things I seem to only notice in my email reader...

Previously, we did not test for the storage being actually freed and
freeing only happened for passes, either explicit or a hit on hfm or
hfp, but not the for the object becoming the hfm/hfp.

We now handle incremental freeing for hfm/hfp as for private objects and
add a test which actually checks that storage is being freed on the go.

Fixes #2653

also polished http_swallow_body() and made the body dump prefix name
more consistent to indicate the receive method:

	(new)		(old)

	chunked		body
	c-l		body
	eof		rxeof

per-VCL, so cleanup should only happen on the last discard event.

This brought confusion to #2576, and while that ticket is probably
easier to debug if the counters were per-vcl, I think it is important
to keep the "per-vmod" behaviour around as an example.

the child process.

Inspired by: #2471

mutex, for instance the VFP might have nipped out for more storage.

Fixes: #2572

On the client side, we impose a total timeout, yet on the client side we
use between_bytes_timeout and do not care about the total (<- we might
want to reconsider this).

Yet HTC_RxStuff only implemented a total timeout, for which we
effectively used first_byte_timeout + between_bytes_timeout. Yet if
first_byte_timeout was not used up, the effective timeout between bytes
could be substantially longer than between_bytes_timeout (initially
analyzed by @daghf).

We now add a duration argument td to HTC_RxStuff which will be used in
addition to (or instead of) the existing total timeout tn. Either td or
tn must be given.

Testcase originally by @fgsch, slightly modified to avoid an assertion
failure in vtc_server due to the connection being closed by varnish.

Fixes #2395

Fixes #2914

Ref: a0436a43

Lines starting with a space are interpreted as a "block quote" and
not ecabulated, this is used for probe details.

Because it looks better.

noté en passant

VOID is only valid as a return type. If VOID was used as an argument,
the C compiler would bail out, yet we should better check this early.

Looking through CTYPES, VOID seems to be the only type not valid for
arguments, thus the explicit check.

Ref 45239c24

Refs 110db401 from the 6.0 branch.

This line ain't too long to be broken.

And update the comment explaining what happens. This is to make it
apparent that a VFP has changed C-E in the stored object, so C-E
should not be changed by an IMS fetch, and a new ETag from IMS has
to be weakened. The flag is appropriate for use by any VFP that
needs to manipulate C-E this way (not just gzip).

Closes #2910

Thank you, @fgsch

Also add a duration argument type and the operators >, >=, < and <=
for use with it (besides the existing == and !=).

obj.ttl and obj.age are compared relative to the time the ban was
issued, as required by the fact that bans are evaluated some arbitrary
time after they are created. For this reason, the ban_dups parameter
has no effect on obj.ttl and obj.age bans, duplicates of bans using
these fields are never removed.

obj.grace and obj.keep are compared as absolute values.

In an effort to support The Proprietary Stevedore (tm) [or any other storage
engine supporting persistent bans - if any?], the existing ban serialisation
format is preserved and upgrades _should_ work seamlessly. But downgrades won't.

Using the BANS_HAS_* macros, we try to make the relationship between the ban
serialisation in cache_ban_build.c and access in cache_ban.c more obvious.

We now generate all fields and operators from table files as well as
the definition of allowed operators for fields. The user help text
about allowed operators is generated at init time.

The second vcl did not actually do anything useful because the second
client would hit a cached response.

as agreed on -commit.

Closes #2906

Refs 832d6da2