core(network-recorder): handle QUIC requests #5256
Conversation
| @@ -107,9 +107,15 @@ class NetworkRecorder extends EventEmitter { | |||
| return; | |||
| } | |||
|
|
|||
| // QUIC network requests don't always "finish" even when they're done loading data | |||
| // Use receivedHeaders instead, see https://github.com/GoogleChrome/lighthouse/issues/5254 | |||
| const isQUIC = record._responseHeaders && record._responseHeaders | |||
There was a problem hiding this comment.
would be nice pulled out into a function to separate determining if request is QUIC vs what to do with QUIC requests
There was a problem hiding this comment.
it could also be a function that's like isQuicAndFinished(record) to absorb logic below, too :)
| // QUIC network requests don't always "finish" even when they're done loading data | ||
| // Use receivedHeaders instead, see https://github.com/GoogleChrome/lighthouse/issues/5254 | ||
| const isQUIC = record._responseHeaders && record._responseHeaders | ||
| .find(header => header.name.toLowerCase() === 'alt-svc' && /quic/.test(header.value)); |
There was a problem hiding this comment.
how certain is the 'alt-svc' check? Any other way to tell?
There was a problem hiding this comment.
seems to be fairly certain, in the spec it's not technically required to announce it buuuuuut...
While HTTP/QUIC doesn't formally require a client to implement Alt-Svc, there's no discovery mechanism other than Alt-Svc provided, so you're not going to get very far without
it.
| @@ -119,7 +125,7 @@ class NetworkRecorder extends EventEmitter { | |||
| .sort((a, b) => a.time - b.time); | |||
|
|
|||
| let numInflightRequests = 0; | |||
| let quietPeriodStart = 0; | |||
| let quietPeriodStart = -Infinity; | |||
There was a problem hiding this comment.
what's the advantage of a [-Infinity, 0] period? Time starting at 0 also seems like a reasonable state of things
There was a problem hiding this comment.
that's fine just made tests have [0, 0] quiet period so we either handle the empty period case, move test cases to past 0, or -Infinity, -Infinity seemed like most reasonable conceptually but I do not have strong feelings
There was a problem hiding this comment.
it may be fine, and I haven't spent enough time here recently to have a good gut instinct, but I think I would expect the earliest quiet period to start at 0.
I guess I don't understand what this is addressing. Seems independent of the QUIC change?
|
is this something we should also fix in the upstream |
DevTools frontend handles this case correctly, I didn't dive into the code there to see how since its ~2 years or so diff :) |
closes #5254