How distcheck configuration should work

Marcin Siodelski marcin at isc.org
Wed Apr 16 19:17:28 UTC 2014


It appears to me that Kea distcheck configuration may be wrong.

Typically, when we run distcheck we do:

autoreconf --install
./configure <switches>
make distcheck

On some systems we don't need switches because everything is in the
default location and can be located by the appropriate macros.

However, in many cases we do put switches to set locations of the third
party software and to configure what bits of Kea we want to build.
Currently, we have optional switches --with-dhcp-mysql and
--with-dhcp-pgsql which enable building of MySQL and PostgreSQL lease
database backends. We have been already discussing compilation time
switches to enable one of the crypto backends to be supported.

The reality is that the distcheck, as we have it today, WILL NOT use
configuration switches specified in:

./configure <switches> (see above)

This means that if you require to specify location of log4cplus to build
Kea, the configure will go just fine, but the distcheck will fail
complaining about the lack of log4cplus, because the --with-log4cplus
will be dropped. If you specify --with-dhcp-mysql or --with-dhcp-pgsql
the switches will also be silently dropped by distcheck. As a result,
the backends will not compile and unit tests will not be run.

There are solutions for that, like setting
AM_DISTCHECK_GTEST_CONFIGURE_FLAG variable prior to distcheck. But, I
don't really see a compelling reason why the configuration switches
can't be carried forward from ./configure to make distcheck phase.

Thoughts?

Marcin


More information about the kea-dev mailing list