Overview
Provide a single, well-known location with information about whether or not Rawhide is broken, and a link to the last known-good Rawhide tree.
Problem Space
Sometimes Rawhide is broken. Brokenness typically comes in one of these types:
- Installer
- Installer images missing
- Installer doesn't work
- Repo
- Metadata is missing
- Package/file conflicts
- Dependency problems
- Functional
- Major bugs with packages in the "Critical Path" (see Critical Path Packages Proposal)
When Rawhide is broken it's much harder for people to become Rawhide testers, harder for developers to get fixes into the hands of testers, and other such badness.
Proposed Solution
First, we should create test cases for each of the types of breakage described previously. These should be automated or scripted when possible.
These tests should be run regularly (daily if possible) by QA team members and other interested parties. The results should be submitted and displayed prominently on a public website. Kyle McMartin registered israwhidebroken.com for this purpose.
Second: currently we keep copies of several weeks' worth of daily Rawhide trees in the Koji build system. When a tree is deemed good (i.e. when it passes all the tests) someone from QA should file a ticket with rel-eng to create a "known-good" symlink for that rawhide tree. This link should be public and well-known. Someone suggested the name "chewtoy" for this known-good Rawhide.
Scope
- QA Team members (i.e. testers) will need to do regular (daily) testing
- rel-eng will need to move the symlink when requested
Active Ingredients
- Define requirements for Good vs. Broken
- Write test cases to check each requirement
- Set up a web app to allow testers to submit/view test results
- Publicize a link to the "chewtoy" known-good tree
- Hosting for israwhidebroken.com - or a redirect.
Roadmap
See https://fedorahosted.org/autoqa/milestone/israwhidebroken.com
Discussion Points
Need to list requirements / solicit mockups for the webapp.
Comments?
To leave a comment, use the Talk page for this proposal.