Fix Event-deletion from the admin

b/c of the introduction of on_delete=DO_NOTHING, this was broken.
this is the quick & dirty fix (no tests)

events/admin.py

@@ -1,11 +1,17 @@
+import json
+
 from django.utils.html import escape, mark_safe
 from django.contrib import admin
+from django.views.decorators.csrf import csrf_protect
+from django.utils.decorators import method_decorator
 
-import json
+from bugsink.transaction import immediate_atomic
 
 from projects.admin import ProjectFilter
 from .models import Event
 
+csrf_protect_m = method_decorator(csrf_protect)
+
 
 @admin.register(Event)
 class EventAdmin(admin.ModelAdmin):
@@ -90,3 +96,28 @@ class EventAdmin(admin.ModelAdmin):
 
     def on_site(self, obj):
         return mark_safe('<a href="' + escape(obj.get_absolute_url()) + '">View</a>')
+
+    def get_deleted_objects(self, objs, request):
+        to_delete = list(objs) + ["...all its related objects... (delayed)"]
+        model_count = {
+            Event: len(objs),
+        }
+        perms_needed = set()
+        protected = []
+        return to_delete, model_count, perms_needed, protected
+
+    def delete_queryset(self, request, queryset):
+        # NOTE: not the most efficient; it will do for a first version.
+        with immediate_atomic():
+            for obj in queryset:
+                obj.delete_deferred()
+
+    def delete_model(self, request, obj):
+        with immediate_atomic():
+            obj.delete_deferred()
+
+    @csrf_protect_m
+    def delete_view(self, request, object_id, extra_context=None):
+        # the superclass version, but with the transaction.atomic context manager commented out (we do this ourselves)
+        # with transaction.atomic(using=router.db_for_write(self.model)):
+        return self._delete_view(request, object_id, extra_context)

events/models.py

@@ -8,12 +8,15 @@ from django.utils.functional import cached_property
 
 from projects.models import Project
 from compat.timestamp import parse_timestamp
+from bugsink.transaction import delay_on_commit
 
 from issues.utils import get_title_for_exception_type_and_value
 
 from .retention import get_random_irrelevance
 from .storage_registry import get_write_storage, get_storage
 
+from .tasks import delete_event_deps
+
 
 class Platform(models.TextChoices):
     AS3 = "as3"
@@ -282,3 +285,13 @@ class Event(models.Model):
         return list(
             self.tags.all().select_related("value", "value__key").order_by("value__key__key")
         )
+
+    def delete_deferred(self):
+        """Schedules deletion of all related objects"""
+        # NOTE: for such a small closure, I couldn't be bothered to have an .is_deleted field and deal with it. (the
+        # idea being that the deletion will be relatively quick anyway). We still need "something" though, since we've
+        # set DO_NOTHING everywhere. An alternative would be the "full inline", i.e. delete everything right in the
+        # request w/o any delay. That diverges even more from the approach for Issue/Project, making such things a
+        # "design decision needed". Maybe if we get more `delete_deferred` impls. we'll have a bit more info to figure
+        # out if we can harmonize on (e.g.) 2 approaches.
+        delay_on_commit(delete_event_deps, str(self.project_id), str(self.id))

events/tasks.py

@@ -0,0 +1,53 @@
+from snappea.decorators import shared_task
+
+from bugsink.utils import get_model_topography, delete_deps_with_budget
+from bugsink.transaction import immediate_atomic, delay_on_commit
+
+
+@shared_task
+def delete_event_deps(project_id, event_id):
+    from .models import Event   # avoid circular import
+    with immediate_atomic():
+        # matches what we do in events/retention.py (and for which argumentation exists); in practive I have seen _much_
+        # faster deletion times (in the order of .03s per task on my local laptop) when using a budget of 500, _but_
+        # it's not a given those were for "expensive objects" (e.g. events); and I'd rather err on the side of caution
+        # (worst case we have a bit of inefficiency; in any case this avoids hogging the global write lock / timeouts).
+        budget = 500
+        num_deleted = 0
+
+        # NOTE: for this delete_x_deps, we didn't bother optimizing the topography graph (the dependency-graph of a
+        # single event is believed to be small enough to not warrent further optimization).
+        dep_graph = get_model_topography()
+
+        for model_for_recursion, fk_name_for_recursion in dep_graph["events.Event"]:
+            this_num_deleted = delete_deps_with_budget(
+                project_id,
+                model_for_recursion,
+                fk_name_for_recursion,
+                [event_id],
+                budget - num_deleted,
+                dep_graph,
+                is_for_project=False,
+            )
+
+            num_deleted += this_num_deleted
+
+            if num_deleted >= budget:
+                delay_on_commit(delete_event_deps, project_id, event_id)
+                return
+
+        if budget - num_deleted <= 0:
+            # no more budget for the self-delete.
+            delay_on_commit(delete_event_deps, project_id, event_id)
+
+        else:
+            # final step: delete the event itself
+            issue = Event.objects.get(pk=event_id).issue
+
+            Event.objects.filter(pk=event_id).delete()
+
+            # manual (outside of delete_deps_with_budget) b/c the special-case in that function is (ATM) specific to
+            # project (it was built around Issue-deletion initially, so Issue outliving the event-deletion was not
+            # part of that functionality). we might refactor this at some point.
+            issue.stored_event_count -= 1
+            issue.save(update_fields=["stored_event_count"])