On Mon, May 23, 2016 at 12:31:36PM +0000, Emmanuel Dreyfus wrote: > On Mon, May 23, 2016 at 08:04:08AM -0400, Kaleb S. KEITHLEY wrote: > > My point is that it's not really clear ? to me ? that that's the best > > (or most practical) choice. I'd like to hear from Patrick and Emmanuel. > > My two cents: the canonicalway of releasing software is to run > autotools on it before releasing so that reelased tarball contains > no autotools-related dependency. > > I am not sure of what happened with config.sub here, but we are lucky > since we can just add it as empty file as a second best solution.
The problem is that the releases are made on a RHEL/CentOS-6 system. The version of autotools is pretty old, and those scripts do not provide support for new architectures/platforms. Not all build environments require the config.{gues,sub} scripts, depending on the parameters that get passed to ./configure. If NetBSD works with empty scripts, the script may even not get executed at all. The patch at http://review.gluster.org/14503 replaces those scripts with one that writes a message and does a 'exit 1'. Could you test that on a NetBSD environment and verify if building the port still works fine? You can get a tarball with the change from here: https://devos.fedorapeople.org/tmp/glusterfs-3.9dev.tar.gz Thanks, Niels
signature.asc
Description: PGP signature
_______________________________________________ maintainers mailing list maintainers@gluster.org http://www.gluster.org/mailman/listinfo/maintainers