On Fri, Jun 07, 2013 at 07:42:27PM +0200, Marc Espie wrote: > On Fri, Jun 07, 2013 at 11:00:12AM -0600, Luis Coronado wrote: > > Or http://openports.se/ > > I wouldn't recommend it. > > It still tries to parse the ports tree by hand, instead of using any number > of correct solutions like sqlports or dump-vars, so they get details wrong. > > You will end up with missing packages, erroneous flavor/subpackage > combinations, missing categories, etc.
Just to make the point: they don't try to figure out DESCR-* from the ports tree. So go look at gcc: they don't have *any* DESCR. Why ? because the DESCR* comes from the lang/gcc directory thx to a Makefile.inc. Or net/avahi, which comes from a handful of multi-packages, but mostly pseudo-flavors that command the existing flavors... well. Som of the most useful information, such as knowing whether a port is currently BROKEN, or some subpackage only exists for some arches, or won't be available as a binary package because of restrictions, is downright missing. The packing-list information is also pretty much useless, seeing as it's viewed in src form before variable substitution, nor does it include ANY fragments. it *could* interface to pkglocatedb and provide a "search by file". There's a list of MASTER_SITES, but it does not even show the HOMEPAGE (which is often WAYS more useful). ... nor does it include MASTER_SITES0-9, but then you knew I was going to say that. just go read the statistics: it says 6889 packages. my latest bulk built... 8141 packages. Well, the only thing it has that ports-readmes doesn't is explicit cvs history...
