Follow-up from "fix some problems related to weston_log_context"
The following discussion from !358 (merged) should be addressed:
-
@pq started a discussion: (+2 comments) I think this should be the last thing after everything else about subscribers and scopes are destroyed, because they are created from/for/into the log context.
It looks like both
weston_log_subscriber_destroy_log()
andweston_log_subscriber_destroy_flight_rec()
are broken: both leakstruct weston_log_subscriber::subscription_list
members. I think the current code relies on all scopes to have been destroyed first, which is implied by destroying the log context first. So maybe leave this for another time.Ok, this MR is good as is here. If you want to fix the destruction order and those functions in another MR, that would be welcome.