SGUG RSE Repo discussion

[Unxmaal]

Here's my proposed RSE repo structure. Tell me how it sucks.

.
└── 0.0.5
    ├── repodata
    └── base
    |   ├── RPMS
    |   └── SRPMS
    └── sp5
    |   ├── RPMS
    |   └── SRPMS
    └── wip
        ├── RPMS
        └── SRPMS

In this example

• base contains the rpms from sgug-rse-rpms-0.0.5beta.tar.gz
• sp5 contains the rpms from the latest service pack
• wip contains all wip rpms
• repodata contains actual real repodata that will work with yum (when we get it ported)

My process so far, working on wip, is

• copy the packages I've built from ~/rpmbuild/RPMS to my workstation
• sync them to S3:

  • aws s3 sync . s3://sgi-dev-public/repo/  --profile sgug --delete

 

[hammy]

Seems quite reasonable.

Some comments of the top of my head.

1. It would be nice that we retain the original fedora SRPM - once we've butchered the spec and disabled patches, they won't appear in our SRPM - and at some point in the future we want to "go back" - we'll need the original fc31 SRPM
2. I'm not sure what the utility of having WIP per release is - IMHO WIP is an eternal, ongoing release where stuff is constantly superceded - perhaps you can explain a bit your thinking there

Ideally, I'd like that we have some way where things in WIP are expired (maybe on check in of newer version, maybe only explicitly).

It's a little tricky, as we don't (currently) have any discipline in WIP about version numbers, which makes things a bit of a mess.