Lars Eggert
Lars Eggert
> One way forward would be to merge this with a default to "on", so we can more easily flip that later. I did this now.
Closing in favor of #3129.
Hm. Transfer test regression, but one bench shows an improvement. Time to look at flamegraphs...
Hm. Simplifying `inbound_frame` drastically doesn't seem to make things slower.
So what we save in `memcpy` we spend on `malloc` now :-)
IIRC, any `STREAM_DATA_BLOCKED` means we are waiting for an RTT before we can send more data, which at high throughputs is quite a bit. Are you saying you are expecting...
> At least a few (server) implementations support QUICv2 already (e.g., [ngtcp2](https://github.com/ngtcp2/ngtcp2?rgh-link-date=2025-04-08T07%3A47%3A36.000Z#quic-protocol-extensions) and [lsquic](https://github.com/litespeedtech/lsquic?rgh-link-date=2025-04-08T07%3A47%3A36.000Z#quic-protocol-extensions)). Firefox ([neqo](https://github.com/mozilla/neqo)) supports QUICv2 as a browser, too.
I'm testing a fix to #224 now, and will send a PR if that works. But even if that works, I have the remaining issue that I can't use a...
Hm. Worse than the default allocator.
The issue is that the log init call in the `qdebug` macro does not execute in test code.