[OTR-dev] compilation on NetBSD too hard, but gaim plugin seems to work

[Greg Troxel]

  What is the usual NetBSD packaging mechanism?  Can you make one of
  whatever that is?  We can host it, and then NetBSD users will just
  have to install it.

See www.pkgsrc.org.  Control scripts build/install and then make
binary packages.  Yes, I can do that, and get the control bits in theb
NetBSD pkgsrc CVS repo, and then they can just type 'make package' and
have it compile.  There is no single binary package; netbsd runs on
tons of cpu architectures.  (Have you run otr on sparc64?  I wonder if
gaim works there even.)

But, a typical package's control file is small, and just runs
configure.  I would have to apply patches, and pkgsrc lets the user
control the prefix, so I'd have to pass that in as a make variable.

  This is definitely the way to go.  Does anyone know how to do this
  off the top of his/her head?

I could probably do this in a couple of hours.  I wouldn't want to do
it unless it would be accepted, since if not it's a waste of time.
Basically, one has to write a configure.ac to look for things, and
Makefile.am to build and install them.

One issue is the windows port.  If you use alternate build control
files on the same source, this should be ok.  autoconf enables all
sorts of HAVE_FOO tests, but right now you just have to have all the
foos that are used.  I don't know how gaim is built on windows.  If
cygwin, autoconf'd packages are fine.

  [I'd actually prefer to keep libotr as a static library, at least for
  now, if only to not have to worry about versioning issues as the library
  undergoes almost-certain changes at this early stage.]

I recall now that this isn't hard at all.

autoconf-ization will impose dependencies on autoconf, automake, and
libtool, but only for those compiling 'from CVS'.  released tarballs
won't have any extra dependencies.