Code review comment for lp:~mterry/deja-dup/no-python2

Michael Terry (mterry) wrote :

@didier, thanks for looking at this!

re: tests, guh! LP doesn't see the failures and I don't see the failures. I don't know what you and Ken have going on. (timezones? locale? -- I will look for something along those lines that could fail the tests).

Regarding your broader point about this approach:

- Yes, duplicity is the only supported backend.

- We don't plan to demote duplicity, we'll leave it in main. It will still be supported. This is just a step on the path to python3-only-in-main. Eventually we will have to port it. (to keep it off the image, we'll move it to a Suggests for deja-dup -- and let the other derivatives know that they need to seed duplicity if they want to keep shipping it on their own images)

- Landscape is a precedent for shipping a hook for functionality that then installs the rest.

- I wasn't going to be able to commit the time necessary to port duplicity to python3 for 16.04. So it's either this or dropping deja-dup unless a volunteer does it -- patches welcome... ;) And I think keeping a user interaction point to nudge people towards backups is better than dropping it from the image.

- We aren't lying to ourselves so much as iterating and reducing the scope of python2 installs. It's still in main. We're working on that over time. And maybe if duplicity remains python2, we'll eventually drop deja-dup. But for 16.04, we're shooting for the base install not having python2 anymore (which is not nothing -- CVEs & bugs affecting python2 will no longer affect every single Ubuntu user; only some of them).

- Note that this MP itself doesn't actually change anything in Ubuntu. This debian/ packaging is just for the upstream PPA (and doesn't even drop duplicity to a Suggests). Your broader concerns are really about the planned followup MP for the deja-dup packaging that updates to this code and then will drop duplicity to a Suggests for Ubuntu. But talking about them here is fine and good. But maybe shouldn't be a blocker for *this* change.

« Back to merge proposal