[18:41] testing bot [18:47] 1 more test [18:54] Dmitrii-Sh: welcome! [18:54] ubuntulog3: also welcome [18:54] jlosito: thanks [18:55] np === Chipaca changed the topic of #smooth-operator to: general discussion of canonical's operator framework || github.com/canonical/operator [19:11] phew [19:12] woop, everything works ;-) [19:13] * Chipaca goes to make dinner [19:13] I'd just say "on the operator framework" [19:13] :P === Chipaca changed the topic of #smooth-operator to: general discussion of the operator framework || github.com/canonical/operator [19:13] niemeyer: like so? [19:13] Thanks [19:14] i'm sure there's a way of just telling chanserv to do that [19:14] instead of op/topic/deop [19:14] anywhoo, dinner-making [19:15] Have fun! [19:15] It's already beer o'clock here.. [19:15] and debugging something.. but at this stage any findings are a bonus :) [19:22] no beer here because it's a light pre-yoga dinner [19:22] might have a beer after yoga; can't be _too_ healthy [19:22] I'm sure your zeneth capabilities would be much improved [19:23] Not so sure about balance though [19:23] i could pretend it's about that, but it's just trying to keep my back un-messed-up [20:16] niemeyer, did we think already about the following situation? 1. the event is deferred, it's `.snapshot` is called; 2. the charm is upgraded, the event code itself changes! 3. the event is restored, a new code needs to handle previous "format" [20:16] I mean, are we just leaving that consideration to the developer, providing any guide? any structure? [20:19] Facu: Yeah, we don't expect devs to be using snapshot directly very often [20:19] Facu: We have the StoredState mechanism for that [20:19] People can just do _stored = StoredState() at class level [20:20] And only store there the data they want to persist [20:20] but everytime you define an Event you need to provide its snapshot/restore pairs, right? [20:20] Naming is important precisely to highlight that this is persisted [20:20] (thus the choice of "_stored") [20:20] Facu: Every time you define an event *type* that needs parameters [20:21] perfect [20:22] Facu: We should probably make EventBase a child of Object, so StoredState would work there as well [20:22] But I haven't thought through the consequences of that