Update to 2.1.4, provided by Matthew Sporleder in PR 44165.
Includes new startup scripts.

Summary of changes from 2.1.3 to 2.1.4

    * A bug in the binary heap layout caused inflated object counts, this has been fixed.
    * Much more comprehensive documentation.
    * A DNS director that uses DNS lookups for choosing which backend to route requests to has been added.
    * The client director now uses the variable client.identity for choosing which backend to send a given request to.
    * String representation of now, making it easier to add Expires headers
    * Portability fixes for Solaris.
    * Various bug fixes.

Summary of changes from 2.1.2 to 2.1.3

    * The scalability of critbit, the default hashing method, has been improved.
    * A bug in varnishd would in some cases confuse varnishncsa leading to lost or wrong log lines.
    * Some bugs in the handling of Range requests has been fixed. This only matters if you enable Range support.
    * Add 束log損 command to VCL which will log to the Varnish log.

Summary of changes from 2.1.1 to 2.1.2

    * When adding Range support in 2.1.1 an error was introduced. Garbage was appended to some objects. This affected some load balancers - but clients seemed not to be affected.

Summary of changes from 2.1.0 to 2.1.1

    * Experimental support for the Range header, must be enabled with the http_range_support parameter.
    * A bug in workspace rollback prevented ESI from working correctly in all situations. This is now fixed.
    * A race condition and a deadlock in the critbit hasher have been fixed.
    * HEAD requests are no longer converted to GET requests for pass and pipe.
    * Support for completely obliterating objects including all variants (束nuke損). See http://varnish-cache.org/changeset/4668 for details

Summary of changes from 2.0.6 to 2.1.0

Varnish 2.1.0 has just been released. It contains lots of changes relative to 2.0.6, the most important ones being:

    * Experimental support for persistent cache
    * The regular expression engine is now PCRE
    * Saint mode, where we can serve a cached object if the backend is down or otherwise faulty.
    * A more scalable hashing method called critbit
    * Increased scalability, removing a limitation of maximum 64k connections
    * obj_workspace is removed, this is now scaled automatically
    * Hashing and client IP based directors

Note: The release has some syntaxchanges in VCL, please read the full changelog


(wiz)

#!@RCD_SCRIPTS_SHELL@
#
# PROVIDE: varnishd
# REQUIRE: DAEMON

. /etc/rc.subr

name="varnishd"
rcvar=$name
command="@PREFIX@/sbin/varnishd"

load_rc_config $name
run_rc_command "$1"

#!@RCD_SCRIPTS_SHELL@
#
# PROVIDE: varnishlog
# REQUIRE: varnishd

. /etc/rc.subr

name="varnishlog"
rcvar=$name
command="@PREFIX@/bin/varnishlog"
commands_args="-D"

load_rc_config $name
run_rc_command "$1"

#!@RCD_SCRIPTS_SHELL@
#
# PROVIDE: varnishncsa
# REQUIRE: varnishd

. /etc/rc.subr

name="varnishncsa"
rcvar=$name
command="@PREFIX@/bin/varnishncsa"
commands_args="-D"

load_rc_config $name
run_rc_command "$1"

$NetBSD: patch-ac,v 1.3 2010/12/06 14:25:44 wiz Exp $

--- configure.ac.orig	2010-11-28 15:22:39.000000000 -0500
+++ configure.ac
@@ -30,7 +30,7 @@
    AC_MSG_WARN([xsltproc not found – not building documentation])
 fi
 AM_CONDITIONAL(HAVE_XSLTPROC,[test "x$XSLTPROC" != "xno"])
-AC_CHECK_PROGS(RST2MAN, [rst2man], "no")
+AC_CHECK_PROGS(RST2MAN, [rst2man rst2man.py], "no")
 if test "x$RST2MAN" = "xno"; then
    AC_MSG_WARN([rst2man not found – not building man pages])
 fi