Fix funny generics in request size limit test

[nik9000]

No description provided.

[nik9000]

@danielmitterdorfer I was trying to debug a non-reproducable failure I had on this test locally (no node available while doing after test cleanup) and I found some funky stuff going on with the generics. Would you mind having a look?

[danielmitterdorfer]

I think the better option is to add the '@SafeVarargsannotation toNettyHttpClient#post()(which requires the method to befinalbut I'd prefer to make even the classfinal` as it is not intended for inheritance).

[danielmitterdorfer]

@nik9000 I wonder why the problem appeared in the first place. Do you use Eclipse? I did not have any problem in IDEA and on the command line with Oracle JDK 1.8.0_74-b02.

I left one comment, otherwise LGTM.

Can you please backport this to 2.x too as I just pushed the backport of #17133 where this change comes from?

[nik9000]

I wonder why the problem appeared in the first place. Do you use Eclipse?

Yes I do, but Eclipse didn't catch it. I add type parameters to things that are missing them when I touch code out of habit and that caused a chain reaction. I suspect the that I hit is something you've already fixed by raising the request buffer but it never reproduced locally for me.

[nik9000]

2.x: 9486bb0 08bd55b