On 18.08.2016 08:53, Michael Olbrich wrote:
> And check your configure options. You need to specify everything that is
> optional or may be auto detected. Please look at some other rules for
> guidance or ask me if you are unsure if an option should be specified.

Uh, good point. Most of the --disable/--enable options seem to be built
in a way that their state is autodetected, except when overwritten on
the command line. But not for every --disable there is also a matching
--enable (that would let the configure fail if things are wrong)!

For example, in ibrcommon, only --disable-netlink really works and
disables netlink support, but --enable-netlink also disables netlink
support, because AC_ARG_ENABLE only checks for the existence of the
--*-netlink switches, not for their actual value. [0]

One could argue that configure (without --disable-netlink) will try to
detect libnl3 and find it every time because it is listed as a
dependency in the rules and is therefore built before ibrcommon.  On the
other hand, if things go wrong somehow, libnl3 could not be detected
although it should be, and configure will not interpret that as a
failure and happily build without netlink support.  So it seems like a
better solution to me if we patch those AC_ARG_ENABLE calls so they
respect the actual value of the command line switch, like in [1].

Same applies e.g. for --enable-debug and --enable-gcov (no
--disable-debug or --disable-gcov), and I didn't look into the other
ibrdtn packages yet...

[0]:
https://github.com/ibrdtn/ibrdtn/blob/master/ibrcommon/configure.ac#L217
[1]: https://autotools.io/autoconf/arguments.html

 - Roland