also detect gnusocial instances w/o nodeinfo

backend/lib/backend/crawler/crawlers/gnu_social.ex

@@ -8,8 +8,15 @@ defmodule Backend.Crawler.Crawlers.GnuSocial do
   @behaviour ApiCrawler
 
   @impl ApiCrawler
-  def is_instance_type?(_domain, nodeinfo_result) do
-    nodeinfo_result != nil and Map.get(nodeinfo_result, :instance_type) == :gnusocial
+  def is_instance_type?(domain, nodeinfo_result) do
+    if nodeinfo_result != nil do
+      Map.get(nodeinfo_result, :instance_type) == :gnusocial
+    else
+      case get_and_decode("https://#{domain}/api/statuses/public_timeline.json") do
+        {:ok, statuses} -> is_list(statuses)
+        {:error, _other} -> false
+      end
+    end
   end
 
   @impl ApiCrawler

backend/lib/backend/scheduler.ex

@@ -84,7 +84,7 @@ defmodule Backend.Scheduler do
     # We want the earliest sucessful crawl so that we can exclude it from the statistics.
     # This is because the first crawl goes up to one month into the past -- this would mess up the counts!
     # The statistics from here assume that all statuses were written at exactly the crawl's inserted_at timestamp.
-    earliest_successful_crawl_subquery =
+    earliest_crawl_subquery =
       Crawl
       |> group_by([c], c.instance_domain)
       |> select([c], %{
@@ -94,7 +94,7 @@ defmodule Backend.Scheduler do
 
     instances =
       Crawl
-      |> join(:inner, [c], c2 in subquery(earliest_successful_crawl_subquery),
+      |> join(:inner, [c], c2 in subquery(earliest_crawl_subquery),
         on: c.instance_domain == c2.instance_domain
       )
       |> where(