Add final boundary for multipart entities with no parts #1257

[arnohaase]

Fixes #1257

[akka-ci]

Can one of the repo owners verify this patch?

[ktoso]

OK TO TEST

[akka-ci]

Test PASSed.

[arnohaase]

I signed the Lightbend CLA,

[jrudolph]

Hi, @arnohaase, thanks for the contribution.

The change seems reasonable, though it is unclear to me if multipart entities without any part would (or should) be allowed at all? Do have any reference for that?

[arnohaase]

The spec does not specifically forbid it, at least as far as I can see. But it does specify that the final boundary must always be present. https://www.w3.org/Protocols/rfc1341/7_2_Multipart.html

The problem appeared in a real-world problem where a multipart response was used to represent documents matching a search, with "no parts" being a valid response document.

Btw, MultipartUnmarshallers.defaultMultipartGeneralUnmarshaller can handle Multipart entities without entries already, provided the final boundary is present.

[raboof]

Looks good, especially since the change is only in strict and it looks like it already behaved that way for streamed.

It'd be nice to have an easy way to execute the test both in 'strict' and 'streamed' mode, do we have a way to do that?

[jrudolph]

@arnohaase

The spec does not specifically forbid it, at least as far as I can see.

It says:

In the case of multiple part messages, in which one or more different sets of data are combined in a single body

I understand it that "one or more" wouldn't include zero. But I guess that's only a nitpick.

I agree we should do it consistently for marshallers and unmarshallers and also for strict and streamed. Would be nice to have tests being run for strict and streamed variants. Is that something you would like to work on, @arnohaase or @raboof?

[raboof]

@jrudolph I think that'd be interesting but am not planning to do it right now - perhaps we should merge this and keep that idea for later? Created #1309 to track it.

[jrudolph]

Yep, sounds good.