Content-Length should be set in the request when using reactive WebClient [SPR-16949]

[spring-projects-issues]

Mirko Adebahr opened SPR-16949 and commented

As of #21431, the content-length header should be set in the request for Mono publishers when using the reactive WebClient. Still, as far as I can see the header is not set in the HTTP request.

This is what I think is the problem:

1. AbstractClientHttpRequest::doCommit: applyHeaders() is called as commit action -> Headers are applied to httpRequest in ReactorClientHttpRequest
2. After this, body.doOnNext(data -> headers.setContentLength(data.readableByteCount())) is called (see EncoderHttpMessageWriter::write)

The doOnNext seems to be executed too late to have an effect on the headers in the HTTP request.

---

Affects: 5.0.7

Issue Links:

• AbstractJackson2Encoder::getContentLength should not return null for MonoJust/FluxJust publisher [SPR-16892] #21431 AbstractJackson2Encoder::getContentLength should not return null for MonoJust/FluxJust publisher

Referenced from: commits 280da61, 6189e17, 4a26f93, d1c9401

[spring-projects-issues]

Brian Clozel commented

Hi Mirko Adebahr,

As of #21431, we've got integration tests covering this use case (see the changes attached to the commit fixing that issue).

Could you provide us with a small application that reproduces this issue?

Thanks,

[spring-projects-issues]

Mirko Adebahr commented

Hi Brian,

please correct me if I'm wrong, but if you're referring to RequestMappingMessageConversionIntegrationTests, as far as I can see all tests check the content-length header on the HTTP response. #21431 as well as the present issue was/is about the header not being set in the outgoing HTTP request.

I understand that an application for reproduction of this issue would be helpful, but we're talking about a client/server scenario. Maybe it would be sufficient if someone familiar with webflux could take a look and confirm the problem?

[spring-projects-issues]

Brian Clozel commented

Mirko Adebahr,

This is now fixed in 5.0.x and master (5.1.x) - new SNAPSHOT builds should be available shortly if you'd like to test it against your project before the next releases.

Thanks!

[spring-projects-issues]

Juergen Hoeller commented

Mirko Adebahr, please give this a try against a recent snapshot: ideally 5.0.8.BUILD-SNAPSHOT, available via Maven from https://repo.spring.io/snapshot... We'll release 5.0.8 in about three weeks, and it'd be great to know upfront whether it works for you now...

[spring-projects-issues]

Mirko Adebahr commented

OK, I'll try to test it until Wednesday.

[spring-projects-issues]

Mirko Adebahr commented

It's working, we're looking forward for the next release. Thanks!

[spring-projects-issues]

Brian Clozel commented

Thanks a lot for checking Mirko Adebahr, this is helping a lot.