ban.url is confusing as it takes a regular expression rather than a
fixed string, so get rid of it in favour of people being explicit and
using ban req.url ~ /foo

10 loops.  Simplify the structure of the code some more.

Prodded by:	martin

rst2man doesn't support the `glossary` markup, it's a sphinxism.
Avoid using that and sort the list manually instead.

Merge virtualised into a new platform specific notes section. Add transparent hugepages info from ticket 1054.

Original patch by:	martin

don't forget about it so it hangs around with the ban, doing nothing.

(This can happen with /etc/hosts)

Inspired by patch from:  Tollef

Prodded by:	Martin

backend VXIDs to request VXIDs.

Still TBD: where ESI fits into this.

belong to the same batch of VSL records and therefore have the
same vxid.

This will allow us to optimize libvarnishapi, in particular when
doing ordered/transaction mode.

This resolves the race where the pools might accept connections and
start processing them before a lot of the initialization routines has
finished.

it seems edit/readline/readline.h and readline/history.h are
incompatible on FreeBSD, so only include one.  Hopefully this works on
OpenBSD still.

Hi,

The diff below fixes one warning during configure and two during compilation on OpenBSD (and maybe elsewhere):

checking pthread_np.h usability... no
checking pthread_np.h presence... yes
configure: WARNING: pthread_np.h: present but cannot be compiled
configure: WARNING: pthread_np.h:     check for missing prerequisite headers?
configure: WARNING: pthread_np.h: see the Autoconf documentation
configure: WARNING: pthread_np.h:     section "Present But Cannot Be Compiled"
configure: WARNING: pthread_np.h: proceeding with the compiler's result
configure: WARNING:     ## -------------------------------------------- ##
configure: WARNING:     ## Report this to varnish-dev@varnish-cache.org ##
configure: WARNING:     ## -------------------------------------------- ##

varnishadm.c:190: warning: implicit declaration of function 'add_history'

cache/cache_main.c:82: warning: implicit declaration of function 'pthread_set_name_np'

Comments? OK?

f.-

Some minor cleanups, no functional changes.  Thanks to Federico
Schwindt for pointing this out.

When looking at varnishlog it looks like it is send_timeout that was reached,
but in fact it is idle_send_timeout. Clarify the language so the user
understands which parameter to adjust.

The ban.list output would hide the very last ban on the list unless
the lurker debug flag was set. This hides useful debugging information
(how many objects are stuck on the last ban), without giving any
apparent benefit. Remove the hide code.

When using varnishtop -1 on a large log file, the old linked list
approach was very slow.  Using a tree increases performance
significantly.  In a test, old was 15 minutes, 52 seconds, new code
changes that to about 1.8 seconds.

including truncating the length.

Inspired by:  patches and talks from martin

Add a function for comparing ban-strings, mostly to have a place to
put the explanatory comment.

Replace some magic numbers by #defines and segregate the ban-string

Base patch from:	martin
Fixes #1225

this information is no longer available.

Fixes #1228

Remove the XXXAZ(smp_chk_sign) after smp_append_sign(), as it causes
heavy unnecessary CPU usage after each sign update. No assertion
reports have come from this test, so there is no reason to expect the
signs not working.

size of the 'http_req_size' so that we don't run out of WS in case
of client pipelining.

Fixes	#1185

setppriv(PRIV_OFF) succeeds when waiving privileges.

fixes varnish builds on any Solaris OS older than onnv_140 due to
too strict assertions from [c613b135]

add history of solaris privileges

better comments

Submitted by:	Nils Goroll