[01:05] <lifeless> wgrant: sure
[12:05] <wgrant> cjwatson: So that branch isn't terrifying at all.
[12:16]  * cjwatson attempts to apply sarcasm detector to wgrant
[12:17] <StevenK> cjwatson: Surely that always returns true.
[12:17] <wgrant> No it doesn't.
[12:18] <StevenK> wgrant: Liar.
[12:19] <wgrant> I don't know why I'd be terrified of a branch that changes the behaviour of lots of scripts that haven't been touched since I refactored this four years ago :)
[12:19] <cjwatson> I've done a lot of grepping trying to think of patterns that might be problematic; direct use of test_emails, catching SMTPExceptions, aborted transactions, that kind of thing
[12:19] <wgrant> Yeah
[12:19] <wgrant> We'll need to do a lot of weird QA.
[12:19] <wgrant> But it's manageable.
[12:20] <cjwatson> Can you think of any other patterns I might have missed?
[12:20] <wgrant> I don't think so.
[12:20] <wgrant> The only other big risk I see is that some things might somehow break the email commit hook.
[12:21] <wgrant> Or not commit at all.
[12:21] <cjwatson> Well, if they don't commit at all then the operation being notified about doesn't happen either.  Unless they commit before sending mail
[12:21] <wgrant> Right, but errors might do that.
[12:22] <wgrant> Some errors abort because they don't LBYL
[12:22] <wgrant> but others will.
[12:22] <cjwatson> I'm definitely spooked by the oops/error handling in jobs, I don't really understand that yet
[12:23] <cjwatson> (which is why I applied the context manager to those for now)