This lets the request's vfp stack survive a VCL label change.

Fixes: #2367

Spotted by @fgsch

Refs #2362

This is moving some of the generally useful functions of vmod debug to a
new one that is installed for use out of the Varnish source tree. Things
weren't just moved from one source file to another, some functions were
sligthly renamed, and the behavior of `barrier_sync` was changed to
adhere to the "universal fail" and no longer returns a boolean.

While at it, I added missing __match_proto__ macros to vmod debug.

Refs #2276
Closes #2362

When versioning appeared in the VRT API, the goal was to allow loose
ABI compliance on loaded VMODs based on the major/minor revision against
which it was built. Strict checking was performed if Varnish was built
from the master branch in the VCC code, but omitted by the child.

This however has two flaws:

1) Release management might go wrong like it happened in 5.1.2 that got
   released from the master branch.

2) This doesn't solve the original problem that some VMODs might rely
   on supported symbols encompassed by the VRT major/minor while others
   may choose to integrate deeper with Varnish and lose guarantees.

This patch retires the `VCS_Branch` macro that is no longer needed and
provides a new `$ABI` stanza that defaults to strict when omitted. To
help discovery, in-tree modules advertise a strict match.

To indicate that a VMOD needs the exact Varnish build to be loaded,
the VMOD's metadata contains 0.0 for the VRT major/minor revision.

In addition, both the VCC and child now perform the full ABI compliance
check, picking the strict or vrt option depending on the VMOD's metadata.

Closes #2330

This does not include VMOD space, but that is pretty much everything
else.

Refs #2325

Refs #2325

There's already an impedence mismatch between both languages and it's
going to increase soon.

Refs #2325

For that, an indirection via VRT_acl_match is needed, because ACLs are a
special case being functions. However they've had "regular" symbols ever
since VMODs learned to use them.

Refs #2325

Refs #2325

Instead of a printf format, a mandatory prefix is expected.

Refs #2325

Add a barrier sync to ensure we get the WINDOW_UPDATE and the DATA
frames in the expected order.

We could still trigger the integer underflow condition in #2349 by
supplying a padding length equal to the frame size.

Fixes: #2350

In a sequence of HEADERS and CONTINUATION frames, the END_STREAM flag is
set in the HEADERS frame, and the END_HEADERS flag is set in the last
CONTINUATION frame.

Make sure we flag the request as scheduled when dispatched from the h/2
Upgrade code.

Without this, the h/2 code will happily clean it up once the client
hangs up even though it is still being processed in a different thread.

Fixes: #2310

This adds a call to VHT_Fini just before we let go of the h2_sess in
h2_del_req.

Unchecked decoding of the padding length and also the priority flag
turns into an integer underrun when the input length is smaller than
what we end up subtracting.

rfc7540 doesn't seem to adress what should be the error condition when
the priority flag is enabled but the payload is smaller than what could
contain the priority bits. Seeing as a HEADERS frame changes connection
state I assume closing down via PROTOCOL_ERROR is the proper handling.

Fixes: #2349

The use case are cluster requests: Intra-cluster bgfetches should
trigger a synchronous fetch on the peer-varnish in order to avoid
additional short-lived / expired objects being created. Or, in
other words, a bgfetch should actually get a fresh object and not
one in grace from another cache.

Merges #2376

Merges #2381

Fixes #2380.

Fixes #2377.