On Thu, Dec 2, 2010 at 9:55 AM, Ray Rashif <[email protected]> wrote:
> On 2 December 2010 22:01, Kaiting Chen <[email protected]> wrote: > > A little tangent but from this page it seems to me that a '-git' or > '-svn' > > suffix should only be applied when there is a version of the package > without > > that suffix in the name; this is to differentiate between the 'stable' > and > > the 'development' version of the same package. > > > > https://wiki.archlinux.org/index.php/VCS_PKGBUILD_Guidelines > > > > In his AUR page there are some packages with '-svn' or '-git' suffixes > that > > do not have non-suffixed counterparts. Is this correct? I would like to > > update that wiki page to explain the convention more clearly. --Kaiting. > > OK, let me get this right. You mean that when for eg. a software only > has a development source tree and no tarball, it should just be > 'package' and not 'package-vcs'? > > If so, I don't think that would be proper. If a PKGBUILD fetches > development sources, it should have a development suffix. However, > exceptions can be made sometimes. > > Personally, I know of at least one upstream that does not directly > offer a tarball, but instead has (or rather had) an SVN tag that > distributors could check out. This package would then be named without > a vcs suffix. > My original view had been that a package would be simply called 'package' regardless of whether or not a source tarball was offered. Then if someone makes a version that builds against upstream VCS, that package would be called package-vcs. In light of this new discussion however, I feel like the proper policy is to name a package without a suffix if there is a 'versioned release', no matter where this comes from (source tarball, vcs tag, etc.). Then the converse is that if a package has *no release* but just a rolling development trunk, then it is given a suffix. --Kaiting. -- Kiwis and Limes: http://kaitocracy.blogspot.com/
