On 17 May 2011 18:14, Michael Schubert <mschu.dev@gmail.com> wrote:
2011/5/17 Ray Rashif <schiv@archlinux.org>
That's the way it's done. You have those optdeps as makedeps so you can finish building the package.
Sorry if I was a bit unclear earlier. The make process does not need all binding languages as dependencies as cmake and configure take care of preventing those from being built. So the make works fine with no optional dependencies installed.
When you're distributing any modular package that, when installed with all usable dependencies, would provide full software functionality, someone (the builder/packager) has to initially configure and build the software with everything in place. * So, you first install all usable deps, build it, then remove the unimportant deps (thus making them makedeps + optdeps). If a dep cannot be removed (when a particular part of the software complains about something missing or does not work as expected), then it cannot be an optional dependency. Sometimes, in order to appease everyone, you must build and depend on deps that not everyone would need, because those deps are link-level or hardcoded once used (to build). . You don't have to build multiple packages just because of this, unless we're talking about a rather large software compilation/bundle. But yes, you may split the package in the end. If it is easier for you then build as many times within the PKGBUILD as needed, to for example a different directory each time. Then you can package separately and the result is a split package. Do this only if building in the optional support actually leads to unwanted bulk for those not needing the optional functionality. -- GPG/PGP ID: 8AADBB10