draft: switch to n0-qlog with support for latest drafts & multipath #2460
Conversation
This changes the open_path function to no longer return the PATH_ABANDON frame, but instead put it on the pending data. And adds functionality to send it in poll_transmit.
This changes the open_path function to no longer return the PATH_ABANDON frame, but instead put it on the pending data. And adds functionality to send it in poll_transmit.
This always accepts the status set by the peer, the draft says it is only a suggestion.
- PTO is computed for the right path - crypto keys update uses 3 * max PTO of all paths - draining state uses 3 * max PTO of all paths
- PTO is computed for the right path - crypto keys update uses 3 * max PTO of all paths - draining state uses 3 * max PTO of all paths
- [x] Basic idle timeout scaffolding. - [x] Handle `PathIdle` timer expiry. - [x] Basic keep alive scaffolding. - [x] Handle `PathKeepAlive` timer expiry. - [x] quinn-proto API for setting path idle and keepalive. - [x] quinn API for setting path idle and keepalive. - [x] Allow setting path idle & keepalive defaults in `TransportConfig`. The following might perhaps better flow over to another PR - [ ] Abandon the path correctly. - [ ] Close the connection if there is no remaining path.
Updates up to quinn-rs@8bf6825
Tests are currently fast, that's a feature we'd like to keep ideally. We can tune it if this ends up being flaky for slow CI hosts.
* feat: add vantage point to qlog traces * feat: record outgoing quic frames in qlog feat: add ConnectionStarted to qlog feat: track received packets in qlog fix: cid groups, padding size, cleanups fixup vantage_point method signature chore: clippy * address review! * fixup * refactor: use crate::Frame for recording sent frames in qlog * fixup * fixup * docs: fixup * chore: cleanup
Frando
requested review from
Ralith,
djc,
gretchenfrage and
mxinden
as code owners
|
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #2460 +/- ##
==========================================
- Coverage 77.23% 76.60% -0.64%
==========================================
Files 78 83 +5
Lines 18501 22759 +4258
==========================================
+ Hits 14289 17434 +3145
- Misses 4212 5325 +1113 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
|
Any particular reason folks keep opening misdirected PRs today? |
|
I assume because they are maintaining a fork: |
Yes, but they've been doing that for years and it happened twice in the same week that they opened a PR here... |
I blame GitHub's UX. It happened to me a couple of times with some forks. When you push a branch to your fork and click on the link returned from git's output, GitHub's default is to open a PR against upstream. You need to then manually select your own fork from a small drop-down instead. As far as I know, there is no way in GitHub to change that default which is annoying. |
10 participants
Switches from
qlogton0-qlog, which is a fork ofqlogwith these two PRs applied: