On Wed, Apr 02, 2008 at 09:10:09AM -0400, James Carlson wrote: > METADATA > > Just a nit, but it seems that Norm Jacobs is the only one following > this "new" format you're using. Everyone else has this: > > Owner: Joe User <email at domain.org> > License: MUMBLEFROTZ > > Notes on the porting job (or lack thereof) here. > > ... and the gate readme seems to singularly unhelpful in guidance. > Is there any rhyme or reason to what we're doing here? Or are these > files just for fun? ;-}
There's no real guidance for it. I've been forgetting to do them myself, and so I grabbed one from a component I knew had one. > 5: I think this should say "GPLv2," not just "GPL." (Assuming > that's the license in use. I don't really care, but apparently > legal is in a tizzy about the license versioning.) Fixed. > pkginfo.tmpl > > 40: (nit, and I understand why you wouldn't want to change it) I > wish we didn't put the version number into both the title and > the package version. This is handled haphazardly in SFW -- some > packages have DESC="... version" and others do not. It's a bit > of a mess. Where is the version number in the package version? It should only be in the description. I've been putting the component version number in when I think of it because it's useful to know, and otherwise there's no place to put it in the package metadata. The downside is that the metadata can't be patched. Which isn't really a problem here. > Makefile and prototype_com: > > Doesn't usr/demo/mercurial/hgwebdir.fcgi need to be executable in > order to be useful? Yes, but I'm shipping it just as it is in the mercurial distro. It would have to be moved out of /usr/demo, anyway. > Files that seem to be missing from prototype_com: > > usr/lib/python2.4/vendor-packages/hgext/color.py > usr/lib/python2.4/vendor-packages/hgext/color.pyc D'Oh! Thanks. > Other than that, I assume you've built this, created new packages, and > installed them locally for a test. I've done a local build of just mercurial, run the built-in test suite, and have started using it on my development box. I still need to do a full sfw build (which would have caught that I didn't remove the "old" directory from Targetdirs). I still need to work in the patch that Rich built up to fix 1052. I'll send out another code review when I have that figured out. Thanks, Danek
