Lars Eggert
Lars Eggert
I'm going to close this, since we removed a lot of related `unwrap`s.
> I'm not seeing PMTUD tests, which would be necessary for this. Yeah, tests are definitely still needed. > There is probably a case for sending probes when you have...
On a positive note, look at the benchmark improvements! Looks like our crappy performance against msquic was mostly because we didn't send 64KB packets over loopback :-)
Making this a draft until the test failure is fixed.
Working on increasing test coverage.
This can wait until after.
This is the crash https://github.com/nss-dev/nspr/blob/dead7133d4535bf4fade1814dd8c261c01b76635/pr/src/io/prlayer.c#L619 Called from https://github.com/mozilla/neqo/blob/36fae6282b2214e4fea425ee4a952c08acf1c445/neqo-crypto/src/agent.rs#L329
Patch to NSPR in https://bugzilla.mozilla.org/show_bug.cgi?id=1890927
Good topic to discuss on the next call. I don't have history of why the current state is as it is, but I'd personally like to reduce duplication as much...
Still happening, e.g., https://interop.seemann.io/logs/2024-07-18T01:39/quic-go_neqo/zerortt/output.txt