Use updated headers from server state

[Kludex]

• Closes #1618

The idea here is to use the ServerState object instead of passing the server_state.default_headers to the <protocol>.default_headers attribute.

How to test it?

• Use the same snippet provided on #1618.

[Kludex]

You should probably add the test from the discussion so that we're sure it is ok in the future.

This said, I played a little with the test, the PR fixes the failed test, BUT, I find that as soon as the sleep is below 1s, the test fails again, any idea ?

Yeah, the on_tick function runs every second.

Ideas for testing this?

[euri10]

ok so just the test added will be fine I suppose: with a 1.1s sleep it should do the job, maybea add a note refering to the on_tick so we know that's a limitation and headers cannot be refreshed more that once per second because of this. Ideally on_tick should be configurable, but not too sure

[Kludex]

This doesn't work. I've set 1-second sleep, and I got 2 seconds difference.

I tried to play with mocking the asyncio.sleep(0.1), and using time-machine package, but none work properly.

[euri10]

works fine on linux, I just added prints, what mac is doing ???

============================= test session starts ==============================
collecting ... collected 1 item

test_default_headers.py::test_date_headers_update 

============================== 1 passed in 1.48s ===============================

Process finished with exit code 0
PASSED                 [100%]INFO:     127.0.0.1:42396 - "GET / HTTP/1.1" 200 OK
INFO:     127.0.0.1:42396 - "GET / HTTP/1.1" 200 OK
2022-09-15 03:08:36
2022-09-15 03:08:37
INFO:     Started server process [1871053]
INFO:     Waiting for application startup.
INFO:     ASGI 'lifespan' protocol appears unsupported.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)
INFO:     Shutting down

[Kludex]

By "I got" I mean the pipeline got. 😟

The job failing is with 2 seconds diff.

[euri10]

can we add microseconds on the CI just to check is it possible to reproduce locally, I have no mac >_)

[Kludex]

Why would the microseconds matter here? Also, the pipeline on mac also fails with timedelta < 1 :sweat_smile:

[Kludex]

Why would the microseconds matter here?

It's a real question, seeing the on_tick and main_loop I don't know how that would help (nor where I should be printing it).

[euri10]

I was thinking maybe there could have been a rounding effect in the difference leading to having a difference of 2 instead of one, But apparently there is no microseconds set in the headers so that wont help. However removing the test is probably not a solution, would be interesting to dig further to see why macos is doing that, it's waiting for one more seconds so there's likely somethig fishy going on

[Kludex]

However removing the test is probably not a solution, would be interesting to dig further to see why macos is doing that, it's waiting for one more seconds so there's likely somethig fishy going on

Not necessarily... As I said:

Also, the pipeline on mac also fails with timedelta < 1 :sweat_smile:

The pipeline is failing for >1 and <1 for Mac.

[euri10]

The pipeline is failing for >1 and <1 for Mac.

but locally it's ok ?

[Kludex]

The pipeline is failing for >1 and <1 for Mac.

but locally it's ok ?

I don't know. I've been using Ubuntu. I'll try on the Mac later this week.

[Kludex]

@euri10 Can you check it? It should be less flaky now.