Forgotten when compatibility for autoconf < 2.63b was added.

and V1F, it has nothing to do with the VDI layer.

e.g. <undef> >= 0.1 should fail for being an invalid comparison

the director has it.

Make VDI_GetBody() and VDI_Finish() operate on the implied bo->director_resp.

IP numbers.

This reverts commit 468a0472.

The next commit will make obj.hits reliable

The common method to check obj.hits is racy: While the update to
the hits counter happens under a lock, a hit could have happened
before the missing request gets to read obj.hits.

... by example of the backend object flags

Outcome of a discussion of the suggestion to treat empty headers like
non-existing for Vary processing:

https://www.varnish-cache.org/lists/pipermail/varnish-dev/2015-January/008131.html

req->wrk->stats->cache_hit only gets reset when workers go to idle
and sum up stats.

This additional space-wasting byte-flag will get collapsed with the
other space-wasting flags soon, it is just added temporarly to separate
commits.

Inspired by fix from JonathanHuot@github

problem, it may be a platform thing, in which case Jenkins should
go Technicrome after this commit...3...2...1...

found at: http://www.open-mpi.org/community/lists/devel/2010/09/8472.php

Fixes: #1378

Fixes: #1648

The streaming fetch of an object can fail between lookup time and when
we actually use it. Relax this assertion.

Fixes: #1647