Fix a stupid misstake where the segment counter in the VSL segment was
set just before the entire struct was cleared during
initialization. This didn't cause problems in the common case because
it would be set correctly on the very first log line produced. But it
did open a race where an unfortunate early log client would assume it
was lagging indefinetely behind varnishd on the log reading and report
overflow error. This was most commonly experienced with varnishtest.

Fixes: #2241

have to test with 16k frame sizes.

in What's New for 5.1.

This was the rollback of the fix for #1206 brought about in #2042.

Remains of copy/pasta made that v2 listened with PROXY protocol which is
not needed. It's a passive instance that only does health checking using
a probe.

not as smart as LLVM about const.

I guess if you can't say anything else good about it, H2 is forcing
me to invent new ways to implement table-driven object-oriented programming.

which per-frame flags only get sent on the last frame of a
continued frame-sequence.

This rolls back the fix for #1206 -- the decision is made regardless
of the pass state of the backend fetch. So delivery sets resp.status
to 304 at ESI level 0 iff the conditions in req.http.{INM,IMS} and
resp.http.{ETag,Last-Modified} allow it. To restore the behavior
from #1206, remove the If-* headers in vcl_pass.

Fixes #2042

Fixes #2151

It consists in the source and destination being the same socket address.

Refs #2151