Ref #2896

include coffee.vcl

Noticed for #2929: travis make check failed, but there is no information
as to why, because that is hidden in the test-suite log

Based on patch from @haproxyFred

towards our max-session limit, and by adjusting the counts
at frame rx/tx time.

Fixes: #2923

Thanks to: xcir

Add another column for the probe result

Ref #2896

backend.list now shows "probe" only if a probe is present.

Ref #2896

The sick state of the director and the healthy callback are the same
thing coming from different directions: Either we query the status
dynamically or we already have it.

For layering directors, the health state is determined by their
backends, so having a director-layer sick state does not make much sense
and duplicates logic.

Also, the sick field duplicates admin_health to some extend (see
cache_director.c do_set_health).

This is also relevant in the context of streamlining the backend.list
output: admin_health "probe" only makes sense if backends actually do
have a probe (= some dynamically determined health state). It appears
streaight forward that the presense of a vdi_healthy_f callback is the
signal for dynamically determined health state.

So we move the sick field into VBE and retire VRT_SetHealth().

We also remove the ctx argument from VRT_SetChanged() because I
previously overlooked that it is required, for example, in probe code
where we got no ctx.

Ref #2896

highest open fd by reading /proc/`getpid`/fd directory.

If it fails we fall back to sysconf(_SC_OPEN_MAX) as usual.

as soon as a RST_STREAM has been received.

Testcase by:	xcir

Fixes:	#2923

Implicitly fixes the same issue for VDI_GetHdr

Ref: #2896

by example of the round-robin director

Also changes the backend probe state

	good->healthy
	bad->sick

for consistency

Ref #2899

Ref #2859 #1799

Ref #2859

Ref #1799

In order to avoid the cross product problem with conversion functions
(from x to y types would require x*y functions), we add flexibility
regarding the input parameters to conversion functions:

Each convertion function named after the destination type now takes all
sensible arguments by name.

int is named integer because of the reserved symbol name "int" in C.

All functions should be fully backwards compatible (existing vtcs
continue to work), but compile time checks are now effectively removed.
The conversion functions now trigger vcl errors if used incorrectly or
if conversion errors occur and no failback is provided. However, if a
failback is provided, vcl errors are only raised for usage errors.

For consistency, the conversion functions now only ever truncate if
necessary. std.round() is added for explicit rounding where required.

Existing functions which are now obsolete are marked deprecated.

Ref #2899
Ref https://github.com/varnishcache/varnish-cache/wiki/VIP12:-vmod-polymorphism-(for-type-conversions)

I have reimplemented this based on Nils's #2858, because I found
it too complex and intrusive.  (In particular we try to avoid
unions in Varnish).

Testcase m00051 by:	Nils Goroll

Closes:	#2858

It may have sounded like it's been added for the first time.

the previous code would round 9007199254740991 to 9007199254740992

Tested on linux and the four vtest SunOS variants

vtest would segfault due to a null pointer access on empty
headers