies for all possible *awk versions?)

The cleaner way is of course to make the configure script only look for
"awk". I could handle this in the port, but local patches are never fun
to maintain.

The question I have is therefore: would it be a good idea to make the
upstream version prefer awk over gawk/nawk/mawk?

It appears to me that systems like Debian make "awk" point to "the
system preferred awk-tool" anyway, so this might make sense there as
well.

While looking at this I also noticed something I did not understand. The
configure script in the kea-1.0.0 tar-ball does this:
===
for ac_prog in gawk mawk nawk awk
===

However, the configure.ac file in that same tar-ball only contains these
awk-related settings:
===
AC_PATH_PROGS(AWK, gawk awk)
===

I find no trace of mawk or nawk in any files other than the supposedely
generated configure script. Where does these come from?

-- 
Patrik Lundin