On Wed, 2011-02-02 at 17:26 -0700, Tim Flink wrote:
On 02/02/2011 01:33 AM, Kamil Paral wrote:
Some of the tools are surely in the repos (mkrizek used nose, it's in Fedora), some of them might not. I believe our requirement is that we want to have AutoQA running on RHEL 5+ and supported Fedora's (13+). Because of RHEL 5.4, we currently limit our server code to Python 2.4 IIRC. But limitation could be soon lifted after we upgrade our production server to RHEL 6.
But this is really a question for James, let's wait for his reply. It would be surely best to use packaged tools. If the tool is not packaged and we need, we should make sure it's possible to have it packaged sooner or later (we will need to have everything in Fedora in the end).
As for supported RHEL releases (5 and 6 or just 6), I believe James have some plans for this.
Thanks for the info, I'll make sure that I'm paying attention to the EPEL repos for tool versions. I'll make sure to talk to the list before getting very far on anything, though.
Other than that, we'll see what James' thoughts are on RHEL versions.
+1 on using existing well established python test frameworks (nose or unittest). My initial thought is we have enough hassles managing content that is not yet packaged. I'd like to leverage existing framework documentation/howto's where possible.
As far as RHEL requirements, nothing comes to mind immediately other than ensuring python compatibility by way of pylint (or similar). We'll likely need to maintain a systemd unit, as well as a SystemV initscript. I'm not sure what unit tests would be required there. I'd like to see more unit tests for each of the watchers to validate expected outputs based on a controlled set of inputs. Since they inputs depend on online services (repodata, koji and bodhi), crafting some known inputs will be a fun exercise.
Thanks, James