Also removed the $Id$, which was both incorrect and now obsolete.

- Bump version numbers to 2.1.5
- Redhat: Don't rebuild doc-stuff (it's done at make dist-time)
- Redhat: Include the varnish_reload_vcl script in extra dist

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5761 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5638 d4fa192b-c00b-0410-8231-f00ffab90ce4

The expiry thread caches its "now" timestamp and if there was a lot of
work to do, this timestamp could get behind times, and send the thread
to sleep, despite being behind on work.

Fix this, by updating the timestamp whenever we run out of work.

Add a parameter ("expiry_sleep") to control how long time the thread
will sleep so this can be tuned down on high-load servers.

Inspired by:    sky



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5637 d4fa192b-c00b-0410-8231-f00ffab90ce4

Merge r5582: Add a binheap function to shuffle an item into correct location after a change of key value.

Contrary to the previous comment in cache_expire.c, this process is
guaranteed to work because the shuffle will always terminate either
in the root position or in the bottom row.

Use this function when we adjust ttl on an object.



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5636 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5635 d4fa192b-c00b-0410-8231-f00ffab90ce4

Advertise the CLI protocol version in the banner and start with
version 1.0 to mark support for here-documents.


git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5634 d4fa192b-c00b-0410-8231-f00ffab90ce4

Do not allow here-documents for unauthenticated CLI sessions to
prevent them from becoming an out of memory DoS.



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5633 d4fa192b-c00b-0410-8231-f00ffab90ce4

The last argument to any CLI command can use this feature.

Typical example:

        vcl.inline vcl_new << 42
        backend foo {...}
        sub vcl_recv {...}
        42

The advantage is that no escaping is needed, as long as the magic
marker, in this case "42" does not match any line anywhere in the 
lines that make up the argument.

Arguments encoded this way are not subject to the "cli_buffer"
parameters size limitation.



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5632 d4fa192b-c00b-0410-8231-f00ffab90ce4

Remove call to VSS_parse() in VSS_open(), as VSS_parse() is called in
VSS_resolve() anyway.

Fixes: #817



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5631 d4fa192b-c00b-0410-8231-f00ffab90ce4

Split up the hitrate chapter into four and added a introduction to
ESI. ESI needs a bit of work wrt params and operational factors.



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5630 d4fa192b-c00b-0410-8231-f00ffab90ce4

Split up the hitrate chapter into four and added a
introduction to ESI. ESI needs a bit of work wrt params and
operational factors.



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5629 d4fa192b-c00b-0410-8231-f00ffab90ce4

Suggested by:   Vincent Wells



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5628 d4fa192b-c00b-0410-8231-f00ffab90ce4

"new-purge" cannot and should not touch busy objects, as they are not
subject to refcounting.

Fixes: #812



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5627 d4fa192b-c00b-0410-8231-f00ffab90ce4

One of the silly overgeneralizations in RFC2616, is that headers which
contain comma-separated lists, can be spread over multiple header
lines.

There is no way of knowing if this rule applies to any header not
in RFC2616, short of chasing down the relevant standards document,
if any, for the particular header.
 
Considering the fact that HTTP header lines have no natural
limitation on length AND that RFC2616 already specifies a mechanism
for header-continuation, this doesn't add any value, at all.
 
It is hardly a surprise that nobody used this either, so until now,
we have ignored this silly stuff and just used the first header we
found.
 
But now Chromium, of all things, seems to find it necessary to
spread its Cache-Control across two lines, and we get to deal
with this crap.
 
Add a function for stitching multiple header lines into one, and
call it on Cache-Control in requests to deal with Chromiums issues.

Since we have it, call it preemptively on Cache-Control and Vary
in backend responses, since the C-code examines these fields.

XXX: At some point, add VCL support for collecting specific headers
this way.

Fixes: #686



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5626 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5625 d4fa192b-c00b-0410-8231-f00ffab90ce4

Don't allow silent failure when neither of the following options are
specified:  -d, -b, -f, -S or -T

In human terms, this means that it is legal to start varnishd without
a Vcl or backend, but only if you have a CLI channel of some kind.



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5624 d4fa192b-c00b-0410-8231-f00ffab90ce4

Merge r5518: documented return, restart, reordered some docs for readability and elaborated on saint and grace mode



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5623 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5622 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5621 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5620 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5619 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5618 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5617 d4fa192b-c00b-0410-8231-f00ffab90ce4

Fixes: #789



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5616 d4fa192b-c00b-0410-8231-f00ffab90ce4

Until now, we have not allowed IMS on objects without LM header but
after due consideration of our role as web-server, that restriction is
found too hard: Varnish will, by definition, not find and object which
is not valid, so we can trust the time we put it into the cache to be
the LM date.

But we can not synthesize a LM header based on this, as this would
allow down-stream client-side caches to make unwarranted decisions
(see RFC2616 13.3.4 p88)

Fixes: #795



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5613 d4fa192b-c00b-0410-8231-f00ffab90ce4

Add a magic ${date} macro, which inserts a RFC2616 format timestamp
of the present time.



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5612 d4fa192b-c00b-0410-8231-f00ffab90ce4

We have to check the magic status before other length indications,
otherwise we cannot pass a 304 with a Content-Length.

(RFC2616 p33 4.4)

Fixes: #806



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5611 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5610 d4fa192b-c00b-0410-8231-f00ffab90ce4

Merge r5501: redhat/fedora initscript: Default-Start and Default-Stop for lsb compliance. The defaults are empty.



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5609 d4fa192b-c00b-0410-8231-f00ffab90ce4

Merge r5500: redhat/fedora initscript: Default-Start and Default-Stop for lsb compliance. The defaults are empty.



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5608 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5607 d4fa192b-c00b-0410-8231-f00ffab90ce4

Fixes: #804



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5606 d4fa192b-c00b-0410-8231-f00ffab90ce4

Add distcleancheck_listfiles to work around BSD make wanting to
rebuild to man pages even though they are in the source directory.



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5605 d4fa192b-c00b-0410-8231-f00ffab90ce4

If requests come in fast enough on a single connection, typically
when running synthetic benchmarks, we need to reset the worker threads
workspace between requests, or we will eventually run out.

This is an embarrasingly old bug.

Fixes: #808



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5604 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5603 d4fa192b-c00b-0410-8231-f00ffab90ce4

Make the vcl.7 man page hopefully build with BSD make.



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5602 d4fa192b-c00b-0410-8231-f00ffab90ce4

git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5601 d4fa192b-c00b-0410-8231-f00ffab90ce4