Channel.publish sometimes returns None after exhausting retries

[sbdchd]

When trying to figure out the failure cases for Channel.publish, I found that
if the method receives a series of 500 requests greater than the retry count,
the method won't raise an error and will instead return None.

This means that to check for all errors with Channel.publish we need to write:

channel = ably.channels.get(channel)
res = channel.publish(...)
if res is None:
    raise requests.ConnectionError

Instead of:

channel = ably.channels.get(channel)
channel.publish(...)

I think the method should be updated to raise an exception after
exhausting all retry attempts.

Something like:

diff --git a/ably/http/http.py b/ably/http/http.py
index 6e7bbbb..ea8067b 100644
--- a/ably/http/http.py
+++ b/ably/http/http.py
@@ -217,6 +217,7 @@ class Http:
                 except AblyException as e:
                     if not e.is_server_error:
                         raise e
+        raise AblyException(message="", status_code=0, code=0)
 
     def delete(self, url, headers=None, skip_auth=False, timeout=None):
         return self.make_request('DELETE', url, headers=headers,

┆Issue is synchronized with this Jira Bug by Unito

[paddybyers]

Hi - thanks for investigating and raising this. We'll have a look as soon as there's someone around.