Don't add rejections to the state_group, persist all rejections #948
Conversation
|
Err, i take that back |
|
I can't speak to the correctness of the changes themselves, but in my experience, when stuff starts getting complicated, lots of comments explaining what is going on can really help. I'm surprised to see quite a lot of code here, with little attention to comments explaining why things are doing what they are doing. |
| @@ -442,6 +433,9 @@ def _persist_events_txn(self, txn, events_and_contexts, backfilled): | |||
| # Handle the case of the list including the same event multiple | |||
| # times. The tricky thing here is when they differ by whether | |||
| # they are an outlier. | |||
| if context.rejected: | |||
There was a problem hiding this comment.
this code appears to have landed between a comment and the code it refers to, thus making the comment confusing. Because this method wasn't confusing enough before, I suppose.
|
I guess the main changes here are: don't add rejected events to event forward extremities table (and don#t and them to the push actions). |
|
Although I'm slightly concerned that we have stopped storing various bits of information about the event, and how that affects other storage functions. I'd prefer if, for now, the changes were as minimal as possible so that we can get the fix out and have refactoring (and commenting) as a separate PR. |
…ts to explain what the code is doing mechanically
No description provided.