Next steps to get RPM Fusion running (V2)

[Adrian Reber]

On Tue, Nov 04, 2008 at 08:21:39PM +0100, Michael Schwendt wrote:
> On Tue, 4 Nov 2008 14:40:39 +0100, Adrian Reber wrote:
> 
> > > But I think I know why: The Debug directory in  
> > > http://download1.rpmfusion.org/free/fedora/development/i386/
> > > was missing until yesterday, as our push script puts the debuginfo files to
> > > http://download1.rpmfusion.org/free/fedora/development/i386/os/debug
> > > I created a symlink in the place where mirrormanager likely expect it to  
> > > be. That way the directory structure also matches the one from Fedora.
> > 
> > Unfortunately symlinks are not good enough. Could the push
> > scripts be changed so that the layout matches the one from Fedora?
> > Or hardlinked files would also work...
> 
> I would advise against using too many manually created links. Better patch
> the scripts to support the "os" subdir where necessary (or make the repo
> paths more customisable).

There is a special function in mirrormanager called
trim_os_from_dirname(), which removes anything including "/os" from the
path string before starting to guess to which repository this path might
belong. I would really prefer if I do not have to provide an extra code
path just for the rpmfusion development release. From what I see on my
Fedora mirror the debug directory is never under the os directory and
for rpmfusion this is only the case for the development directory.

Is it possible to change the push script to create the debug directory
next to the os directory and not below it? Is there a reason why it
creates the debug directory at another place for the development
repository?

> The old default is
> 
>   $treeroot/$target/$basearch
>   $treeroot/$target/$basearch/debug
> 
> with no "os" subdir. That matches Fedora (Updates, Test Updates, EPEL,
> Extras). The "os" subdirs seems to be specific to the "releases" and
> "development" trees, where the repodata path must be constructed
> differently (e.g. $treeroot/$target/Everything/$basearch/os).

		Adrian