-
Notifications
You must be signed in to change notification settings - Fork 179
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Ensure deferred #1565
base: master
Are you sure you want to change the base?
Ensure deferred #1565
Conversation
8fe3303
to
9696925
Compare
9696925
to
e550051
Compare
It may actually fix some random CI failure of |
@AdamISZ Could you take a look on this? |
tACK e550051 However, a caveat is that I can't remember if or why it's helpful to make this change? This comment in #1562:
I don't have a preference for So to be clear, I don't really know the reason for this PR, but, there is no change in test functionality, and, it works (according to my tests). So if others want to merge it, I will not object (hence the ack-). |
If this is true, that's enough reason to merge for me. |
Oh, I was going to write that myself, as I had a memory of that, but then I couldn't find the quote, so I didn't include it. Indeed, although, I'm not sure why that would be the case. If it is @roshii , can you remind us why? |
Reason for this PR is best stated by this post, see
I did not find any explanation for this changed behavior. I did notice that some test happens to timeout after 120 s with old syntax. And timing from this Digging for it though, I found other case of |
@roshii thanks for these thoughts, lots to chew on!
That was an interesting read. Some of the discussion (including the follow up comments) does mention, though, that trying to mix the two is pretty problematic.
That case is interesting. I may be reading it wrong, but i think it shows that the timeout is occurring during wallet sync (see lines 314-315); 120s is after all an absurd amount of time, especially in a test setup, for that. I was going to say "maybe merging #1594 will help", but, well, it might, but we should never need 120 seconds there, so I'm less inclined to worry about if there's a default timeout lurking somewhere because of that. About the long timings: there are definitely some inefficiencies lurking in the test suite, but indeed, I have found some tests take an age to run, part of it, I think, is download and install is sometimes crazy slow. I see 10 minutes used up by that, in the 19 minute run you linked. |
Mixing can be problematic indeed, which is probably why it was not working when I first wrote test, mixing syntax. Therefore this PR, changing the whole unit test suite to
There are good arguments for default timeout, 120 s is indeed absurd. In my opinion though, it should not cause test to fail, code is working as expected. #1594 may help but if Github CI infrastructure is under pressure it may just go as slow and fail.
From what I have seen 10 min is rather standard for the setup phase. On the other hand, more than 12 min is unusual for the test phase. Regardless, I am perfectly fine at leaving the PR as draft for now, gather more data to assess if this helps with test timeout, maybe expand its scope to a full syntax migration, or simply close it if we are just happy with Twisted Deferred. Suggestions welcome naturally. |
e550051
to
2d15ae8
Compare
rebased |
Following #1562 (comment) this PR implements coroutines with async and await syntax.