Nils Goroll

I have updated the PR based on the new preamble-in-endpoint code by @bsdphk I will rewrite the description to reflect the current state of affairs.

FTR, after some discussion about the last comments: * @Dridi is right about test coverage which could (always) be improved * As he pointed out himself, I think using a...

notes from irc: @Dridi has warmed this topic. I would appreciate if @bsdphk could have another look how we could make iterative progress with this. As mentioned before, sensible suggestions...

> We would like to see explicit coverage of via backends connection pooling to be certain of the pooling behavior when multiple backends share the same via backend. Test case...

@bsdphk docs added in 5ee551f2a9bed94a4c037c9667ae58fa702417a8

> I would rather artificially prevent this behavior first and consider allowing it when we have a strong case for it. @Dridi I originally intended this to work, but have...

@Dridi apologies for my bad wording: I meant to say that I originally intended this to work (this PR is last in a series of iterations) and the current implementation...

My suggestions to go forward: * turn `hash_data(x)` into `req.hash.add(x)` (see below regarding vcl versions) * introduce a VRT (runtime) `freeze` type method to prevent changes to objects, e.g. ```...

If we had the log with the 165 failures, we might be able to tell more about what could be going wrong. The single m00000.vtc failure seems to be related...

Options discussed during bugwash: * have vmodtool detect the clash and use the double underscore only if a clash exists. * Expand `VPFX()` into `VPFX_FUNCTION()` and `VPFX_METHOD()` * do not...