Sébastien Deleuze

Please create a related issue with me mentioned with a minimal reproducer and a detailed description of the type hint inference you would expect.

Looks like a duplicate of #29663.

This is more complex than adding a `ContinuationArgumentResolver` since this involves the proxied invocation of a suspended function, I think it depends on #22462.

Superseded by the PR #29527.

I confirm @aooohan analysis, Undertow handling of multiple cookies with the same name looks incorrect. They may have mistakenly interpreted "e.g., that were set with different Path or Domain attributes"...

> I'm not a particular huge fan of the [TechEmpower benchmarks](https://www.techempower.com/benchmarks/#section=data-r21), because they barely reflect real use-cases @dreis2211 Same, but as I am myself working on some benchmarks to compare...

Interesting, better to focus initially on JSON and plaintext indeed. It would be interesting to do for example a new round of JSON test with only `MappingJackson2HttpMessageConverter` and remove the...

@dreis2211 Pretty good results on skipping `instanceof GenericHttpMessageConverter` checks, see #29438. I was able to confirm the gains on 2 different benchmarks, but if you can check on your side...

As you may have seen, more benchmarks on https://github.com/spring-projects/spring-framework/pull/29438 did not achieve to reproduce in a reliable way the gains I initially saw, I think it would maybe more visible...

Great, thanks a lot for your various contributions. Could be great if TechEmpower benchmark could be updated with Spring Boot 3.0.x, explicit content type and compare with the previous figures....