it would seriously miss the point.

Spotted by:	scoof

n_objsendfile is of course no longer relevant.

That makes n_objwrite pointless.

n_objoverflow was unused.

and threads_failed were not incremented where it should be.

the guard.  This makes 'max' parameter mis-sizing much less problematic.

Spotted by:	scoof

unlock/lock sequence when using critbit, at the expense of a
slightly more ornate API for hashers.

Also make the "noh" argument optional, as originally intended.

Needed to test #1123

Fixes: #1123

predictive vary string.

Fixes #1120

Fixes #1125

Varnishtest doesn't see the difference between a VCC core dump and
a VCC error, so a test-case is non-trivial.

Fixes	#1126

This is vastly inspired by #1113 and scoof+martins testcase+patch.

Add a new cnt* state "cnt_restart" to do the mandatory restart work,
and to move the restart limit check out of vcl_recv{}, getting rid
of a nasty XXX comment.

Set the explicit 503 when over limit, reset in other cases.

Expand scoof+martins test-case to cover more vcl methods, which
were also affected, but not fixed in their patch.

Fixes	#1113

The following patch allows the autoconf script to detect the presence of
libedit when there isn't a pkg-config file present and thus allowing
Varnish to detect libedit on OpenBSD/FreeBSD/NetBSD/DragonFly.

Fixes: #1003

I've opted for a less intrusive fix that Martin proposed, but my fix
is largely a rework of his suggested patch.

Testcase by:	Martin

have started to fill it.  Add a param to the STV_trim() method
to indicate acceptability of move.

The only reason I don't remove STV_trim() entirely, is that a site
with gazillions very small objects, will want to disable streaming
and use malloc to squeeze as many objects into memory as possible.

with Req/Resp/Bereq/Beresp/Obj instead.

on everybody agreeing about the order of these fields.