Improve flush-list search by avoiding skipping too many cache-keys. #120
Conversation
As we're using `new_keys` to build up `flush_keys` and `search_keys` we still might need the proper keys in `obj_keys` to avoid objects not being invalidated properly.
|
This essentially replaces #106 |
|
Hi @EnTeQuAk , my apologies for the delay on this. Is this still something you're interested in getting in? If so, could you rebase this onto master and add a test for the issue? The test suite should be passing now; if it's not for you, please let me know and I'll see what I can do to help. |
|
Seeing this project being active again is nice! I'll try to remember what I did and add some tests throughout this week, thanks for the follow-up! |
|
Sounds good, thanks! |
|
I looked at the change here vs what's in #106. I don't think this one is correct - should a flush key be hit twice, then it will end up in both the obj_keys and flush_keys. I think hitting the same flush key second time should result in a NOP, i.e. I did that by having a separate check: if key in flush_keys: |
|
Closing in favor of #127 |
As we're using
new_keysto build upflush_keysandsearch_keyswe still might need the proper keys in
obj_keysto avoid objectsnot being invalidated properly.
There are some tests missing currently unfortunately but even without the patch the test-suite is failing for me locally so it's quite hard to reason about any particular changes. This should be straight forward enough though.
Let me know what you think :)